Relative Price Distortions and Inflation: The Case of Argentina, 1963-76

It is doubtful that a simple rule or equation will succeed in explaining short-run fluctuations in Argentine prices and real output. Frequent policy and institutional changes, the complexities of its semi-industrialized economy, sectorial imbalances and rigidities, suggest that nothing less than a comprehensive disaggregated model for the economy can hope to yield reasonably accurate predictions for the future and explanations for the past.1

Abstract

It is doubtful that a simple rule or equation will succeed in explaining short-run fluctuations in Argentine prices and real output. Frequent policy and institutional changes, the complexities of its semi-industrialized economy, sectorial imbalances and rigidities, suggest that nothing less than a comprehensive disaggregated model for the economy can hope to yield reasonably accurate predictions for the future and explanations for the past.1

It is doubtful that a simple rule or equation will succeed in explaining short-run fluctuations in Argentine prices and real output. Frequent policy and institutional changes, the complexities of its semi-industrialized economy, sectorial imbalances and rigidities, suggest that nothing less than a comprehensive disaggregated model for the economy can hope to yield reasonably accurate predictions for the future and explanations for the past.1

The problem of chronic high inflation has plagued a number of countries since 1945, and perhaps the most persistent examples may be found in Latin America. Several of these countries have had annual rates of price increase that have been consistently higher than 10 per cent and have, in certain cases, exceeded 300 per cent. Traditional analysis of inflation fails to explain a number of phenomena peculiar to these countries. Why, for example, in certain cases, has inflation suddenly risen by several hundred per cent, and why is there such a remarkable asymmetry between the rapidity with which an economy enters into a period of high inflation and the slowness with which an economy leaves it?

The analysis of this paper will be applied to the inflations experienced by Argentina during the last fifteen years, when traditional policy tools have often been difficult to use. Corrective monetary policies have been applied, but they have not been sustained, for a variety of reasons. The government deficit has been, on occasion, cut in real terms but has usually been allowed to rise again with little delay; price controls have been tried as a last resort, but they have provided, at best, only momentary relief.

It is the thesis of this paper that once an inflationary process started in Argentina, it set in motion a self-generating mechanism, which must be fully understood by policymakers if they are to stabilize the economy effectively. This mechanism resulted from distortions in the structure of the price system itself—distortions that were brought about, to a large extent, by selective price controls applied during periods of inflation. The inflationary impact of these distortions is realized through their effect upon the level of domestic credit and, hence, upon the money supply. In the following section, we will describe briefly the recent inflationary process in Argentina, summarize the various attempts to analyze this inflation, and then give a general description of our model. In Section II, we will delineate the macroeconomic model that we will use to test the significance of price distortions in explaining inflation, and we will also describe the method used for a quantitative estimation of these distortions. In Section III, we will give the results of the computations of price distortions and estimations of the macro-economic model and will then describe the outcomes of various simulations.

I. Recent Inflation in Argentina and Its Analysis

During the period 1974-76, Argentina experienced rates of price increase that have been matched by only a few countries in recent history. This inflation was not an isolated phenomenon in the country’s recent past but rather a part of a continuing history of rising and subsiding inflationary tendencies that began in 1945. Since the empirical work of this paper deals only with the period 1963-76, we will confine the historical summary of inflation to those years. In 1963, the economy was operating without any significant price controls at an annual rate of inflation of 23.8 per cent, which was approximately the postwar average; but during 1964-65, the government attempted to use selective direct controls to limit the increase in prices. As Table 1 indicates, this program was unsuccessful. It was terminated at the end of 1966. A major reason for its failure was the inability of the government to control the policies of decentralized government agencies and state enterprises. “Labor featherbedding and excessive wage concessions were alleged to be the main causes of burgeoning deficits in these entities, the most notable of which was the state railway deficit that by 1965 accounted for almost 20 percent of total central government expenditures.”2 This financing of public enterprises’ deficits has had, as will be seen, an important impact upon inflation.

Table 1.

Argentina: Annual Increases in Prices, 1963-76

(In per cent)

article image
Source: International Monetary Fund, International Financial Statistics, various issues.

Following a change of government in 1966, an anti-inflation program that was markedly different from earlier stabilization attempts, both in its design and outcome, was undertaken. Restraints were simultaneously applied to both wages and prices, so that there was not a significant decline in real wages; the fiscal deficit was reduced; and the public sector was able to finance a high level of investment almost entirely from domestic savings. During 1967 and 1968, moral suasion and various selective price controls continued to be used, mainly taking the form of attempts to limit increases to identifiable input cost increases; and in 1968 and 1969, the country experienced the lowest rates of inflation in recent history, while at the same time the government continued to reduce its use of price controls as policy instruments.

This period of relatively low inflation came to an end in 1971, however, as relaxed wage policies were combined with expansionary monetary and fiscal programs. In an attempt to slow inflation, several types of direct price controls were instituted toward the end of the year and, in particular, firms were not allowed to pass on increased wage costs to their output prices. These controls were unsuccessful, however, in slowing inflation, and in the following year the rate almost doubled. The government that came to power in 1973 brought with it the so-called Social Contract, one of the main goals of which was to increase the real wage level; and in June 1973, prices for a large number of consumer goods were lowered by 7 to 20 per cent, while wages were raised by approximately 20 per cent. As a result, business and agricultural profits were squeezed; black markets developed; and, by April 1974, sufficient pressure had been created to force the government to relax its price controls. Since, at the same time, wages were allowed to rise, an inflationary spiral began, weakening the government’s fiscal position and aggravating the price distortions, especially for public-sector tariffs, that were already in evidence. The government attempted to halt wage increases, but its efforts collapsed under union pressure, and inflation sharply accelerated. By the end of 1975, the country’s foreign reserves had sharply declined, and the problem of repayment of foreign debt had reached a critical point. In March 1976, the new government removed all price controls, leading to an initial burst of inflation that was largely a result of official prices reaching what had previously been black market levels. The rate of inflation sharply decelerated thereafter, and prices, which had risen by 175 per cent during the first half of the year, rose by only 63 per cent during the second half.

A number of economic models—both theoretical and empirical—have been advanced to analyze the Argentine inflation. The general monetarist model, which has received the most extensive empirical testing, claims essentially the following: Excess demand in the market for goods can exist only if there is excess supply in the market for financial assets, in the sense that the supply of money is greater than the demand for cash balances. In this case, the excess liquidity spills over into the goods market, inflating the nominal value of demand and driving up prices. Thus, excessive wage demands cannot have any long-run inflationary impact unless they are validated by money creation, since they will otherwise lead to unemployment, which, in turn, will moderate the initial wage demands. In order for monetary policy to be effective as a policy instrument, it is necessary for the money supply to be a variable that can be controlled by the authorities and for the demand for real cash balances to be reasonably predictable. The first requirement, that money be exogenous, has been assumed in most empirical work on Argentina until quite recently, and research has, accordingly, focused on the estimation of the demand for money. Dagnino Pastore (1964) found a significant relationship between demand for money, price changes, and income during the first half of the period 1935-60, but his relationship breaks down during the second half of the period. Diz (1970) used the Cagan (1956) formulation for price expectations to explain the changes in real per capita money balances for the period 1935-62, but he arrived at the conclusion that, as the rate of inflation increased, people tended to base their expectations of future inflation more on past than on current rates.3 Relatively similar estimations were carried out by Díaz-Alejandro (1970), who arrived at the general conclusion that the instability of the demand for cash balances presents a serious problem for monetary management.

The fact that the money supply may not always be an exogenous policy variable in a situation of high inflation has been recognized by several authors recently. Their general idea is that, although the authorities may control certain “real” variables (government expenditures in real terms, in particular), the money supply itself is endogenous, because it depends on the price level and hence on the nominal values of these real variables. This idea was first presented by Olivera (1967, 1970) and was tested empirically for Argentina, with satisfactory results, by Dutton (1971). Further work in this direction, applied to Indonesia, has recently been carried out by Aghevli and Khan (1977). Tanzi (1978) has constructed a model in which the endogenous nature of the government deficit, and hence of the money supply, depends to a large extent on lags in tax collection. The policy implications of viewing money as an endogenous variable are quite significant, since the task of controlling inflation via monetary management is complicated by the problem of directly managing the money supply.

An alternative approach considers money to be endogenous because foreign exchange reserves (a component of high-powered money) are endogenous. This monetary approach has not received much attention in econometric studies of Argentina, because it requires the assumption of a small open economy with a fixed exchange rate system. Since, for significant intervals during the sample period of this paper, there have been both floating exchange rates and restrictions on the foreign sector, this assumption is not necessarily supported. The approach here will therefore be to view foreign exchange reserves as exogenous, although this is, admittedly, a simplification that would be valid for only a part of the sample period.

This paper analyzes the Argentine inflation during the years 1963-76 within the context of the monetarist model, taking money to be endogenous. The paper claims that, along with the central government deficit, a major cause of growth in the money supply and of increases in the rate of inflation was losses suffered by businesses; these losses were caused by price distortions, which, in turn, were caused by selective price controls. It should be emphasized that this model may not necessarily be relevant for years either before or after the sample period, when other factors may have been more significant in determining the rate of inflation. As the brief summary of recent Argentine inflation indicates, a succession of governments have attempted to check the rise in prices by means of selective price controls. These controls led to situations in which a particular industry under government control was not allowed to pass along the increased costs of labor or other inputs to its output price and, as a result of rising costs, found itself operating at a loss. These losses would, under normal circumstances, have led to the bankruptcy of firms but, in the Argentine situation, often produced different results.4

There were essentially two cases to consider: loss-making public enterprises and loss-making private ones. In the first case, of which the Argentine National Railroad was perhaps the best example, the government was unwilling to accept the consequences of major public enterprises going bankrupt and accordingly granted credit in order for them to finance their losses. This credit was granted in two ways. The first was by means of direct transfer payments from the central government to the loss-making public enterprises (and would therefore be included in the government deficit), while the second was by means of central bank lending to public enterprises. Both ways had an impact on the money supply, M1, through an immediate increase in the reserve base.5 For loss-making private firms, the financing situation was not so straightforward, however. In general, the enterprises applied for and received loans from commercial banks and thereby avoided going bankrupt. When the banks were able to make these loans out of their reserves or increased savings and time deposits, there was no increase in the reserve base, and hence in M1. Quite often, however, these sources of funds were not available, and the banks instead received rediscounts from the Central Bank, which led to increases in M1. Thus, while one would expect that there would be an almost one-to-one correspondence between increases in the government deficit and increases in the reserve base, an increase in losses caused by price distortions should have led to a smaller rise in the reserve base. The measurement of the magnitude of the impact of these distortion-induced losses upon the rates of change in money and prices is the focus of this paper.

II. The Model

In this section, a simple macroeconomic model will be introduced, and its microeconomic foundations will be described. The macroeconomic model, which will be outlined in Section II.1, has two stochastic equations; the first explains the rate of change in the money supply, and the second determines the demand for real balances. As a system, the model explains, among other things, the time paths of the rate of change in the stock of money and of the rate of inflation on the basis of three basic exogenous variables: the government deficit, the computed aggregate loss of private industries, and the change in foreign exchange reserves held by the Central Bank—all in real terms.6 The macroeconomic model is kept simple, because the main concern of the paper is in illustrating the linkage between money creation and price distortions. Section II.2 introduces a non-stochastic Leontief model and explains the derivation of the industries’ computed aggregate loss resulting from price distortions.7

1. the macroeconomic model

As in Dutton (1971) and Aghevli and Khan (1977), the money supply is endogenous in the model described in this section. Three important factors in the determination of the rate of change in the money supply are identified, and it is shown that these variables are influenced by the rate of inflation.

Assume a simple relationship

Mt=kHt(1)

where Mt, denotes the money supply, Ht denotes high-powered money, and k denotes the money multiplier, which in this study is a constant. The equation can be written in a difference form as

ΔMt=kΔHt(2)

where

ΔMt = Mt - Mt-1, and ΔHt = HtHt-1

There are at least three major factors that normally cause changes in high-powered money: changes in the central bank’s (1) lending to the government, (2) lending to commercial banks, and (3) foreign exchange holdings. During the sample period (1963-76) in Argentina, a sizable portion of the government deficit was often created by government subsidization of public enterprises. Furthermore, a large portion of the central bank’s lending to commercial banks was necessitated by the subsidization of loss-making private firms. A major cause of these losses was selective price controls. Of course, not all the central bank’s lending to commercial banks was subsidies to loss-making private industries; there were a number of other factors that caused changes in high-powered money. In this paper, however, only three factors—(1) the government deficit, a significant portion of which consisted of the government’s current transfer payments to loss-making public enterprises, (2) the computed aggregate loss of private industries, and (3) the changes in foreign exchange reserves—explicitly enter the equation that determines the rate of change in the money supply. A change in high-powered money is therefore expressed as a sum8

ΔHt=ΔHgt+ΔHdt+ΔHbt(3)

where ΔHgt, ΔHdt, and ΔHbt denote changes in high-powered money caused by, respectively, the government deficit, the losses of private industries, and the changes in foreign exchange reserves. Substitution of equation (3) for ΔHt in equation (2) and division by Mt–1 gives

ΔMtMt1=k(ΔHgtMt1+ΔHdtMt1+ΔHbtMt1)(4)

which can be written as

μt=k(hgt+hdt+hbt)(5)

where

μt=ΔMtMt1,hgt=ΔHgtMt1,hdt=ΔHdtMt1,andhbt=ΔHbtMt1(6)

The three ratios, hgt, hdt, and hbt, may not be observable; to express the rate of change in the money supply as a function of observable variables, it is further specified that

hgt=a1gmt,hdt=a2dmt4,hbt=a3bmt(7)

where

gmt=GtMt1,dmt=DtMt1,bmt=BtMt1

Gt = deficit of the government

Dt = computed aggregate loss in private industries

Bt = change in foreign exchange reserves

All variables Gt, Dt, and Bt are measured in current pesos. The government deficit Gt includes the government’s current transfer payments to the public enterprises; the total deficit Gt is therefore the sum of the deficit Gdt incurred by the government’s subsidization of the public enterprises and the deficit Ggt incurred by the other activities of the government. The computed aggregate loss Dt of private industries is the computed total of losses in both public and private sectors minus the current transfer payments of the government to public enterprises.9 These losses were primarily caused by the price distortions that resulted mainly from selective price controls and were largely financed by loans made through the banking system. There are a number of reasons why a significant time lag may have existed between the realization of losses by the private firms and the subsequent creation of money: the banking system may not have fully responded instantaneously to the higher demand for funds; furthermore, the actual need for additional funds may have lagged because of the practice of delayed payments by loss-making firms. The four-quarter lag introduced into the second equation in (7) reflects this relationship.10 The equation that explains the rate of change in the money supply, and that will be referred to as the money supply equation, is therefore

μt=α0+α1gmt+α2dmt4+α3bmt+εt(8)

where all variables are quarterly rates or averages.11 As indicated earlier, the real deficit of the government, the computed real aggregate loss of private industries, and the change in foreign exchange reserves in constant pesos—denoted, respectively, by gt dt, and bt—are the exogenous variables of the system.12 Three additional identities should, therefore, be introduced to complete the model’s subsection that explains the rate of change in the money supply

Gt=gtPt,Dt=dtPt,Bt=btPt(9)

where Pt denotes the general price level.

The real balance equation is the one Cagan (1956) proposed for hyperinflationary economies

nmtD=b0+b1πtE+ηt(10)

The level of real balances the public desires to hold is denoted by mtD, the rate of inflation anticipated by the public by πtE, and the error term by ηt. By introducing the usual adaptive expectations scheme

πtEπt1E=θ(πtπt1E)(11)

the real balance equation can be transformed into

nmtD=(1λ)b0+λnmt1D+θb1πt+ηtληt1(12)

where λ = 1-θ. Equation (12) can also be written as

nmtD=β0+β1nmt1D+β2πt+ωt(13)

where

β0=(1λ)b0,β1=λ,β2=θb1,ωt=ηtληt1(14)

The equilibrium condition

mtD=mt(15)

where mt denotes the stock of real balances (mt = Mt/Pt), closes the system, which can be written as follows:

μt=α0+α1gmt+α2dmt4+α3bmt+εt;α1,α2,α3>0nmtD=b0+b1πtE+ηt;b1<0μt=MtMt1Mt1,gmt=GtMt1,dmt=DtMt1,bmt=BtMt1(16)Gt=gtPt,Dt=dtPt,Bt=btPtπtEπt1E=θ(πtπt1E);0<θ<1πt=PtPt1Pt1,mtD=MtPt

The system in (16) simultaneously determines the rate of change μt in the stock of money and rate πt of inflation. This property of the system is more apparent if the model is reduced to a three-equation system. Note that

gmt=gtmt1(1+πt),dmt=dtmt1(1+πt),bmt=btmt1(1+πt)(17)

where gt, dt, and bt denote, respectively, the government deficit, the computed aggregate loss of private industries, and the change in foreign exchange reserves in real terms; mt, denotes the stock of real balances, and πt denotes the rate of inflation. The rate of growth in the money supply can therefore be expressed as a function of the three basic exogenous variables of the system, the lagged real balances, and the rate of inflation. Take the first difference of the equation and rearrange the terms. If the equilibrium between demand for, and supply of, money is always maintained, the two stochastic equations in (16) can be written as

μt=α0+(α1gtmt1+α3btmt1)(1+πt)+α2dt4mt5(1+πt4)+εt(18)
n(1+πt)+β2πt=β1n(1+πt1)+β2πt1+n(1+μt)+β2n(1+μt1)+et(19)

where et is a non-invertible moving-average error term

et = – Δ(ηt – λ ηt–1)

Additionally, there should be an identity for the level of real balances

mtmt1=1+μt1+πt(20)

The system is nonlinear and dynamic. The rates of increase in money and prices are jointly determined through time by the time paths of the real government deficit, the computed aggregate loss of private industries, and the change in foreign exchange reserves. Higher paths of these exogenous variables would result in higher paths of the rates of change in both money and prices. The system estimated in Section III is stable; a temporary change in any exogenous variable or parameter would have significant impacts on the two basic endogenous variables for only a limited period of time.13

2. computation of price distortions

In the previous sections, references have been made to zero-profit prices and to distortions of actual market prices. At this point, it is still not clear, however, what should be used to represent these zero-profit prices. Are they to be some unobserved “true” values, as are used in many econometric studies, or are there to be independent numerical estimates of them? Clearly, the second method would be preferable, since it would permit computation of a zero-profit price structure that would be independent of the method of estimating the monetary equations and would thus avoid the problems of introducing a priori assumptions about price formation. In order to compute zero-profit prices for an economy, one could construct a full general equilibrium model that postulates the supply of goods derived from production functions and that assumes the existence of groups of consumers, each of which behaves according to a well-behaved utility function. Such a model, while introducing certain assumptions about competitive behavior, would permit a direct computation, via a method such as the Scarf fixed-point algorithm, of market-clearing prices.14 In our case, however, such a model, while technically feasible, would not yield the type of results desired because of lack of data. We therefore use an input-output system, which is a special case of the general activity analysis model, to derive zero-profit prices. Because of the special characteristics of input-output matrices, it is possible to compute these prices solely by examining the production side of the economy; demand serves only to determine the levels of production.15

Argentina has published a series of input-output matrices of its aggregate production technology, the first constructed for 1950 with new ones being constructed for 1953 and 1963. The technical coefficients derived from the 1963 matrix were revised in 1970, so as to incorporate the changes that had taken place in the intervening years.16 As is typical of most input-output matrices that have been derived from national income accounts, those of Argentina were obtained by using actual cash flows rather than physical units. We apply a standard methodology to derive a set of underlying physical units, so that direct comparisons may be made between the 1963 and 1970 matrices. For the purposes of this study, this direct comparability is important, since it is necessary to construct input-output matrices to represent the technology of each of the years between 1963 and 1976. To do so, the assumption is made that technology changed in a smooth linear fashion during the 1963-70 period and continued to change at the same rate after 1970. Thus, if the years 1963–76 are represented by s and if the input-output technology for year s is represented by As, then:

As=7(s1963)7A1963+s19637A1970(21)

where A1963 and A1970 are the 1963 and 1970 input-output technical coefficient matrices, respectively.

This scheme produces 14 input-output matrices of technical coefficients As, s = 1963, …, 1976, representing the years 1963-76, each of which has been determined by using physical quantities based on 1963 prices. For the purpose of price computation, the total value added per unit of output in each sector is also needed, where output is measured in 1963-based units. The national income accounts give values added in current pesos for each of the years 1963-73 for each of the 23 sectors that appear in the input-output matrices. Rather than the total value added, the portion that is represented by the wage bill is used, because data on total value added are not available for 1974-76. Thus, by making various interpolations, quarterly values added for 1964-76 are obtained. If cjt denotes the technical coefficient of the value added per unit of output for the jth sector in quarter t, vajt denotes the nominal value added17 of the sector, xj0 the sector’s output in 1963, and rjt the rate of growth in real output of the sector from the fourth quarter of 1963 to quarter t, then

cjt=vajtxj0(1+rjt)(22)

The zero-profit prices for each quarter are now computed as

pt=ct(IAt)1(23)

where pt denotes a 1 × 23 vector of prices for quarter t. Thus, these prices are those at which each sector in the economy is operating at zero profit. Since we are assuming that the technology matrix changes only yearly, we will use the same At matrix for all quarters during a year.

Clearly, the prices for a given quarter are dependent upon the physical units chosen as the base for the computations. What is not dependent on units, however, are the rates of change in prices from quarter to quarter; our interest, therefore, is not in the computed prices themselves but in indices of prices based on 1963. For a given commodity, each figure in the series of 52 prices representing its computed zero-profit values for each of the quarters in the years 1964-76 is divided by the commodity’s 1963 price to arrive at a corresponding series of index numbers based on 1963 prices. These index numbers then represent changes in the zero-profit prices of each commodity over the period in question. The index numbers of the zero-profit prices will be represented by pte=(p1te,,p23te) for quarter t.

The key element in the investigation of price distortions now comes with the assumption that 1963 was a year in which prices were at equilibrium levels, in the sense that there were no excess demands and that the profit made by each industry was “normal,” that is, it was sufficient to cover the cost of required investment. This assumption must be based on the rather subjective argument that the rate of inflation was a relatively low 24 per cent in 1963 and that there were no significant restrictions enforced in the economy.18 For our purposes, this is equivalent to assuming that the actual prices were equal to the computed zero-profit prices in that year. By renormalizing, price indices based on 1963 are constructed out of the official indices, which are based on 1960 prices and include indices for 17 of the 23 sectors that appear in the input-output matrices. The difference between the actual and computed price indices for a particular commodity in any quarter of any year may then be used as a measure of the distortion in the price of that commodity. Thus, if, for example, a commodity’s computed price index is higher than its actual price index in some period, one could conclude that its price had not risen sufficiently since 1963 to compensate for the changing technological and cost structure of the economy.

It is possible for all price indices to be below their zero-profit levels if all prices are controlled, while wages, and hence values added, rise at the same time. Because zero-profit prices depend solely upon the technology and the real values added, which are exogenous in the model, the actual relative prices, if they are controlled, may be systematically lower than the zero-profit prices deflated by the actual general price index.

The Argentine input-output matrices consist of 23 sectors.19 In each sector, value added in real terms is taken to be an exogenous variable, determined by technology and political forces; so, given the values added for a particular quarter, the zero-profit prices are the unique prices at which the input-output matrix produces positive outputs. Prices different from these would cause some sectors to make a profit on each unit of output, while other sectors would make a loss. In the first instance, the sector would operate at an infinite level, while in the second, the sector would not operate at all because the model is based on the assumption of constant returns to scale. Hence, each sector must have received the zero-profit price for its output and have paid the zero-profit prices for its inputs. Since the actual prices are different from these, how is the discrepancy to be explained? If a sector’s actual output price is lower than its computed price, then, we claim, the difference between the two must be made up by a loan, in order to bring the effective sales price up to the required level. Conversely, if the sector’s actual output price is higher than its Leontief level, then the difference would have to be collected as a tax, in order to bring the effective sales price down to the required level. Hence the difference, (pjtepjta), between the computed and actual prices at time t represents, if it is positive, the unit loan for20 (or, if it is negative, the unit tax on) the jth sector that is brought about by price distortions.

Thus (pjtepjta)xjt, where xjt represents the gross output of the jth sector at time t, represents the total loss or profit of the jth sector at time t.21 By summing over all positive distortions, that is, all sectors for which (pjtepjta)>0, the aggregate loss caused by price distortions at time t is obtained. Hence the total loss Dppt induced by distortions is

Dppt=Σ(pjtepjta)xjt(24)

where the summation is taken over all j such that pjte>pjta.

These losses enter into the money supply equation described in the previous section as a measure of the amount of financing required at time t because of price distortions.22 The question arises as to why negative distortions, which theoretically represent tax collections, are not subtracted from total losses to arrive at a net figure for credit creation. We claim that the monetary impact of the distortions is asymmetrical, in the sense that positive distortions actually lead to borrowing, while negative distortions are not, in reality, taxed away, but instead lead to profits. The sectors that make these profits do not produce at infinite levels, however (as they would in a model with constant returns to scale), because of capacity constraints and market conditions that have not been explicitly introduced. Thus, negative distortions, representing profits, do not decrease the money supply but rather have no effect on it.

III. Empirical Results

This section will be divided into three parts, the first of which will contain the results of the computation of price distortions; the second will give the estimations of the monetary equations; and the third will be concerned with the various simulations that have been made with the empirical model.

1. computation of price distortions and losses

On the basis of the methodology described in Section II, quarterly series of indices of zero-profit prices based on 1963 were derived for each of the 23 sectors in the Argentine input-output matrices. These computed indices were then compared with the actual indices. The two series deflated by the wholesale price index (WPI) are shown in Charts 1 and 2 for six selected industries. The industries shown were chosen to be representative of certain general types. Thus, agriculture has been, during the periods of price controls, perhaps the most severely restricted part of the private sector, while the food industry has been less closely controlled. Leather goods are export oriented and have not been subject to many price restrictions, while the construction industry has been highly competitive and has set prices relatively freely. Fuels and utilities (electric power and light) are typical of large public enterprises whose employment and pricing policies are regulated by the government.

Chart 1.
Chart 1.

Argentina: Actual and Computed (Zero-Profit) Price Indices Deflated by Actual General Price Level, Selected Industries, 1965-761

(1963 = 100)

Citation: IMF Staff Papers 1978, 003; 10.5089/9781451972559.024.A003

1 Both actual and computed price indices are deflated by the wholesale price index (WPI). Since the computed price indices depend only upon technology and values added, they can be systematically higher than the corresponding actual price indices if the output prices are controlled and values added are allowed to rise.
Chart 2.
Chart 2.

Argentina: Actual and Computed (Zero-Profit) Price Indices Deflated by Actual General Price Level, Selected Industries, 1965-761

(1963 = 100)

Citation: IMF Staff Papers 1978, 003; 10.5089/9781451972559.024.A003

1 Both actual and computed price indices are deflated by the wholesale price index (WPI). Since the computed price indices depend only upon technology and values added, they can be systematically higher than the corresponding actual price indices if the output prices are controlled and values added are allowed to rise.

Agricultural prices have been continuously held below their zero-profit levels, with the difference becoming most pronounced during 1973-75, the period of the Social Contract.23 The prices of agricultural commodities rose rapidly, following the elimination of controls in March 1976, with the result that, by the end of the year, the agricultural sector’s downward distortions had been almost entirely removed. The food industry was also subject to downward distortions in its price structure, although not such severe ones as agriculture’s; but during 1976, these losses were more than compensated for, and the actual price of output in the sector rose beyond its zero-profit level. The two least-controlled sectors of the sample—leather and construction—had prices that from 1964 to 1975 moved very closely in line with their zero-profit prices and that increased to such an extent following the end of price controls in 1976 that both industries were operating at a significant rate of profit. Fuel and utilities, the two public enterprises in the sample, both suffered from downward price distortions during most of the sample period; but in the case of fuel, the distortions had largely been eliminated by the end of 1976, while utility prices still remained considerably below their zero-profit levels.

In general, the 17 sectors for which actual and zero-profit prices could be compared showed significant but relatively constant (in real terms) downward distortions in aggregation for the years of the sample period up to approximately 1971. The aggregate downward distortions increased sharply in the period 1973-75 and declined rapidly thereafter, and had been practically eliminated by the end of 1976. Chart 3 gives the total computed losses owing to distortion in real terms for each quarter during the sample period. The computation of the aggregate nominal loss, Σ(pjepja)xj, where the summation is taken over all loss-making industries, is described in Section II.2. Real losses are derived by deflating the nominal losses by the wholesale price index. It is often claimed that inflation, in and of itself, can cause relative price distortions, but the results for the last half of 1976, when there was a continuing high rate of price increase but a virtual disappearance of distortions, indicate that, given the absence of price controls, inflation alone would not have created significant downward distortions.

Chart 3.
Chart 3.

Argentina: Exogenous Variables, 1965-76

Citation: IMF Staff Papers 1978, 003; 10.5089/9781451972559.024.A003

2. estimation of the macroeconomic model

The two stochastic equations to be estimated in the model are

μt=α0+α1gmt+α2dmt4+α3bmt+εtnmt=β0+β1nmt1+β2πt+ωt(25)

where

μt=MtMt1Mt1,πt=PtPt1Pt1gmt=gtPtMt1,dmt=dtPtMt1,bmt=btPtMt1,andmt=MtPt.

The error term ∊t is assumed to be an independently and identically distributed (i.i.d.) random variable,

E(εt)=0E(εtεt)=σε2fort=t=0otherwise(26)

The real balance equation in (25) is derived from the original real balance equation specified in (10), and the coefficients β0, β1, and β2 are related

Table 2.

Argentina: Quarterly Averages of Major Variables in Inflation (Actual Observations)1

article image

The growth rate, or rate of increase, in a variable xt, is defined as 100 (xt - xt-1)/xt-1. Variables in real terms are measured in 1963 pesos. The total amount of current transfer payments to the public enterprises is interpreted as a reasonable estimate of the aggregate loss of the public sector. The computed aggregate loss dt of the private sector is derived by subtracting the estimated loss for the public sector (a total amount of the current transfer payments) from the computed aggregate loss of the whole economy (described in this section).

to the coefficients of the original equation—b0, b1, and θ—by

β0=(1λ)b0,β1=λ,β2=θb1,andλ=1θ(27)

as indicated in (14). The error term ωt is related to the error term of the original equation ηt by

ωt=ηtληt1(28)

where the error ηt, is also assumed to be an i.i.d. random variable,

E(ηt)=0E(ηtηt)=ση2ift=t=0otherwise(29)

The historical series of the variables are derived as follows:

(1) All the flow variables in nominal terms (the government deficit, the computed aggregate loss of the private sector, and the change in foreign exchange reserves) are derived as quarterly aggregates. (2) The quarterly average series of the monthly wholesale price index centered in midquarters is used as the series for the general price level. (3) The nominal flow variables are deflated by the price index to derive the flow variables in real terms. (4) The quarterly series of the stock of money is derived by averaging the monthly M1 series with midquarters as centers.

The real variables (gt, dt, and bt) derived in step (3) are assumed to be exogenous in estimation. That is, it is assumed that

E(xtut)=0foralltandt(30)

where xt = gt, dt, bt and ut′ = εt′, ηt′

The estimation method is an instrumental variable (IV) method. Although any instruments uncorrelated with the error terms would give consistent estimators under standard conditions, distributed lags of the exogenous variables are used as instruments in this paper.24

The estimated equations are

μt=0.003(0.23)+1.103(18.77)gmt+0.157(2.14)dmt4+0.584(5.15)bmt+ε˜t(31)R2=0.936SE=0.034DW=1.62
nmt=0.319(5.65)+0.804(20.83)nmt10.258(6.62)πt+0.182(4.54)dmt+ω˜t(32)R2=0.930SE=0.068DW=0.65

where the figures in parentheses are t-ratios. Equation (31) has the estimates of the coefficients with anticipated signs and significant t-ratios.25 To examine the magnitudes of the coefficients, the estimated non-stochastic part of the equation can be written as

ΔMt=0.003Mt1+1.103Gt+0.157Mt1Mt5Dt4+0.584Bt(33)

by multiplying the equation through by Mt-1. It is clear from equation (33) that the money supply would increase by 1.103 billion pesos if there were an increase in the nominal government deficit of 1 billion pesos. The impact on the money supply caused by the computed aggregate loss of the private sector cannot be determined, however, unless the rate of growth in the nominal money supply during the preceding year is known. For example, an increase in the money supply of 300 per cent during the year ending in the quarter t–1, combined with an aggregate loss of one billion pesos in the quarter t–4, would lead to an increase of 0.628 billion pesos in the money supply in the quarter t. The impact of the losses on the money supply increases as the rate of growth in the money supply rises, because of the time lag between the occurrence of the losses and the subsequent creation of money. With a higher rate of inflation associated with a higher rate of growth in money, the nominal amount of credit creation required to subsidize a given amount of losses incurred during a quarter that precedes by one year the quarter during which the subsidies are given should increase along with the money supply.26 The results of estimations indicate that the government deficit was the most important of the three exogenous variables in affecting monetary expansion and that the other two factors—subsidization of the losses in the private sector and changes in external reserves—were also highly significant.

In estimating the real balance equation, a dummy variable dmt is added because of the real balances that were conspicuously high in comparison with the levels that would be anticipated in view of the rate of inflation during the third and fourth quarters of 1973, the fourth quarter of 1975, and the third quarter of 1976. The estimated coefficients of equation (32) imply the following coefficient values of the original equation (10):

b˜0=1.628b˜1=1.316θ˜=0.196

The original real balance equation implied by the estimated equation(32), therefore, is

nmtπtEπt1E=1.6281.316πtE+η˜t=0.196(πtπt1E)(34)

The estimated coefficient b˜1 of πtE has the right sign, as does the estimated coefficient θ˜ of price anticipation.

The Durbin-Watson (D-W) statistic for the real balance equation (32) is very low (0.65), but a low D-W should be anticipated in this case, because the error ωt, t = 0, ± 1, ± 2,…, is a first-order moving average (MA) process with a positive parameter. Since the estimators for the coefficients β0, β1, and β2 are consistent, an examination can be made to determine whether the D-W statistic suggests an estimate of the MA parameter compatible with the one indicated by the estimate of β1; which is identical with λ, the MA parameter of the error ωt. (See equation (28).) The results are compatible, at least in their signs, since the D-W statistic suggests a positive value of λ˜.27

3. results of simulations

Several long-run dynamic simulations are conducted for the period extending from the first quarter of 1967 to the fourth quarter of 1976 (1967 I-1976 IV) with various assumptions on the time paths of the exogenous variables to evaluate the model’s capability to describe the historical path of the rate of inflation and to evaluate possible impacts of various policies on inflation. The system described in (16) with the parameter estimates reported in equations (31) and (32) produces quite realistic rates of growth in money and prices in a deterministic long-run dynamic simulation based on the historical series of the exogenous variables for the period 1967 I-1976 IV.28 The results of the dynamic simulation based on the historical series of the exogenous variables are summarized in Chart 4 and Tables 3 and 4. Table 3 compares the actual quarterly averages of the basic endogenous variables with the quarterly averages of the simulated values for the periods 1967 I-1972 IV, 1973 I-1976 IV, and 1967 I-1976 IV. The model simulates the rates of increase in money and prices fairly well. The simulation gives higher average rates of change in money and prices for 1973 I-1976 IV than for 1967 I-1972 IV, as actually happened. As shown in Chart 4, the downward movements of the rate of increase in money during the periods 1967 II-1969 IV, 1973 IV-1974 III, and 1976 I-1976 IV are correctly simulated, as well as the upward movements during 1970 I-1973 III and 1975 II-1975 IV. The model also reproduces the relatively high rates of inflation during 1971 I-1973 II and 1975 IV-1976 II, as well as the relatively low rate of inflation during 1976 III—1976 IV. A notable exception is the third quarter of 1975, when the simulated rate of inflation is far below the actual rate. The unusually high rate of inflation during the quarter was probably a consequence of the collapse of the stabilization program introduced in mid-1975. Correlation coefficients between the actual and the simulated series of some of the endogenous variables and root-mean-squared errors are reported in Table 4, which shows that the simulation of the rate of growth in money is consistently more accurate than the simulation of the rate of inflation. The estimated period means and the standard deviations for the basic endogenous variables are also reported in Table 4 for comparison with the root-mean-squared errors.29 The results of the simulation described above confirm that the model is dynamically stable. A further examination of some of the properties of the model is conducted by recording the dynamic paths of adjustments caused by temporary shocks to the system. The results show that the impacts of these temporary shocks taper off rapidly with oscillations.30

Chart 4.
Chart 4.

Argentina: Results of Long-Run Dynamic Simulation with Historical Series of Exogenous Variables, 1967-76

Citation: IMF Staff Papers 1978, 003; 10.5089/9781451972559.024.A003

Table 3.

Argentina: Quarterly Averages of Major Variables, Actual and Simulated1

article image

All variables are defined in the same way as in Table 2. The results are based on a dynamic simulation for the period extending from the first quarter of 1967 to the fourth quarter of 1976. Actual historical values are used for all exogenous variables.

Table 4.

Goodness of Fit in Simulation1

article image

The results are based on the same dynamic simulation as in Table 3. The units of measurement for the variables in 2, 3, and 4 are the same as in Table 2.

Chart 5 and Table 5 summarize the results of three policy simulations based on (1) the historical series of the real government deficit, (2) a real government deficit that includes only the current transfer payments (i.e., the subsidies to public enterprises) and (3) a completely balanced budget. All simulations are conducted for the entire simulation period 1967 I-1976 IV, although the results are reported for 1967 I-1972 IV and 1973 I-1976 IV, as well as for the entire period. The simulated average quarterly rate of inflation for the period 1967 I-1976 IV is reduced to 4.76 per cent from 17.09 per cent during the entire simulation period when all the sources of the government deficit other than current transfer payments to public enterprises are assumed to be zero. A balancing of the government budget further reduces the simulated average rate of inflation to 3.68 per cent for the 10-year period. As indicated in Chart 5, the relative importance of the current transfer payments to public enterprises was significantly reduced during the second half of the simulation period because of the rapid increase in the government deficit, which was generated mainly by sources other than transfer payments.

Chart 5.
Chart 5.

Argentina: Policy Simulations Based on Different Assumptions on Government Deficit in Real Terms, 1967-76

Citation: IMF Staff Papers 1978, 003; 10.5089/9781451972559.024.A003

Table 5.

Argentina: Policy Simulations Based on Different Assumptions on Government Deficit in Real Terms1

article image

All variables are defined in the same way as in Table 2. All of the controlled simulations are conducted for the whole simulation period, which extends from the first quarter of 1967 to the fourth quarter of 1976.

Chart 6 and Table 6 summarize the results of simulations based on (1) the historical series of the computed aggregate losses of the public and private sectors, (2) zero current transfer payments to the public sector (no price distortions in the public sector, or no subsidization of public enterprises), (3) zero losses of the private sector, and (4) the absence of both current transfer payments and of losses of the private enterprises. The average simulated quarterly rate of inflation for the period 1967 I-1976 IV decreases from 17.09 per cent to 14.12 per cent when condition (2) is imposed and to 10.29 per cent and 8.54 per cent as the conditions (3) and (4), respectively, are imposed on the simulations. As Chart 6 shows, and as equation (33) predicts, the inflationary impact of losses owing to distortions tends to be greater during periods of rapid inflation than during periods of stable prices. Thus, for example, a complete elimination of price distortions reduces the simulated rate of inflation in the first quarter of 1976 from 96 per cent to about 43 per cent, while during the first quarter of 1969, a complete elimination of price distortions reduces the simulated rate of inflation by only 3 per cent.

Chart 6.
Chart 6.

Argentina: Policy Simulations Based on Different Assumptions on Computed Aggregate Losses in Real Terms, 1967-76

Citation: IMF Staff Papers 1978, 003; 10.5089/9781451972559.024.A003

Table 6.

Argentina: Policy Simulations Based on Different Assumptions on Computed Aggregate Loss in Real Terms1

article image

All variables are defined in the same way as in Table 2. All of the controlled simulations are conducted for the whole simulation period, which extends from the first quarter of 1967 to the fourth quarter of 1976.

The simulations confirm the findings that were briefly discussed on the basis of the results of the estimation in the previous section; the government deficit in real terms has probably been the most important cause of the monetary expansion and inflation during the sample period, but the simulations indicate that the distortion-induced subsidization of public and private firms has also been an important cause of the monetary expansion. For example, a complete elimination of the causes of the subsidization (price distortions) reduces the simulated quarterly average rate of monetary expansion to 17.90 per cent (from 30.62 per cent) during 1973 I-1976 IV, and to 8.55 per cent (from 16.01 per cent) during 1967 I-1976 IV. The simulated quarterly average rate of inflation is reduced to 19.66 per cent (from 34.77 per cent) during 19731-1976 IV and to 8.54 per cent (from 17.09 per cent) during 1967 I-1976 IV. A quarterly rate of inflation of 19.66 per cent is still quite high, but the size (15.11 per cent) of the decrease in the rate of inflation is also significant. Furthermore, the simulated quarterly rate of inflation of 5.86 per cent, or 25.58 per cent at an annual rate, which is obtained with the assumption of a balanced government budget, suggests that the simple balancing of the government budget may not have been a sufficient condition for stable prices. Therefore, only a completely balanced government budget and the elimination of distorted price structures would probably have created a policy environment that would have been more favorable to an effective management of the monetary aggregate.

IV. Summary and Conclusions

This paper has constructed a macroeconomic model within which the impact of price distortions on the rate of inflation may be analyzed. These distortions are computed by using input-output estimates of average zero-profit prices for each major industry and observing the deviations of actual prices from the zero-profit prices. The model is applied to Argentina for the period 1963-76, and its hypotheses are supported by the estimation results. It is found that price distortions increased dramatically during the period of selective price controls, especially during 1973-75, and that they had a significant impact on the rate of monetary expansion. This impact, which was realized via the credit that was created when distortion-induced losses were financed through loans made by the banking system, increased in proportion to other causes of inflation as inflation accelerated. The central government deficit—the other cause of domestic credit creation in the model—is found to have had a greater impact on inflation than did price distortions. This result is plausible, since there was an almost one-to-one correspondence between deficit financing and increases in high-powered money, while losses owing to price distortions were partially financed through the commercial banking system and thus had a lesser impact on high-powered money.

A number of simulations are carried out with the estimated model to determine the effects of various policies. It is found that, during the period of highest inflation, 1973-76, a complete elimination of all price distortions reduces the simulated average annual rate of inflation from 230 per cent to 105 per cent. If, on the other hand, price distortions are retained at their historical levels while the central government budget is assumed to be balanced, the simulated annual rate of inflation is reduced to 26 per cent; it is reduced to 32 per cent if it is assumed that subsidies of public enterprises are maintained at their actual levels. Since the simulated rate of inflation obtained is still high, even after a complete elimination of the deficit, it may be concluded that only the elimination of price distortions, combined with a balanced government budget, would have led to stable prices.

If price distortions had been allowed to continue and, at the same time, an attempt had been made by the Government to check their inflationary impact by refusing to grant financing to loss-making firms, the result would have been widespread business failures. Thus, applying credit policies to alleviate the expansionary pressures caused by price distortions, without eliminating those distortions at the same time, would have led, in the long run, to a recessionary situation. If the authorities had attempted to hold to a particular set of selective price controls as inflation continued, distortions would have increased and would have led to yet more inflationary pressure. Thus, price controls, in and of themselves, could bring about a self-generating inflation.

APPENDIX

Computation of price distortions

The technology in the model is represented at time t by an n × n input-output matrix At = [aijt], each column of which represents the nonlabor inputs used in a particular production activity of the economy. Since the assumption is made that each activity produces one, and only one, output (no joint production), the technical coefficients [aijt], i = 1, … n in the jth column represent the physical inputs of each produced commodity that are required to turn out one unit of the jth commodity at time t. There is one non-produced input to production—labor, of which the jth column (or production activity) requires an input vajt to produce one unit of good j at time t (where va denotes the vector of values added). Suppose further that the prices for all goods are normalized, so that the price of labor is 1 and the prices of the other commodities are [pit], i = 1,…, n at time t. The total cost of operating the jth activity at unit level at time t is then given by Σi=1npitaijt+vajt. Since a condition for equilibrium in an economy characterized by constant returns to scale is that each activity must make a nonpositive level of profit, the following inequality must hold for each, j = 1, …, n:

pjtΣi=1npitaijt+vajt(35)

Since there is only one activity producing each commodity, and since we shall assume that there is positive demand for each good, each activity must make a nonnegative profit, as a loss-making activity would not be operated. Thus, at equilibrium, the equation becomes

pjt=Σi=1npjtaijt+vajtj=1,,n(36)

or, in matrix notation,

pt=ptAt+vat(37)

where pt = (p1t,…, pnt)

At=[aijt]i,j=1,,nvat=(va1t,,vant)

Rearranging terms, we obtain

pt(IAt)=vat(38)

where I denotes the n × n unit matrix, and thus

Pt = vat (I – At)-1

The nonsubstitution theorem31 guarantees that there will be no substitution of production activities, so the technology completely determines prices while demand serves only to determine the level of production.

In computing the total loss caused by price distortions in equation (24), use was made of xjt, the gross output of the jth sector at time t. In order to have consistent units of output, certain modifications of the national income accounts data are required. Suppose that x¯jt represents the output, in constant 1963 pesos, of the jth commodity at time t. The Leontief price pjt, as computed in equation (23), then is the price of one unit of good j at time t, where these units are measured in constant 1963 pesos. Define an index by

pjte=pjtpj0(39)

It follows that pjtx¯jt=pjtepj0x¯jt. Define xjt by xjt=pj0x¯jt, to arrive at units of output that are consistent with the normalized computed and actual price indices.

It is important to analyze fully the general equilibrium implications of price distortions on a theoretical level, since doing so leads to a better understanding of both the direction of income flows in this model and of the ways in which a noninflationary solution to the problem of price distortions could be found. Suppose that all units of production have been translated so that they are consistent with the 1963-based price indices. If y = (y1,…,y23) is the vector of final demand in a particular quarter, and if x = (x1,…,x23) is a vector the components of which are the activity levels of each of the 23 sectors, then the condition for market clearing is

xAx=y(40)

where A denotes the technology in the particular period.

Assuming that the input-output matrix A is such that at least one strictly positive output vector can be produced—a condition that will be satisfied by any empirically derived input-output matrix—a positive solution to this equation will always exist.32 Suppose now that, given these final demands and production levels, prices pa=(p1a,,p23a) are charged for each commodity, where pa represents, as before, the actual prices charged during the particular quarter. Then equation (40) implies

paxpaAx=pay(41)

Now if pe=(p1e,,p23e) represent the correct Leontief prices for the period in question—the unique prices at which the nonpositive profit condition is satisfied—then

paxpaAx=[pe(pepa)]x[pe(pepa)]Ax(42)=pexpeAx[(pepa)x(pepa)Ax]

The Leontief zero-profit price vector pe is computed as

pepeA=c(43)

where c denotes the vector of values added per unit of output; so, in particular,

pexpeAx=cx=totalvalueadded(44)

Since cx equals the total value added and, in our case, the total wage bill as well, it also equals the total value of final demand if there are no taxes levied on, nor any subsidies given to, consumers. Combining these equations, we obtain

pay = cx – [(pe – pa) x – (pe – pa)Ax]

or

pay=cx(pepa)y(45)

Since y denotes final net output, pay equals the value of supply at the actual market prices if producers receive these prices for their outputs; so, the equation (45) may be interpreted as stating that consumers must either be taxed or subsidized by the amount (pe – pa) y in order for Walras’ Law to hold. Since in our case (pe – pa) y is generally positive (i.e., prices are being held below their zero-profit levels), this term would represent a tax on consumers’ incomes. The tax would then be transferred to the loss-making enterprises, so that factors of production could be paid without the central bank having to create money.33 If, on the other hand, no taxes are levied, but the losses are financed by borrowing, then equation (45) should be rewritten as

pay+(pepa)y=cx(46)

The left-hand side of equation (46) now represents the value of supply, including the credit that has been created, while the right-hand side represents the total wage bill, and hence the value of final demand.

One may thus conclude that it would be possible for the government to finance losses caused by price distortions purely by means of increased taxes, without resorting to the inflationary mechanism of credit creation. Such a situation might occasionally be desirable if it were deemed necessary to control certain prices for a period of time without producing inflation.

Characteristics of macroeconomic system

Equation (18) determines the rate of change in the money supply as a function of the rate of inflation that has given coefficients, a realized value of the error term, and values of the predetermined variables. The straight line μ in Chart 7 is a graphical illustration of the equation. The slope dμt/dπt, of the line is denoted by

α1gtmt1+α3btmt1
Chart 7.
Chart 7.

Argentina: Equilibrium Rates of Increase in Money and Prices

Citation: IMF Staff Papers 1978, 003; 10.5089/9781451972559.024.A003

which should be positive unless bt, is both negative and sufficiently large or unless both gt and bt are negative. Equation (19) determines the rate of inflation as a function of the rate of change in the stock of money, with given realized values of the error term and of predetermined variables and coefficients. The slope dπt/dμt of the curve on a π–μ plane is denoted by

1+πt(1+μt)[1+β2(1+πt)]

and the rate of inflation satisfying the equation rises as the rate of growth in the stock of money increases, unless 1 + β (1 + πt,) is less than zero. The μ-line has a positive slope in Chart 7 because a higher rate of inflation implies a higher rate of money creation that is necessitated by a higher nominal deficit and a higher rate of monetization (in domestic currency) of foreign exchange reserves. The π-curve is positively sloped because a higher rate of growth in the money supply will cause a higher rate of increase in expenditures, which will be induced by the excess supply of money.34 Because the system is nonlinear, there can be two equilibrium points, as indicated in Chart 7. It is obvious, however, that only point e1 in Chart 7 is a stable equilibrium.

Since the level of desired stock of money is always equal to the actual stock of money, the equilibrium point e1 in Chart 7 represents an equilibrium between the supply of, and the demand for, money, as well as an equilibrium between rates of growth in money that satisfies both stochastic equations in (16). The chart helps illustrate the impacts of various changes in the exogenous variables and parameters of the system. An increase in the real deficit gt of the government will cause a rightward shift of the μ-line in Chart 7 because of a shift of the μ-intercept and the slope dμt/dπt, and both the rate of inflation and rate of growth in the stock of money will rise. An increase in the computed loss dt will cause a rightward shift of the μ-intercept of the line four quarters later, resulting in similar rises in the rates of increase in money and prices.

The time paths of the endogenous variables (μt, πt, and mt) are completely determined by the system; in contrast to a linear system, it is difficult to examine analytically the dynamic properties of the model, even if the coefficients are given.

The results of the simulations reported in Section III, however, suggest that the model is dynamically stable. A further examination of some of the properties of the model is conducted by recording the dynamic paths of adjustment caused by temporary shocks to the system. Three kinds of shocks are applied for this examination: (1) temporary (for the first quarter of 1967 only) increases in the government deficit in real terms (Δg67 I = 0.1, 0.3, and 0.5 billion pesos), (2) similar increases in the computed aggregate loss of the private sector in real terms (Δd67 I, = 0.1, 0.3, and 0.5 billion pesos), and (3) a temporary reduction in the coefficient of the rate of inflation in the real balance equation (Δβ˜2=0.1 for the first quarter of 1967). All the experiments with shocks (1) and (2) result in acceleration in the rates of increase in money and prices, during the first quarter of 1967 for increases in gt and during the first quarter of 1968 for increases in dt, with the magnitudes of the impacts approximately proportional to those of the shocks. The impacts oscillate through time in both cases and taper off quickly. The temporary decrease in β˜2 (lower demand for real balances) also gives higher rates of increase in money and prices during the first quarter of 1967, as anticipated, and the impacts quickly taper off with oscillations. A shock added to the supply side has a higher impact on prices than on money, as does a shock added to the demand side.

Chart 8.
Chart 8.

Argentina: Impacts of Temporary Changes in Policy Variables and Parameters on Rates of Change in Money and Prices, 1967-76

Citation: IMF Staff Papers 1978, 003; 10.5089/9781451972559.024.A003

Simulation of rate of change in money supply, with and without distortion variable

To evaluate the net increase in the explanatory power that is brought to the money supply equation by the distortion variable, the money supply equation is re-estimated with an alternative specification in which the government deficit and the change in foreign exchange reserves are the only causes of the changes in the money supply. The estimated equation is

μt=0.032(4.50)+1.149(20.36)gmt+0.608(5.09)bmt+ε˜t(31)

Equations (31) and (31′) are then used to conduct dynamic simulations of the rate of change in the money supply for the period 1967 I-1976 IV on the basis of the historical series of the rate of inflation and the exogenous variables. The distortion variable reduces the root-mean-squared error for the rate of change in the money supply by 7.9 per cent, from 3.93 per cent to 3.62 per cent. The following table contains root-mean-squared errors for the two subperiods:

Table 7.

Argentina: Simulated Rates of Change in Money Supply, With and Without Distortion Variable

(In per cent)

article image

BIBLIOGRAPHY

  • Aghevli, Bijan B., and Mohsin S. Khan, “Government Deficits and the Inflationary Process in Developing Countries,” Staff Papers, Vol. 25 (September 1978), pp. 383-416.

    • Search Google Scholar
    • Export Citation
  • 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.

    • Search Google Scholar
    • Export Citation
  • Amemiya, Takeshi, “The Nonlinear Two-Stage Least-Squares Estimator,” Journal of Econometrics, Vol. 2 (July 1974), pp. 105-10.

  • Banco Central de la República Argentina, Modelo Econométrico Sectorial Dinámico: Actualization de la Matríz Insumo Producto a 1970 (unpublished, Buenos Aires, 1973).

    • Search Google Scholar
    • Export Citation
  • Banco Central de la República Argentina, Sistema de Cuentas del Producto e Ingreso de la Argentina, Vol. 2 (Buenos Aires, 1975).

  • Banco Central de la República Argentina, Cuentas Nacionales de la República Argentina; Series Históricas, Vol 3, (Buenos Aires, 1976).

    • Search Google Scholar
    • Export Citation
  • Cagan, Phillip, “The Monetary Dynamics of Hyperinflation,” in Studies in the Quantity Theory of Money, ed. by Milton Friedman (University of Chicago Press, 1956), pp. 25-117.

    • Search Google Scholar
    • Export Citation
  • Dagnino Pastore, José Maria, Ingreso y Dinero: Argentina 1935-1960 (Buenos Aires, 1964).

  • Díaz-Alejandro, Carlos F., Essays on the Economic History of the Argentine Republic (Yale University Press, 1970).

  • Diz, Adolfo C., “Money and Prices in Argentina, 1935-1962,” in Varieties of Monetary Experience, ed. by David Meiselman (University of Chicago Press, 1970), pp. 71-162.

    • Search Google Scholar
    • Export Citation
  • Dutton, Dean S., “A Model of Self-Generating Inflation: The Argentine Case,” Journal of Money, Credit and Banking, Vol. 3 (May 1971), pp. 245-62.

    • Search Google Scholar
    • Export Citation
  • Frenkel, Jacob A., “Inflation and the Formation of Expectations,” Journal of Monetary Economics, Vol. 1 (October 1975), pp. 403-21.

  • Gale, David, The Theory of Linear Economic Models (New York, 1960).

  • Harberger, Arnold C., “The Dynamics of Inflation in Chile,” in Measurement in Economics: Studies in Mathematical Economics and Econometrics in Memory of Yehuda Grunfeld (Stanford University Press, 1963), pp. 219-50.

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund, International Financial Statistics, various issues.

  • Khan, Mohsin S., “Variable Expectations and the Demand for Money in High-Inflation Countries,” Manchester School, Vol. 45 (September 1977), pp. 270-93.

    • Search Google Scholar
    • Export Citation
  • Mallon, Richard D., and Juan V. Sourrouille, Economic Policymaking in a Conflict Society: The Argentine Case (Harvard University Press, 1975).

    • Search Google Scholar
    • Export Citation
  • Maynard, Geoffrey, and Willy van Rijckeghem, “Stabilization Policy in an Inflationary Economy—Argentina,” in Development PolicyTheory and Practice, ed. by Gustav F. Papanek (Harvard University Press, 1968), pp. 207-35.

    • Search Google Scholar
    • Export Citation
  • Miller, Marcus H., and John E. Spencer, “The Static Economic Effects of the U.K. joining the EEC: A General Equilibrium Approach,” Review of Economic Studies, Vol. 44 (February 1977), pp. 71-93.

    • Search Google Scholar
    • Export Citation
  • Morishima, Michio, and F. Seton, “Aggregation in Leontief Matrices and the Labour Theory of Value,” Econometrica, Vol. 29 (April 1961), pp. 203-20.

    • Search Google Scholar
    • Export Citation
  • Mussa, Michael, “Adaptive and Regressive Expectations in a Rational Model of the Inflationary Process,” Journal of Monetary Economics, Vol. 1 (Ocotber 1975), pp. 423-42.

    • Search Google Scholar
    • Export Citation
  • Olivera, Julio H. G., “On Structural Inflation and Latin-American ‘Structuralism’,” Oxford Economic Papers, Vol. 16 (November 1964), pp. 321-32.

    • Search Google Scholar
    • Export Citation
  • Olivera, Julio H. G., “Money, Prices and Fiscal Lags: A Note on the Dynamics of Inflation,” Banca Nazionale del Lavoro, Quarterly Review, Vol. 20 (September 1967), pp. 258-67.

    • Search Google Scholar
    • Export Citation
  • Olivera, Julio H. G., “On Passive Money,” Journal of Political Economy, Vol. 78 (July-August 1970), pp. 805-14.

  • Prebisch, Raúl, “Structural Vulnerability and Inflation,” in Leading Issues in Economic Development: Studies in International Poverty, ed. by Gerald M. Meier (Oxford University Press, Second Edition, 1970), pp. 238-41.

    • Search Google Scholar
    • Export Citation
  • Samuelson, Paul A., “Abstract of a Theorem Concerning Substitutability in Open Leontief Models” in Activity Analysis of Production and Allocation, ed. by Tjalling C. Koopmans (New York, 1951), pp. 142-46.

    • Search Google Scholar
    • Export Citation
  • Seers, Dudley, “Normal Growth and Distortions: Some Techniques of Structural Analysis,” Oxford Economic Papers, Vol. 16 (March 1964), pp. 78-104.

    • Search Google Scholar
    • Export Citation
  • Stone, Richard, and J.A.C. Brown, “A Long Term Growth Model for the British Economy,” in Europe’s Future in Figures, ed. by Robert C. Geary (Amsterdam, 1962), pp. 287-310.

    • Search Google Scholar
    • Export Citation
  • Tanzi, Vito, “Inflation, Lags in Collection, and the Real Value of Tax Revenue,” Staff Papers, Vol. 24 (March 1977), pp. 154-67.

  • Tanzi, Vito, “Inflation, Real Tax Revenue, and the Case for Inflationary Finance: Theory with an Application to Argentina,” Staff Papers, Vol. 25 (September 1978), pp. 417-51.

    • Search Google Scholar
    • Export Citation
  • Wachter, Susan M., Latin American Inflation: The Structuralist-Monetarist Debate (Lexington, Massachusetts, 1976).

  • Whalley, John, “The United Kingdom Tax System 1968-1970: Some Fixed Point Indications of Its Economic Impact,” Econometrica, Vol. 45 (November 1977), pp. 1837-58.

    • Search Google Scholar
    • Export Citation
*

Mr. Chu, economist in the Commodities Division of the Research Department, is a graduate of Kyung Hee University (Seoul) and Columbia University. Before joining the Fund, he was an instructor at Columbia University.

Mr. Feltenstein, economist in the Special Studies Division of the Research Department, received degrees from Harvard and Yale Universities. Before joining the Fund, he taught at the University of Massachusetts at Amherst.

The authors would like to thank Phillip Cagan, Michael Parkin, and John Whalley for their helpful comments and suggestions on an earlier draft of this paper. They are also grateful to a number of colleagues in the Fund and to the members of the monetary workshop at the University of Western Ontario, where an earlier version of this paper was presented, for valuable comments.

3

The hypothesis of regressive expectations has been used to explain this conclusion. The reader may wish to see Frenkel (1975) or Mussa (1975) for a description of this hypothesis.

4

The most extreme selective price controls came during the Social Contract of 1973 to mid-1976.

5

It is, of course, theoretically possible for the loss-making public enterprise to be financed through the private capital market, but in the Argentine case, such financing was not generally forthcoming.

6

As we mentioned before, the assumed exogenous nature of the changes in foreign exchange reserves is not correct for a part of the sample period.

7

The aggregate loss, a key variable in the model, should not be interpreted as the aggregated value of the actual “accounting” losses of the firms. It is an aggregate of the ex ante losses that would be incurred by the firms under a Leontief system with exogenous real values added and distorted actual prices. Actual losses of firms may have been smaller than these ex ante losses, not only because of subsidies but because of input prices maintained under the zero-profit levels by price control. Actual losses may have been larger than ex ante losses for those industries in which the practice of delayed payment for purchase of inputs was widespread and in which input prices were inflated in anticipation of inflation. The aggregate loss is referred to hereafter as “computed aggregate loss.”

8

Decomposing a change in high-powered money into only these three factors is an obvious simplification. There apparently have been other sources of increase in high-powered money. The estimated equation, however, suggests that these three factors adequately explain the variations in the rate of increase in money during the sample period.

9

The government deficit incurred by subsidization of public enterprises is represented in this study by the government’s current transfer payments to public firms. In reality, the subsidization also took other forms, such as exemption of some public enterprises from import tariffs and toleration of long delays in payment of taxes during inflationary periods. These other forms of subsidies should also have increased the government deficit to some extent. They are not included in the estimate of the subsidies in the model, however, because it is difficult to measure them.

10

To obtain a reasonable estimate of the length of the lag, equations with various lengths of lag were estimated in various forms by both the ordinary least-squares and the instrumental variables estimation techniques. The data strongly supported the one-year lag.

11

The money supply equation (8) is not an accounting identity. First of all, the equation linking the money supply and high-powered money in equation (1) would not be an identity unless the money multiplier k were constant. Furthermore, the equations introduced in (7), especially the second one, are essentially stochastic, although to simplify notation they are specified without error terms.

12

In a more expanded model, the real government deficit can be endogenized. See Tanzi (1978) and Aghevli and Khan (1977) for expanded treatments of the public sector in inflationary economies.

13

See Section III for a description of the impacts of changes in various policy variables and the Appendix for an analytical, as well as an empirical, examination of the characteristics of the model.

14

Such methods have been used by Whalley (1977) and Miller and Spencer (1977) to estimate the effects of changes of tax structure in various countries.

15

This result is demonstrated in the Appendix.

16

The input-output matrices for 1950, 1953, and 1963 are given in Banco Central de la República Argentina (1976), and the 1970 matrix is given in Banco Central de la República Argentina (1973).

17

Value added vaj refers to total value added rather than value added per unit of output.

18

The empirical results in Section III strongly support this assumption.

19

These 23 sectors are as follows: *(1) agriculture, hunting, and forestry, *(2) mining, *(3) food, drinks, and tobacco, *(4) textiles, *(5) clothing and shoes, *(6) wood products and furniture, *(7) paper and printing materials, *(8) hides and skins, *(9) rubber, *(10) chemical products, *(11) petroleum derivatives, *(12) non-metallic minerals, *(13) metal, *(14) machinery, (15) machinery and electrical apparatus, (16) transportation equipment, (17) others, *(18) electricity, gas, and water, *(19) construction, (20) commerce, restaurants, and hotels, (21) transportation and communication, *(22) housing, (23) personal and financial services. The sectors preceded by asterisks are those for which official price indices are available. (See Banco Central de la República Argentina (1976).)

20

More correctly, this difference represents the required total credit creation brought about by the particular price distortion per unit of output.

21

See the Appendix for a derivation of output unit that is consistent with the price indices.

22

The total loss Dppt is an estimate of the losses incurred by public and private enterprises. It is not unreasonable to assume that the losses of the public enterprises would be promptly subsidized by the government. Therefore, the amount of actual current transfer payments of the government to public enterprises is subtracted from the computed aggregate loss. The remainder Dt is an estimate of the losses of private industries.

23

At one point during 1975, the actual price of output in the agricultural sector had fallen to barely 30 per cent of its zero-profit level.

24

The IV estimators for the coefficients would be consistent under standard conditions if the instruments were not correlated with the errors. (See Amemiya (1974) for a description of the conditions for IV estimators of the parameters of general nonlinear models.) In this study, distributed lags of the basic exogenous variables (gt, gt-1, gt-2, dt-4, dt-5, dt-6, bt, bt-1 and bt-2) and a dummy variable (explained later) are used as instruments for both equations, because by assumption they are uncorrelated with the errors and because the time paths of the endogenous variables for which the instruments are used are characterized largely by the past time paths of these exogenous variables.

25

See the concluding paragraph of this section (III.2) for a discussion of the meaning of the Durbin-Watson (D-W) statistic for equation (32).

26

It often happened that suppliers of inputs, knowing that they would be paid only after a long delay, deliberately raised their prices to compensate for the anticipated increase in the general price level. Hence, the actual repayment of a loss was usually larger than the value of that loss at the time it was incurred.

27

Although the D-W statistic cannot be used to test serial correlation of the error for the real balance equation because the equation has a lagged dependent variable and because the estimation method is an IV technique, a useful asymptotic relationship can be established: It can be shown that the D-W statistic computed for a first-order MA process converges in probability to 3 if the MA parameter is 1, to 2 if the parameter is 0, and to 1 if the parameter is -1. The MA parameter suggested by the D-W statistic in equation (32) is greater than 1, but a perfect compatibility cannot be anticipated in this case because the estimates are not constrained.

28

In all simulations in this section, gt, dt, bmt, and dmt, (dummy variable) are assumed to be exogenous, and all simulations are dynamic from the beginning of the simulation period (the first quarter of 1967) to the end (the fourth quarter of 1976). The assumption of the exogenous nature of bmt is introduced because the paper mainly concerns itself with the impacts of price distortions on inflation.

The error ωt = ηt – λ ηt–1 for the real balance equation and the i.i.d. error ∊t of the money supply equation are suppressed for all quarters in the simulations. In a more accurate simulation, the residual n˜t1 for t = 1 would be used as an initial condition for starting the simulation. The suppression of n˜t1 would not have a large lingering effect on the simulated paths of the endogenous variables, however, because the model is stable.

29

The significance of the distortion variable (dmt-4) in explaining the money supply is indicated by the estimated money supply equation. The Appendix introduces the results of simulations conducted for a similar purpose.

30

See the Appendix for a further discussion of the results.

32

For a proof of this result, see Gale (1960), p. 296.

33

This tax is always feasible, that is, it does not take more than the total income of consumers, since, in the extreme case, when pa = 0, we have cx - (pe – pa) y = cx – pey = 0.

34

The stable equilibrium point could be in other quadrants. It is conceivable that both μt and πt, could be negative. According to the results obtained, however, μt was always positive, and πt was negative for only two quarters during the period 1964-76 in Argentina.

IMF Staff papers: Volume 25 No. 3
Author: International Monetary Fund. Research Dept.
  • View in gallery

    Argentina: Actual and Computed (Zero-Profit) Price Indices Deflated by Actual General Price Level, Selected Industries, 1965-761

    (1963 = 100)

  • View in gallery

    Argentina: Actual and Computed (Zero-Profit) Price Indices Deflated by Actual General Price Level, Selected Industries, 1965-761

    (1963 = 100)

  • View in gallery

    Argentina: Exogenous Variables, 1965-76

  • View in gallery

    Argentina: Results of Long-Run Dynamic Simulation with Historical Series of Exogenous Variables, 1967-76

  • View in gallery

    Argentina: Policy Simulations Based on Different Assumptions on Government Deficit in Real Terms, 1967-76

  • View in gallery

    Argentina: Policy Simulations Based on Different Assumptions on Computed Aggregate Losses in Real Terms, 1967-76

  • View in gallery

    Argentina: Equilibrium Rates of Increase in Money and Prices

  • View in gallery

    Argentina: Impacts of Temporary Changes in Policy Variables and Parameters on Rates of Change in Money and Prices, 1967-76