A Multilateral Exchange Rate Model for Primary Producing Countries
Author:
Mr. Andrew Feltenstein
Search for other papers by Mr. Andrew Feltenstein in
Current site
Google Scholar
Close
,
Mr. Morris Goldstein https://isni.org/isni/0000000404811396 International Monetary Fund

Search for other papers by Mr. Morris Goldstein in
Current site
Google Scholar
Close
, and
Ms. Susan M Schadler
Search for other papers by Ms. Susan M Schadler in
Current site
Google Scholar
Close

In a world of significant exchange rate variability, the problem of evaluating the impact of exchange rate changes on a country’s trade balance carries immediate operational relevance. Even those countries that continue to peg the value of their currencies find that frequent changes in the exchange rates of floating currencies affect the outlook for a substantial segment of their own exports and imports. For example, in 1977, countries that pegged their currencies to the U. S. dollar, the pound sterling, the French franc, the SDR, or some other currency composite still conducted roughly 40 per cent of their export trade with countries that either allowed their currencies to vary independently or did not fix their currencies to the same peg. On a global basis, trade between countries that fixed the value of their currencies in relation to one another accounted for less than one fifth of world exports in 1977.

Abstract

In a world of significant exchange rate variability, the problem of evaluating the impact of exchange rate changes on a country’s trade balance carries immediate operational relevance. Even those countries that continue to peg the value of their currencies find that frequent changes in the exchange rates of floating currencies affect the outlook for a substantial segment of their own exports and imports. For example, in 1977, countries that pegged their currencies to the U. S. dollar, the pound sterling, the French franc, the SDR, or some other currency composite still conducted roughly 40 per cent of their export trade with countries that either allowed their currencies to vary independently or did not fix their currencies to the same peg. On a global basis, trade between countries that fixed the value of their currencies in relation to one another accounted for less than one fifth of world exports in 1977.

In a world of significant exchange rate variability, the problem of evaluating the impact of exchange rate changes on a country’s trade balance carries immediate operational relevance. Even those countries that continue to peg the value of their currencies find that frequent changes in the exchange rates of floating currencies affect the outlook for a substantial segment of their own exports and imports. For example, in 1977, countries that pegged their currencies to the U. S. dollar, the pound sterling, the French franc, the SDR, or some other currency composite still conducted roughly 40 per cent of their export trade with countries that either allowed their currencies to vary independently or did not fix their currencies to the same peg. On a global basis, trade between countries that fixed the value of their currencies in relation to one another accounted for less than one fifth of world exports in 1977.

The primary purpose of this paper is to present a small simulation model that can be used to estimate the medium-run effects of exchange rate changes on the trade balance of a primary producing country (PPC). The model is intended to be sufficiently general in structure to be applicable to a variety of primary producing countries, but for the purposes of empirical illustration it is applied here to four copper producing countries, Chile, Peru, Zaïre, and Zambia. Since reliable empirical estimates are not yet available for all the parameters appearing in the model, the reported simulation results should be regarded more as indications of the types of questions that can be answered by the model than as firm estimates for these particular countries.

The model has four general features that were designed to accommodate some of the more important characteristics of primary producing economies and of the international environment in which they operate. First, the model is multilateral in that it is capable of estimating the effect of a number of simultaneous exchange rate changes on a given country’s trade balance. The tendency for many exchange rates to change, either simultaneously (as in a negotiated currency realignment) or at close intervals (as in the continual readjustment of floating exchange rates), makes the multilateral approach necessary.

Second, the model adopts a commodity-by-commodity approach to primary producing countries’ export earnings. This aspect reflects the facts that, relative to industrial countries, the exports of most primary producing countries are concentrated in a small number of commodities or products, 1 and that these export products are ones that are not distinguished by their place of origin; for example, copper exported by Zambia is a perfect substitute for copper exported by Chile. This latter feature means that the price an exporter receives for his product depends on conditions in the world market for that commodity. Hence, in our view, the interplay between individual primary producing countries’ export receipts and world supply and demand conditions for the products they export can best be analyzed by taking a commodity-by-commodity approach. In addition, disaggregation of exports by commodity allows one to break free of the restriction that each primary producer be regarded either as a price taker or as a price setter for all its exports. In other words, disaggregation of exports by commodity permits a given primary producing country to be a price setter for some of its exports and a price taker for others, with the distinction based on the country’s ability to affect the world supply and hence the world price for that commodity.

The third general feature of the model is the attention paid to the extremely high rates of inflation experienced in many primary producing countries in the past, 2 the effects of which have partially, or even completely, offset the competitive advantage normally gained by devaluation (as regards the country’s external or internal terms of trade). 3 In the model the domestic rate of inflation in each primary producing country is expressed as a function of its excess money balances. The rate of change of domestic prices can then be used to adjust all nominal exchange rate changes into real exchange rate changes, since it is the real exchange rate that is likely to affect the decisions of exporters and importers. The model can perhaps best be understood as providing estimates of the impact of exchange rate changes on a country’s trade balance given some assumed stance of monetary policy in that country that determines its inflation rate. In simulation exercises with the model, it is therefore possible to estimate the trade balance effects of exchange rate changes under alternative rates of domestic monetary expansion. 4

Finally, the model was deliberately kept small and its structure was designed to be relatively simple, so that the information needed to solve the model would be readily available for at least many (less developed) primary producing countries. Thus, the model focuses on only a few key behavioral relationships, such as the demand for money, the demand for imports, the determination of real expenditure, and the supply of exports.

The approach here is to specify simple models of world trade in each of several primary commodities that make up the bulk of a primary producer’s export earnings. Exchange rate changes, which are assumed to be exogenous, together with changes in domestic price levels that are assumed to result from excess money supplies, shift the import demand and export supply schedules for each commodity. These shifts lead to changes in the world price of each commodity. Translating the world price into real domestic currency terms, the change in producer countries’ export earnings for that commodity depends on the size of the domestic supply elasticity. Summing over all of a country’s commodity exports yields the change in total export earnings induced by exchange rate and money supply changes. This change in export earnings, as well as the change in relative prices again resulting from exchange rate and money supply changes, determines the change in the country’s total imports. All of these effects are assumed to take place over the medium run, defined as the length of time over which the country can control both its money supply and its exchange rate, and over which the exchange rate has real effects on trade flows via the exchange rate induced change in relative prices. 5 For the purposes of the simulation exercises, the medium run has been rather arbitrarily labeled as three years, although if a country has relatively high rates of inflation, open capital markets, limited means for sterilizing reserve inflows or outflows, and a small nontraded sector, the relevant medium run may be measured in months rather than years.

The rest of the paper is organized into four sections. Section I describes the structure of the model, while Section II identifies its limitations as well as some of the problems encountered in constructing the model. Section III presents the results of three simulation exercises based on the model and also points out some additional potential uses of the model, including the generation of indices of effective exchange rates for individual primary producing countries. Some concluding remarks are offered in Section IV.

I. The Model

To explain the structure of the model in detail, this section has been divided into four parts describing export changes, import changes, the domestic rate of inflation, and changes in the trade balance.

EXPORT CHANGES

The simple supply and demand system used to determine the change in the world price of each primary commodity can be written as follows: 6

S i = S i ( P i r X ) S 1 > 0 ( 1 )
D i = D i ( E r M ,  P i r M , P s r M ) D 1 , D 3 > 0 , D 2 < 0 ( 2 )
S i = D i ( 3 )
P i r X T X Σ k α i k ( P i k T k / P D k ) P i r W ( 4 )
P i r M T M Σ j β i j ( P i j T j / P j D ) P i r W ( 5 )

where

Si = the supply of exports of commodity i (in volume terms)

Di = the demand for imports of commodity i (in volume terms)

PirX = the real price of commodity i in terms of exporters’ (PPC) currencies

PirM = the real price of commodity i in terms of importers’ currencies

PsrM = the real price of substitutes for commodity i in terms of importers’ currencies

Pik = the price of ith good in the kth country in domestic currency

T = the exchange rate, that is, U. S. dollars per unit of importers’ or exporters’ currency

αik = the share of exporter k in total exports of commodity i

βij = the share of importer j in total imports of commodity i

ErM = real expenditure of importing countries

Superscripts X and k (M and j) represent aggregate and individual exporters (importers), respectively, superscript W denotes world prices in terms of U. S. dollars, and subscripts r and D indicate real magnitudes and the domestic value of a variable, respectively.

equation (1) is the global export supply function for commodity i, where the quantity exported depends on the profitability of producing export commodities; the ratio of the commodity export price (in local currency) to the overall domestic price level is used as a proxy. The commodity export price is deflated by the overall domestic price level to capture the effect of exchange rate and money supply changes not only on the price exporters receive but also on their costs of production. The more sensitive are domestic wage rates and other factor costs to devaluation, and the more expansionary is the accompanying monetary policy, then the smaller will be export supply induced by the devaluation. 7 In some cases, of course, the cost of production may vary widely from commodity to commodity, so that the overall price index may be a misleading cost indicator for an individual commodity. However, data on other indicators of cost that would be more commodity specific usually are not available. 8 In some trade models, other variables are included in the export supply function, for example, an index of the industry’s production capacity, trend exports, or seasonal factors. 9 These variables are excluded from this model because they are unlikely to be affected by either exchange rate changes or money supply changes.

In equation (2) the global demand function for commodity i is postulated to depend on the level of real expenditure in import markets, the (real) world price of commodity i, and the (real) world price of substitutes for commodity i. equation (3) is the condition for commodity market equilibrium, while equations (4) and (5) are identities that give the relationships between the price in importers’ currencies, the price in exporters’ currencies, and the world price for commodity i.

This model can be solved for the proportionate change in the world price of commodity i (P˙iW) by totally differentiating equation (3), using identities (4) and (5) to translate importers’ and exporters’ prices into weighted averages of real exchange rates, and by making some substitutions to yield

P ˙ i W = n s i K ˙ i n d i R ˙ i + n e i E ˙ i r M + n a i P ˙ s r M n s i n d i ( 6 )

where a dot (·) over a symbol indicates a percentage change, that is, x˙=dxx, and where

K˙i = the average proportionate change of exporters’ real exchange rates (exchange rate changes deflated by the domestic rate of inflation) in terms of U. S. dollars, weighted by their share in total world exports of commodity i

i = the average proportionate change of importers’ real exchange rates in terms of U. S. dollars, weighted by their share in total world imports of commodity i

nsi = world price elasticity of supply for commmodity i

ndi = world price elasticity of demand for commodity i

nei = world elasticity of demand with respect to expenditure for commodity i

nai world price elasticity of demand for commodity i with respect to the price of substitute commodities

Several aspects of equation (6) are worth emphasizing. First, the supply and demand elasticities are world elasticities. Unless these elasticities are assumed to be identical across countries, the world elasticities should be obtained as weighted averages of individual country elasticities. The procedure followed in the empirical work was to weight the country elasticities according to the country’s share of world exports or world imports of that commodity. Estimates of country elasticities come primarily from commodity studies.

Second, partial differentiation of the world price (PiW) with respect to exchange rate changes (e.g.,dPiW/dKi=[nsi+neidEi/dKi+naidPs/dKi]/(nsindi)) reveals that the effect of exchange rate changes on the world price of a commodity depends not only on the elasticities nsi, ndi, nei, and nai but also on the changes in expenditure and the change in prices of substitute commodities induced by exchange rate changes. These two variables are clearly more difficult to estimate than the actual changes in expenditure and in the prices of substitutes alone (dEi/dKi, dPs/dKi versus Ėi and s, since the changes resulting from exchange rate movements are likely to be different from the observed changes. Fortunately, we were able to obtain estimates of the exchange rate induced changes in expenditure in importing countries from the Artus-Rhomberg (1973) industrial country MERM which produces the relevant change in expenditure for each of the industrial countries and for the oil exporters as a group. Each of these expenditure changes can then be weighted by the country’s share in total imports of the commodity to produce the (approximate) change in world expenditure in importing countries resulting from exchange rate changes. The exchange rate induced changes in the world prices of substitutes could in principle be taken from the solution of a supply and demand system for the appropriate substitute commodity. In practice, the substitution elasticities (nai) did not appear to be large enough, at least for the five primary commodities considered here, to warrant including these effects in the simulation exercises.

Once the proportionate change in the world price of commodity i is known, the change in the kth country’s export earnings from commodity i (X˙ik) is determined as follows:

X ˙ i k = n s i k ( P ˙ i W T ˙ k ( 1 + P ˙ D k ) P ˙ D k )    + P ˙ i W ( 7 )

where nsik is the kth country’s supply price elasticity for commodity i. equation (7) confirms the a priori expectation that the change in the country’s export earnings (in numeraire currency) will be larger (i) the larger the domestic supply elasticity for the commodity, (ii) the larger the increase in the world price of the commodity, (iii) the smaller the rate of domestic inflation, and (iv) the smaller the devaluation with respect to the numeraire currency (the U. S. dollar).

From equation (7), it follows directly that the change in country k’s total export earnings induced by exchange rate changes, world expenditure changes, and the rate of domestic inflation is

X ˙ k = Σ i = 1 I w i k X ˙ i k ( 8 )

where wik is the share of commodity i in country k’s total exports and I is the total number of (primary) export commodities.

IMPORT CHANGES

There are a number of channels by which exchange rate changes can affect a country’s expenditure on imports. The model focuses on three of these. First, exchange rate changes stimulate or depress export earnings, in turn affecting income, expenditure, and the demand for imports. Second, exchange rate induced changes in the price of exports relative to the price of imports (i.e., terms-of-trade changes) change real income, which feeds through via expenditure changes to affect the value of imports demanded. Third, as long as there is some substitution in consumption between imports and domestically produced goods, an increase in import prices induces some substitution away from the former toward the latter and thus affects the volume of imports demanded. 10

Following Bélanger (1976), these effects on import expenditure can be represented in the following submodel of income, expenditure, and the demand for imports:

Y ˙ k = α D k D ˙ k + α X k ( X ˙ k T ˙ k ) α M k ( M ˙ k T ˙ k ) ( 9 )
D ˙ k = P ˙ D k + n D Y k ( Y ˙ k P ˙ Y k ) ( 10 )
M ˙ k T ˙ k = ( P ˙ M k T ˙ k ) + n M D k ( D ˙ k P ˙ D k ) + n M S k ( P ˙ D k P ˙ M k ) ( 11 )

where

Yk = national income in country k

Dk = domestic expenditure in country k

Xk = value of country k’s exports in terms of U. S. dollars

Mk = value of country k’s imports in terms of U. S. dollars

PDk = price deflator for domestic expenditure

PYk = price deflator for national income (PYk=αDkPDk+αXkPXkαMkPMk)

PMk = price of imports in country k

αDk = share of domestic expenditure in national income of country k

αXk = share of exports in national income of country k

αMk = share of imports in national income of country k

nDY = elasticity of domestic expenditure with respect to national income

nMD = elasticity of imports with respect to domestic expenditure

nMS = elasticity of imports with respect to the ratio of the price of imports to the domestic price level

equation (9) is the national income identity and equation (10) relates the change in real expenditure to the change in real national income. equation (11) describes the change in the volume of imports as a function of both changes in real expenditure and changes in the relative price of imports. In primary producing countries where import demand is affected by nonprice factors or restrictions (other than those associated with export earnings), a term to represent them could be included in equation (11), although in practice such factors are difficult to measure.

Substituting equations (9) and (10) into equation (11) produces an expression for changes in import expenditures in terms of changes in export earnings, changes in exchange rates, and changes in domestic and import prices; specifically,

M ˙ k = ( 1 α D k n D Y k ) ( P ˙ M k T ˙ k n D Y k P ˙ D k ) 1 n D Y k ( α D k n M D k α M k ) + n M D k n D Y k [ α D Y k ( 1 n D Y k ) P ˙ D k + α x k ( X ˙ k T ˙ k ) ] 1 n D Y k ( α D k n M D k α M k ) + T ˙ k ( 12 )

One difficulty that arises in using equation (12) in simulations is how to measure the part of the change in import prices that is due to exchange rate changes. This problem is analogous to that mentioned earlier with respect to expenditure changes in importing countries, and the solution is again to use results generated by the industrial country MERM. More precisely, the industrial country MERM produces the change in each industrial country’s export prices induced by exchange rate changes. By viewing a primary producing country’s import prices as a weighted average of industrial country export prices, it is possible to approximate the change in country k’s import prices due to exchange rate changes. The weights used are the share of country k’s imports coming from industrial country j, so that

P ˙ M k = Σ j = 1 22 u k j P ˙ X j ( 13 )

where PXj is country j’s export prices, and ukj is its weight in country k’s total imports.

Although equation (12) is rather cumbersome, it supports the a priori expectation that the change in import expenditure following an exchange rate change will be greater (i) the greater the change in export earnings induced by the exchange rate changes, (ii) the greater the share of exports in national income, (iii) the greater the elasticities of domestic expenditure with respect to national income and of import demand with respect to domestic expenditure, (iv) the greater the rate of domestic inflation, and (v) the smaller the increase in country k’s import prices. 11

DOMESTIC RATE OF INFLATION

As emphasized earlier, the effect of a nominal exchange rate change can be evaluated only against the backdrop of the change in the domestic price level that occurs at the same time. For this reason, it is important to investigate the determinants of the rate of inflation within the model itself. 12

We have postulated that the domestic rate of inflation in primary producing countries is a function of the excess demand for money balances. While this approach may seem unduly rigid in that there are a variety of factors that act on the inflation rate, many of these seemingly nonmonetary factors, such as government budget deficits and relative price distortions, are implicit in the excess demand for money insofar as they affect the money supply. 13 The inflation determination process can be represented as follows:

log mo t D = a 0 + a 1  log  ( Y D k / P D k ) t + a 2 P E ˙ t k ; m o = ( M o / P D k ) a 1 > 0 , a 2 < 0 ( 14 )
P E ˙ t k = P E ˙ t 1 k + b ( Δ log P ˙ D t k P E ˙ t 1 k ) 0 < b < 1 ( 15 )
m o D = m o ( 16 )

equation (14) expresses the demand for real money balances (mo) as a function of the level of real income and of the expected rate of inflation (k), where the latter represents the opportunity cost of holding money. 14 Implicit in this formulation is the assumption that in many primary producing countries the relevant decision is between holding money and goods rather than between holding money and other financial assets. equation (15) is an adaptive expectation scheme whereby the current inflation forecast is equal to the past period’s forecast plus some fraction (b) of the past period’s forecast error. Finally, equation (16) closes the inflation submodel by stating that the price level moves so as to equate the actual level of real balances to the demanded level. By substituting equations (16) and (15) into equation (14) and solving for log PDtk, it is possible to obtain an expression for the price level in terms of only observable variables. Using the consumer price index as a proxy for the overall domestic price level, and calculating the percentage change in the price level over the period for which the model is simulated, yields the domestic rate of inflation (P˙Dk) that is to be plugged into the export and import determination equations.

One important outcome of both adopting a “monetary” explanation of domestic inflation and assuming that the exchange rate and the money supply are independent policy instruments in the medium run is that it becomes possible to estimate the trade balance effects of alternative combinations of exchange rate and domestic money supply changes. Three particular combinations are worth mentioning explicitly. First, if the objective is to determine the trade balance impacts of two alternative sets of exchange rate changes, this can be done by specifying a given money supply growth rate and then comparing the trade balance changes under the two alternative exchange rate scenarios. The pure “exchange rate effect” thus emerges as the difference between two policy scenarios with identical money supply assumptions rather than as a single estimate that reflects both policy instruments. Second, in a similar vein, if one wishes to estimate the trade balance effects of two alternative money growth rates, this can be accomplished by specifying a given exchange rate change (including zero) and comparing the estimated trade balance changes corresponding to the two alternative money growth rates. Again, the calculation of the independent effect of one of the policy instruments requires a comparison of two trade balance outcomes where one of the instruments is being held constant at some prescribed rate of change. In short, while the model posits the trade balance as a joint product of monetary and exchange rate policies, it is still possible to estimate the effect of exchange rate changes (or money supply changes) alone on the trade balance, albeit in a particularly defined way. Finally, there is the third general case where the objective is to estimate the simultaneous or joint effect of exchange rate and money supply changes on the trade balance. This is, in fact, the case analyzed in the simulation section. In this instance, one can still speak of the trade balance effects of exchange rate changes, but it must be recognized that these exchange rate effects apply only to a given set of money supply and inflation rates in the subject countries.

TRADE BALANCE CHANGES

The expressions derived earlier for the proportionate changes in a country’s export receipts and import expenditures (equations 8 and 12, respectively) can be combined to yield the net effect of exchange rate and money supply changes on the kth country’s trade balance (TB˙k). This can be compactly written by using the following identity:

T B ˙ k = X 0 k X ˙ k M 0 k M ˙ k ( 17 )

where X0kandM0k are the values of country k’s exports and imports in some arbitrary initial period.

One of the more important potential uses of the framework summarized in equation (17) is the derivation of the kth country’s real effective exchange rate. 15 This can be done by substituting for TB˙k in equation (17) the change in country k’s trade balance that was estimated to have occurred over a given period (as a result of all exchange rate and money supply changes) and then inverting the equation to isolate the change in the value of country k’s currency on the left-hand side. Setting all other exchange rate changes equal to zero and similarly setting inflation rates in all countries at zero, one can then obtain the change in the kth country’s exchange rate (vis-à-vis the numeraire) that would have produced the same change in country k’s trade balance that was estimated to have taken place as a result of the actual changes in all exchange rates and money supplies. This is precisely the definition of country k’s (real) effective exchange rate in the model.

II. Limitations of the Model

Before presenting an application of the model to four copper producing countries, some of the problems encountered in building the model will be reviewed. These problems can be presented under three headings, namely, the characteristics of the commodity models, the influence of nonprice factors on trade flows, and the appropriate criteria for evaluating the performance of the model. 16

COMMODITY MODELS

One simplifying assumption in the highly generalized commodity model of equations (l)–(5) is that prices alone act to clear commodity markets. The model thus abstracts from changes in stocks and changes in order backlogs which may also act to eliminate disequilibria, particularly for durable commodities. Once stockpiling is admitted, related factors—such as the influence of expected versus actual price changes on stocks, demand for consumption versus demand for investment (or speculative) purposes, and the role of secondary markets for recycled commodities—must be modeled. 17 Unfortunately, models of this type exist for very few primary commodities, so that it is difficult to measure the errors involved in using a simpler model. Since the influence of stocks and order backlogs vis-à-vis that of prices on market clearing probably recedes as the time period under question increases, the assumption that prices alone clear markets may not be very inaccurate over the time horizon of one to three years considered in this study.

A second difficulty is that the present commodity model accounts for the behavior of exports and imports without recognizing that countries can be both major consumers and suppliers of a commodity, as, for example, the United States and Canada in the copper market. By producing a significant quantity of a particular commodity for the domestic market, some countries can have an appreciable influence on the world price of a commodity even if they are not major exporters (or importers) of that commodity. In principle, the solution to this problem is fairly straightforward. Equations (1) and (2) could be rewritten in terms of total demands and supplies rather than total exports and imports; the relevant elasticities in the expression for world price changes (equation (6)) then would be total demand and supply elasticities rather than import demand and export supply elasticities. In this broader context, exports and imports would represent excess domestic supply and demand, respectively. In practice, however, this alternative approach presents difficulties because data on total domestic production and consumption are harder to obtain for many commodities than those on imports and exports. Similarly, estimates of elasticities of import demand and export supply are generally more readily available than those for total demand and supply.

Imperfectly competitive commodity markets present another departure from the assumptions underlying the supply and demand model. In the context of this study, imperfections are of two types—those arising from a dominant supplier that essentially determines the world price of a given primary commodity and those arising from product differentiation. The first imperfection applies to Zaïre’s position in the world cobalt market, while the second applies to manufactured exports in each of the four copper producing countries. The dominant-supplier case can be handled by deriving the expression for changes in world price and quantity when the dominant supplier maximizes its profit subject to the market demand and supply response of other producers. These expressions and estimates of the relevant demand and supply elasticities for Zaïre were taken from earlier studies by Bélanger (1976) and Charles River Associates (1969).

Changes in export receipts for manufactures and “other” exports can be handled in a number of ways. One option, taken by Bélanger (1976), is to use the industrial country MERM to estimate changes in export prices of these goods, and then to combine them with an estimate of the domestic supply elasticity (for the economy as a whole) to derive the change in export earnings of manufactured goods owing to exchange rate changes. A second option would be to use a mark-up price equation to estimate the change in export prices of manufactures which can then be plugged into the relative price term in a traditional export demand equation to obtain the volume effects of an exchange rate change. 18 Finally, if manufactured exports represent a small proportion of total exports (as, for example, in the case of Zambia), they can be ignored; that is, one can simply normalize the weights in equation (8) so that exports of primary commodities exhaust total exports. The last option was used for the simulations.

NONPRICE FACTORS

Import demand and export supply decisions in primary producing countries are frequently not made on the basis of relative prices alone. In particular, nonprice allocative mechanisms, such as quotas, special credit and tax programs, and bilateral trading agreements, also affect countries’ trade flows. These nonprice factors, which are notoriously difficult to quantify, can have an important effect on countries’ trade balances.

In appraising the effect of nonprice factors on the model’s estimates of the trade balance effects of exchange rate changes, three considerations are important. First, to the extent that quantitative restrictions on import demand vary systematically with the availability of foreign exchange receipts, the model has already taken some account of them by allowing export earnings to affect aggregate import demand. As the model evolves, it should be possible to add foreign aid flows, receipts from exports of invisibles, and other inflows in the balance of payments to export earnings to get closer to a true measure of foreign exchange receipts. 19 One might also consider these additional items to be exogenous and incorporate them in the model as now formulated. It may also be preferable to permit a weighted average of foreign exchange receipts in the current year and previous years, rather than just in the current year, to affect import demand, thereby making imports less sensitive to temporary abrupt changes in export earnings. Second, the objective of the model is to estimate the trade balance effects of exchange rate (and money supply) changes not to forecast or to explain changes in the trade balance as a whole. Thus, even when non-price factors are important, they will distort the results of the model only when their changes are highly correlated with exchange rate changes; it is only in this case that the model will incorrectly attribute to the exchange rate those changes in the trade balance that were in fact due to other factors. Third, there are some country studies that explicitly introduce nonprice factors into the estimates of import demand and export supply equations. 20 When available, these studies provide estimates of demand and supply price elasticities holding other factors, including nonprice influences, constant. Using these estimated elasticities in the equations for global price changes, and for changes in export earnings and import expenditures, alleviates the problem of spurious correlation that could result from failure to account for nonprice factors. The force of the above comments is not to imply that nonprice factors do not seriously complicate the task of estimating the trade effects of exchange rate changes but rather to suggest that their presence in primary producing countries need not invalidate the model’s estimates.

EVALUATING THE MODEL’S PERFORMANCE

While most models can be evaluated by comparing the results predicted by the model with actual data, this approach has some drawbacks for our purposes. First, our model estimates the trade balance changes that result only from exchange rate and money supply changes. Thus, during any period when factors other than exchange rates and money supplies (e.g., world oil prices) have an important influence on a country’s trade balance, there is unlikely to be a close correspondence between predicted and actual values. One possible way around this problem would be to find a particular year in which exchange rate and money supply changes were the dominant factors affecting the trade balance and then compare actual and simulated values for that year. The difficulty here, however, is that the simulations involve four primary producing countries, and what might be an equilibrium year for one country may not be one for the others. In fact, we were unable to find a common equilibrium year for Chile, Peru, Zaïre, and Zambia.

A second factor that complicates the comparison of actual and predicted trade balance changes is the treatment of time lags. Actual trade balance changes are, of course, observed only in a single year, yet the trade balance change for that year can be the result of exchange rate changes occurring in previous years as well as in that year; therefore, in any historical simulation where exchange rates are changing every year, there is a matching problem. One approach would be to assume that exchange rate changes affect the trade balance over a three-year period. The total change in the trade balance in, say, 1973 would then be given by the sum of the trade balance effects of exchange rate changes occurring in 1971, 1972, and 1973 with the (estimated) changes in 1972 and 1971 reduced by one third and two thirds, respectively. This is, however, only one approximation to the true time lags.

Another approach to evaluating the model’s output would be to compare its results with that of other models. In practice, because other models either are not designed to estimate the effects of a set of exchange rate changes or do not apply to the four copper producing countries under study here, this approach is not feasible. Still another alternative would be to analyze some of the key structural equations in the model, such as the ones for the domestic rate of inflation, real expenditure, and aggregate import demand. Again, however, there are definite limits to what can be learned from this exercise because the data for the four countries are poor (both in quality and in the number of observations) and because there are few other empirical studies with which to compare our estimates. 21

III. Simulation Analysis

In this section the results of three simulations using the model are reported. In the first simulation, the trade balance effects are estimated for each of the four copper producing countries of the actual set of real exchange rate changes that occurred in exporting and importing countries over the 1971–76 period. The second simulation considers the trade balance effects of the nominal exchange rate changes that took place over this period. By comparing the results of the first two simulations, it is possible to derive a measure of the importance of including the domestic rate of inflation in the model. The third simulation estimates the changes in world copper prices that would result from an increase in the world price elasticity of demand for copper.

Before proceeding to the simulations themselves, the values that were assigned to the key parameters in the model are presented. Table 1 shows the assumed supply price elasticities for each of the five primary commodities in the model. There is a supply elasticity for each of the four primary producing countries as well as a weighted elasticity for the world as a whole. Also, each of the elasticities has one-, two-, and three-year components to be consistent with the (admittedly tentative) assumption in the model that the trade balance effects of exchange rate and money supply changes take place over three years.

Table 1.

Assumed Supply Elasticities for Primary Commodities

article image

Cobalt is treated as a commodity with a single dominant producer, namely Zaïre, following Bélanger (1976).

World supply elasticities for copper are significantly higher than those for any of the four countries, reflecting the relatively high supply elasticities in certain other major producing countries, such as the United States and Canada.

Table 2 provides the assumed world demand elasticities for the same five primary commodities. There is a price elasticity and an expenditure elasticity for each commodity. Viewing the numbers in Tables 1 and 2 together, the key feature is that copper (the most important export commodity in each of the four countries) is quite price inelastic, especially on the demand side; note that it carries the lowest price elasticity of demand of the five commodities considered. These low price elasticities reflect the facts that present technology permits very limited substitution for copper in electrical use (aluminum is the only practical substitute) and that copper production can usually not be expanded rapidly in the short to medium run (because of the time lags and resource constraints associated with opening new mines).

Table 2.

Assumed World Demand Elasticities for Primary Commodities

article image

Table 3 gives the assumed values for those elasticity parameters that link national income, domestic expenditure, and the demand for imports in our primary producing countries (see equations 9, 10, and 11). It can be seen that the main difference among the four copper producers lies in the assumed expenditure elasticity for imports rather than in either the relative price elasticity of demand for imports or the elasticity of expenditure with respect to income. The significantly higher import expenditure elasticity in Zaïre vis-à-vis Zambia is based on Bélanger’s (1976) estimated import equation for Zaïre combined with a new estimation of the Zambian import equation.

Table 3.

Assumed Elasticity Values for Income, Expenditure, and Imports, 1971–76

article image

At this point we should mention again that (aside from two demand-for-money equations for Chile and Zaïre and an import equation for Zambia) we have not estimated any of the parameters of the model ourselves. 22 Rather, we have assumed “consensus values” for the individual country and global parameters based on published empirical studies and preliminary discussions with commodity and country specialists. 23

HISTORICAL REAL EXCHANGE RATE CHANGES, 1971–76

The first step in both this and the following simulation exercise is to compute the weighted average percentage changes in importers’ and exporters’ exchange rates (relative to the U. S. dollar) for each of the five primary commodities examined here. As described in Section I, the changes in exporters’ (K˙i) and importers’ (i) exchange rates can be expressed in real terms as

K ˙ i = Σ j V i x j ( T ˙ i x j ( 1 + P ˙ D j ) + P ˙ D j ) ( 18 )
R ˙ i = Σ k V i M j ( T ˙ i M j ( 1 + P ˙ D j ) + P ˙ D j ) ( 19 )

where Vixj is a weight equal to the share of country j in global exports of commodity i, Vimj is a weight equal to the share of country j in global imports of commodity i, j is the percentage change in country j’s exchange rate relative to the U. S. dollar (expressed in U. S. dollars per unit of domestic currency), and P˙Dj is the jth country’s domestic rate of inflation. Applying expressions (18) and (19) to the actual pattern of exchange rates during 1971–76 yields the weighted exchange rate changes given in Table 4. In brief, the exchange rate figures in Table 4 indicate that the currencies of both exporters and importers of primary commodities appreciated in real terms vis-à-vis the U. S. dollar over the period. The rate of appreciation was often, however, quite different between the two groups, with, for example, exporters of copper recording significantly larger appreciations than importers in 1971, 1973, and 1976, and a significantly smaller appreciation in 1975. Given the weighted changes in real exchange rates shown in Table 4, and given the parameters in Tables 1, 2, and 3, the change in the world price of each primary commodity is computed from equation (6). These global price changes are presented in Table 5. Of special note are the exchange rate induced increases in the world prices of all five primary commodities over the 1971–76 period. This is precisely the result one would expect given the appreciation of both exporters’ and importers’ currencies relative to the U. S. dollar over the period, since appreciation of exporters’ currencies reduces the world supply of copper while appreciation of importers’ currencies increases the world demand for copper. As for the sizes of the price increases, it is worth recalling the well-known proposition that inelastic supply and demand curves (low global price elasticities of demand and supply) imply relatively large price changes in response to shifts in demand and/or supply. Reference to Tables 1 and 2 confirms that most of the primary commodities fall into this low-elasticity category.

Table 4.

Weighted Average Changes in Importers’ and Exporters’ Real Exchange Rates Relative to the U. S. Dollar, 1971–761

(In per cent)

article image

It should be recalled that exchange rates, for the purpose of this study, are in U. S. dollars per unit of domestic currency. Thus, a positive figure in the table denotes an appreciation of exporters’ (real) exchange rates vis-à-vis the U. S. dollar.

Table 5.

Estimated Changes in World Prices, 1971–76 1

(In per cent)

article image

Simulation using historical real exchange rate changes. World prices expressed in U. S. dollars.

The next step is to combine the individual country supply elasticities with the change in world prices (and the estimated domestic rates of inflation) in order to produce the percentage change in export earnings for each of the five commodities. These can then be aggregated according to each commodity’s share in total exports for that country to derive the changes in each country’s total export earnings. The results for Chile, Peru, Zaïre, and Zambia are presented in Tables 6, 7, 8, and 9, respectively. While there are some differences across commodities, countries, and years, these results strongly suggest on balance that the pattern of historical exchange rate changes in 1971–76 led to an increase in export earnings for the four copper producing countries (with Zaïre being the only subject country not showing increases in total export earnings in five of the six years).

Table 6.

Chile: Estimated Changes in Export Receipts, 1971–76 1

(In per cent)

article image

Simulation using historical real exchange rate changes. Export receipts expressed in U. S. dollars.

Table 7.

Peru: Estimated Changes in Export Receipts, 1971–76 1

(In per cent)

article image

Simulation using historical real exchange rate changes. Export receipts expressed in U. S. dollars.

Table 8.

Zaïre: Estimated Changes in Export Receipts, 1971–76 1

(In per cent)

article image

Simulation using historical real exchange rate changes. Export receipts expressed in U. S. dollars.

Table 9.

Zambia: Estimated Changes in Export Receipts, 1971–76 1

(In per cent)

article image

Simulation using historical real exchange rate changes. Export receipts expressed in U. S. dollars.

Turning to the effects of the exchange rate changes on imports in the four copper producing countries, it is first necessary to calculate the change in import prices facing these countries. As noted earlier, these import price changes are obtained by weighting the export price changes of the country’s trading partners (predicted by the industrial country MERM) by its import shares with them. The results are shown in Table 10.

Table 10.

Estimated Changes in Import Prices, 1971–76 1

(In per cent)

article image

Simulation using historical real exchange rate changes. Import prices measured in terms of U. S. dollars.

Given the assumed parameters in Tables 1, 2, and 3, and given the estimated changes in import prices and export earnings, the change in each country’s imports can be calculated as in equation (12). These are given in Table 11. Since changes in export earnings have a positive effect on import demand in the model, the percentage changes in imports usually have the same signs as the changes in total export earnings (the correspondence is not exact because other factors also affect imports). Further, because only part of the change in export earnings finds its way into the demand for imports (i.e., the elasticities of expenditure with respect to income and of imports with respect to expenditure are generally both less than one), the induced change in imports (in value if not in percentage terms) is usually less than the induced change in export earnings. 24

Table 11.

Estimated Changes in Value of Imports, 1971–76 1

(In per cent)

article image

Simulation using historical real exchange rate changes. Value of imports measured in terms of U. S. dollars.

Combining the changes in export receipts and in import expenditure yields the estimated change in the balance of trade for each of the four countries, as shown in Table 12. Again, it is difficult to generalize across years but we find that changes in the trade balance tend to be dominated by changes in export receipts in the cases of Zambia and Zaïre, and by changes in import expenditures in the case of Chile; 25 no dominant result emerges for Peru. The results in Table 12 also suggest that the historical pattern of (real) exchange rate changes in 1971—76 did not generate the same trade balance effects for the four countries studied here, with Zambia recording significantly more favorable trade balance changes than either Zaïre or Peru, and with Chile emerging as the most adversely affected country. In-depth country analysis would no doubt make it possible to trace these intercountry (trade balance) differences to differences in rates of domestic inflation, differences in rates of depreciation, and differences in commodity compositions of exports.

Table 12.

Estimated Changes in Balance of Trade: Real Exchange Rate Simulation, 1971–76

(In millions of U. S. dollars)

article image

Trade data for Zaïre for 1976 are not yet available, so computation of balance of trade changes for that year was not possible.

HISTORICAL NOMINAL EXCHANGE RATE CHANGES, 1971–76

The next simulation examines the same period as the previous one but constrains domestic rates of inflation to be zero for all countries in the model. As mentioned earlier, the use of nominal exchange rate changes was the prevailing practice in previous models of this type. 26 It was also pointed out that using nominal exchange rate changes would be likely to produce misleading (unduly optimistic) estimates of trade balance changes for those countries and those periods for which domestic rates of inflation are high. A comparison of the results in Table 13 (for nominal exchange rate changes) with those in Table 12 (for real exchange rate changes) supports this contention. The trade balance changes for the low-inflation country in the sample (Zambia) are considerably more favorable for real than for nominal exchange rate changes, whereas the opposite result holds for the high-inflation country (Chile). For a given country, the differences between the real and nominal simulation results are generally greatest (as expected) in those years when domestic rates of inflation are most abnormal; for example, the large difference in 1973 vis-à-vis 1972 for Chile reflects the increase in Chile’s inflation rate from about 73 per cent in 1972 to about 320 per cent in 1973. The difference between the real and nominal exchange rate shows up not only in the absolute dollar amounts but also in the sign of the trade balance changes. In fact, of the 69 trade balance changes that appear in Tables 12 and 13, more than half (39) have a different sign in the real exchange rate case than in the nominal case. While these results are not surprising, they do emphasize the importance of considering exchange rate changes in conjunction with domestic rates of inflation when dealing with primary producing countries.

Table 13.

Estimated Changes in Balance of Trade: Nominal Exchange Rate Simulation, 1971–76

(In millions of U. S. dollars)

article image

Trade data for Zaïre in 1976 are not currently available, so computation of balance of trade changes for that year was not possible.

INCREASE IN WORLD PRICE ELASTICITY OF DEMAND FOR COPPER

In the third simulation the model is used to estimate the trade balance consequences of a rise (specifically, a tripling) in the world price elasticity of demand for copper. Such an increase could result, for example, from the discovery of a new technology that would allow another metal to be more easily substituted for copper.

For the purpose of the calculations, the world copper demand price elasticities were increased to –0.32, –0.74, and –1.0 after one, two, and three years, respectively. The increase was assumed to take place in 1974, so that the results prior to 1974 are exactly the same as in the previous simulation. As can be seen by examining equation (6), an increase in the world demand price elasticity for copper has two opposing effects on the change in world copper prices. First, since the demand elasticity (ndi) is multiplied by the change in importers’ exchange rates (), an increase in the elasticity leads to a more pronounced shift in the world demand curve for copper for any given ; this demand-shift effect will work to increase the size of the world price change. Second, and working in the opposite direction, an increase in the demand elasticity leads to a flatter world demand curve for copper, and this acts to reduce the size of the world price change (i.e., an increase in ndi increases the denominator in equation (6)). 27 One might call this the demand-slope effect. The net effect on the world price of copper thus depends on which of these two effects dominates.

With these considerations in mind, interpretation of the results in Table 14 becomes straightforward. This table indicates that estimated increases in world copper prices were larger under the high demand elasticity assumption in both 1974 and 1975, but smaller in 1976. The reason for this pattern of outcomes is that the demand-shift effect dominates the demand-slope effect in 1974 and 1975, whereas the reverse is true in 1976. The evidence for this can be seen by referring to Table 4, which gives the weighted changes in importers’ real exchange rates (). Specifically, note that for copper carries values of 10.7 per cent and 16.1 per cent for 1974 and 1975, respectively, but only 1.1 per cent for 1976. In the earlier two years, the rightward (outward) shift of the world demand curve for copper (cum the higher demand elasticity) is more than sufficient to counter the flatter slope of this demand curve induced by the higher elasticity; but in 1976 the shift is only about 1 per cent, and this is not enough to offset the tripling of the demand elasticity. In more general terms, this simulation demonstrates that moderate changes in either the (global) demand or supply price elasticity for primary commodities will usually have ambiguous implications for the size of world price changes for these commodities.

Table 14.

Estimated Changes in World Copper Prices, Low Versus High World Price Elasticity of Demand, 1974–76 1

(In per cent)

article image

World copper prices expressed in U. S. dollars.

The low-elasticity results are identical with those reported for copper in Table 5.

INDICES OF EFFECTIVE EXCHANGE RATES

As indicated at the beginning of this paper, when many exchange rates change simultaneously it is usually desirable to weight the various exchange rate changes according to some systematic criterion so that the average exchange rate change against all other currencies (i.e., the effective exchange rate) can be calculated for a given country. Two general methods of weighting are most popular: one assigning weights according to bilateral trade patterns, the other using weights derived from some general equilibrium trade model, such as the Fund’s industrial country MERM. 28 Other methods, such as global trade weights, are also occasionally used.

The approach of using bilateral trade weights, be they import shares, export shares, or some average of the two, is clearly the simpler method. The information required to construct trade-weighted effective exchange rates is readily available from published trade data (e.g., the Fund’s Direction of Trade) and the approach is reasonably flexible as to the choice of base period, the partner countries included, etc. The difficulty with using bilateral trade weights, at least in estimating the effects of exchange rate changes on a country’s trade balance, is that they ignore several important factors that determine the “true” effect on the trade balance. Specifically, weights based solely on the size of trade flows ignore (i) the demand and supply responses (by both the country and its trading partners) to exchange rate changes, (ii) the effect of these exchange rate induced supply and demand responses on the prices of imports, exports, and domestic goods, and (iii) the effect of exchange rate changes on trade flows of third countries that have not at that time changed their exchange rates.

Table 15.

Indices of Effective Real Exchange Rates, 1971–76

(1970 = 100)

article image

Trade data for Zaïre were not available for 1976, thereby precluding computation of both balance of trade and effective exchange rate changes for that year.

The practical upshot of the exclusions listed above is that bilateral trade weights will accurately indicate the importance of particular exchange rate changes on a country’s trade balance only under certain conditions, few of which are likely to be satisfied in most countries. The conditions necessary for equality between bilateral weights and weights derived in a general equilibrium framework will vary both with the assumed structure of the import and export equations and with the particular type of bilateral trade weights considered (i.e., import-weighted, export-weighted, or an average of the two). For example, Rhomberg (1976) has indicated that for bilateral trade weights (import- and export-weighted) to equal the weights in the industrial country MERM, the following conditions on price elasticities of supply and demand would be necessary (but not sufficient): (i) all price elasticities of supply of traded goods are infinite, (ii) all price elasticities of demand for a country’s imports and exports are equal to one another (regardless of commodity or country of origin or destination), and (iii) all elasticities of a country’s exports with respect to prices of exports of other countries are zero. Similarly, using an assumed structure where there are no expenditure or cross-price effects and where the country is a price taker for both its imports and exports, Thakur (1975) has demonstrated that MERM weights will be equal to average export- and import-weighted bilateral trade weights only when the price elasticity of demand for the country’s imports is equal to two and when its price elasticity of supply for exports is zero. Thus, while the size of the errors is not yet firmly established, there are good a priori reasons for suspecting that using trade-weighted effective exchange rates (in preference to MERM-weighted effective exchange rates) could involve a considerable sacrifice of accuracy for convenience.

Table 16.

Indices of Effective Nominal Exchange Rates, 1971–76

(1970 = 100)

article image

Trade data for Zaïre were not available for 1976, thereby precluding computation of both balance of trade and effective exchange rate changes for that year.

The indices of effective exchange rates that are generated by the model for each of the four copper producing countries are presented in Tables 15 and 16 for real and nominal exchange rates, respectively. Following Artus and Rhomberg (1973), the change in a country’s effective exchange rate is defined as the unilateral change in that country’s exchange rate that, all other factors held constant, would produce the same trade balance change as that estimated to have taken place as a result of all exchange rate changes during that period.

According to our estimates, the effective real exchange rates of Chile, Peru, and Zaïre appreciated (on average) over the 1971–76 period, whereas the effective rate of Zambia depreciated. 29 This is to be expected given the fact that our earlier estimates of the trade balance effects of historical real exchange rate changes over 1971–76 indicated trade balance deteriorations (on average) for Chile, Peru, and Zaïre and trade balance improvements (on average) for Zambia (see Table 12). Thus, Chile, Peru, and Zambia would require unilateral (real) exchange rate appreciations to match these estimated negative trade balance changes, while Zambia would require a unilateral real depreciation to match the estimated improvements in its trade balance. In contrast, the nominal effective rate indices in Table 16 point to depreciations for Chile and Peru, and appreciations for Zambia and Zaïre. These movements in effective rates again follow from the earlier estimated trade balance changes (see Table 13 for the nominal case), and the differences in direction between the real and nominal cases mirror the different signs of the estimated trade balance changes under the real and nominal cases (compare Table 12 with Table 13). The reason why the movement in the real index is sometimes much larger (smaller) than that in the nominal index is that the estimated trade balance change in the real case is sometimes much larger (smaller) than that in the nominal case; for example, compare the very large estimated trade balance changes for Chile in the real case (Table 12) with the smaller changes in the nominal case (Table 13); in such instances, the unilateral exchange rate changes necessary to reproduce these large trade balance changes will also be large.

IV. Concluding Remarks

This paper has introduced a framework for analyzing the effect of a set of exchange rate changes on the trade balances of a group of primary producing countries. The model that was used draws rather heavily on earlier work in that it retains the commodity-by-commodity approach to export earnings, views exchange rate changes as leading to weighted shifts in global export supply and import demand schedules, and permits export earnings to affect import demand through its effect on domestic expenditure. 30 At the same time, the model goes beyond earlier work in several respects, perhaps the most important of which is the explicit introduction of monetary policy as a second instrument that is used in conjunction with the exchange rate. This modification enables the model to study the medium-run trade balance effects of real rather than just nominal exchange rate changes—a significant advantage when dealing with countries that sometimes record very high rates of inflation.

The basic workings of the model were illustrated by applying it to four copper producing countries, namely, Chile, Peru, Zambia, and Zaïre. Specifically, three simulations with the model were performed. The first simulation calculated the effects of the historical pattern of real exchange rate changes during the 1971–76 period on the world prices of five primary commodities, on total export earnings, on import expenditures, and on the trade balance for each of the four countries. The second simulation carried out the same exercise for nominal exchange rates and demonstrated the importance of using real rates. The third simulation dealt with the trade balance effects of a hypothetical increase in the world price elasticity of demand for copper.

The model is flexible enough to accommodate a wide variety of other simulation exercises, as well as to withstand some changes in underlying assumptions that would make it more consistent with the characteristics of some other primary producing countries. For example, one could use the model to estimate the trade balance implications of a round of competitive devaluations by a group of primary producing countries, or to calculate the world price changes resulting from supply changes in a given primary producing country. Similarly, one might wish to relax the complete sterilization assumption to permit some feedback from exchange rate changes to the money supply, or to specify different inflation determination processes for tradable and non-tradable goods, etc. 31 As the model undergoes further application and development, it would be expected that its usefulness for country work would increase.

BIBLIOGRAPHY

  • Aghevli, Bijan, and Mohsin S. Khan (1976), “Credit Policy and the Balance of Payments in Developing Countries” (unpublished, International Monetary Fund, December 27, 1976).

    • Search Google Scholar
    • Export Citation
  • Aghevli, Bijan, and Mohsin S. Khan (1978), “Government Deficits and the Inflationary Process in Developing Countries,” Staff Papers, Vol. 25 (1978), pp. 383416.

    • Search Google Scholar
    • Export Citation
  • Aghevli, Bijan, and Carlos Alfredo Rodriguez (1979), “Trade, Prices, and Output in Japan: A Simple Monetary Model,” Staff Papers, Vol. 26 (1979), pp. 3854.

    • Search Google Scholar
    • Export Citation
  • Armington, Paul S. (1969), “A Theory of Demand for Products Distinguished by Place of Production,” Staff Papers, Vol. 16 (1969), pp. 15978.

    • Search Google Scholar
    • Export Citation
  • Artus, Jacques, and Anne K. McGuirk (1978), “A Revised Version of the Multilateral Exchange Rate Model” (unpublished, International Monetary Fund, May 16, 1978).

    • Search Google Scholar
    • Export Citation
  • Artus, Jacques, and Rudolf R. Rhomberg (1973), “A Multilateral Exchange Rate Model,” Staff Papers, Vol. 20 (1973), pp. 591611.

  • Askari, Hossein, and John Thomas Cummings (1977), “Estimating Agricultural Supply Response with the Nerlove Model: A Survey,” International Economic Review, Vol. 18 (June 1977), pp. 25792.

    • Search Google Scholar
    • Export Citation
  • Banks, Ferdinand E. (1974), The World Copper Market: An Economic Analysis (Cambridge, Massachusetts, 1974).

  • Bautista, Romeo M. (1977), “Effects of Major Currency Realignment on Philippine Merchandise Trade,” Review of Economics and Statistics, Vol. 49 (May 1977), pp. 15260.

    • Search Google Scholar
    • Export Citation
  • Behrman, Jere Richard (1975 a), Foreign Trade Regimes and Economic Development: Chile (National Bureau of Economic Research, New York, 1975).

    • Search Google Scholar
    • Export Citation
  • Behrman, Jere Richard (1975 b), “Econometric Modeling of National Income Determination in Latin America, with Special Reference to the Chilean Experience,” Annals of Economic and Social Measurement, Vol. 4, No. 4 (1975), pp. 46188.

    • Search Google Scholar
    • Export Citation
  • Bélanger, Gerard (1976), “An Indicator of Effective Exchange Rates for Primary Producing Countries,” Staff Papers, Vol. 23 (1976), pp. 11336.

    • Search Google Scholar
    • Export Citation
  • Black, Stanley W. (1976), “Exchange Policies for Less Developed Countries in a World of Floating Rates,” in The International Monetary System and the Developing Nations, ed. by Danny M. Leipziger (Bureau for Program and Policy Coordination, Washington, 1976), pp. 1373.

    • Search Google Scholar
    • Export Citation
  • Blejer, Mario I. (1977), “The Short-run Dynamics of Prices and the Balance of Payments,” American Economic Review, Vol. 67 (June 1977), pp. 41928.

    • Search Google Scholar
    • Export Citation
  • Branson, William H. (1972), “The Trade Effects of the 1971 Currency Realignments,” Brookings Papers on Economic Activity: 1 (1972), pp. 1558.

    • Search Google Scholar
    • Export Citation
  • Burrows, James C. (1971), Cobalt: An Industry Analysis (Lexington, Massachusetts, 1971).

  • Burrows, James C. and M. Lonoff, Charles River Associates Models of Metal Markets,Paper presented at Ford Foundation Conference on Stabilizing World Commodity Markets, Arlee, Virginia, March 1977.

    • Search Google Scholar
    • Export Citation
  • Charles River Associates (1969), An Economic Analysis of the Cobalt Industry, Prepared for the General Services Administration (Cambridge, Massachusetts, 1969).

    • Search Google Scholar
    • Export Citation
  • Chu, Ke-Young, and Andrew Feltenstein (1978), “Relative Price Distortions and Inflation: The Case of Argentina, 1963–76,” Staff Papers, Vol. 25 (1978), pp. 45293.

    • Search Google Scholar
    • Export Citation
  • Clark, Peter B. (1977), “The Effects of Recent Exchange Rate Changes on the U.S. Trade Balance,” in The Effects of Exchange Rate Adjustments, Proceedings of a Conference Sponsored by OASIA Research, Department of the Treasury, April 4 and 5, 1974, ed. by Peter B. Clark, Dennis E. Logue, and Richard James Sweeney (Washington, 1977), pp. 20136.

    • Search Google Scholar
    • Export Citation
  • Fisher, Franklin M., Paul H. Cootner, and Martin Neil Baily (1972), “An Econometric Model of the World Copper Industry,” Bell Journal of Economics and Management Science, Vol. 3 (Autumn 1972), pp. 568609.

    • Search Google Scholar
    • Export Citation
  • Frenkel, Jacob A., and Harry G. Johnson, eds. (1976), The Monetary Approach to the Balance of Payments (University of Toronto Press, 1976).

    • Search Google Scholar
    • Export Citation
  • Goldstein, Morris, and Mohsin S. Khan (1978), “The Supply and Demand for Exports’. A Simultaneous Approach,” Review of Economics and Statistics (May 1978), pp. 27586.

    • Search Google Scholar
    • Export Citation
  • Hemphill, William L. (1974), “The Effect of Foreign Exchange Receipts on Imports of Less Developed Countries,” Staff Papers, Vol. 21 (1974), pp. 63777.

    • Search Google Scholar
    • Export Citation
  • Hirsch, Fred, and Use Higgins (1970), “An Indicator of Effective Exchange Rates,” Staff Papers, Vol. 17 (1970), pp. 45387.

  • International Monetary Fund (1977), The Monetary Approach to the Balance of Payments (Washington, 1977).

  • Isard, Peter (1977), “The Price Effects of Exchange-Rate Changes,” in The Effects of Exchange Rate Adjustments, Proceedings of a Conference Sponsored by OASIA Research, Department of the Treasury, April 4 and 5, 1974, ed. by Peter B. Clark, Dennis E. Logue, and Richard James Sweeney (Washington, 1977), pp. 36986.

    • Search Google Scholar
    • Export Citation
  • Khan, Mohsin S. (1974), “Import and Export Demand in Developing Countries,” Staff Papers, Vol. 21 (1974), pp. 67893.

  • Labys, Walter C. (1975), ed., Quantitative Models of the Commodity Markets (Cambridge, Massachusetts, 1975).

  • Lipschitz, Leslie (1979), “Exchange Rate Policy for a Small Developing Country, and the Selection of an Appropriate Standard,” Staff Papers, Vol. 26 (1979), pp. 42349.

    • Search Google Scholar
    • Export Citation
  • Paljarvi, Jukka, and Massimo Russo (1977), “The Demand for Money in Zambia” (unpublished, International Monetary Fund, 1977).

  • Rhomberg, Rudolf R. (1976), “Indices of Effective Exchange Rates,” Staff Papers, Vol. 23 (1976), pp. 88112.

  • Richard, Denis (1978), “A Dynamic Model of the World Copper Industry,” Staff Papers, Vol. 25 (1978), pp. 779833.

  • Ridler, Duncan, and Christopher A. Yandle (1972), “A Simplified Method for Analyzing the Effect of Exchange Rate Changes on Exports of a Primary Commodity,” Staff Papers, Vol. 19 (1972), pp. 55978.

    • Search Google Scholar
    • Export Citation
  • Segura, Edilberto Leoncio (1973), An Econometric Study of the Fish Meal Industry (Food and Agriculture Organization of the United Nations, New York, 1973).

    • Search Google Scholar
    • Export Citation
  • Teigeiro, Jose (1977), “The Balance of Payments Performance of Latin America and the Caribbean, 1971–76,” a paper presented at the XIV Meeting of Central Bank Technicians of the American Continent, in San Carlos de Bariloche, Argentina, November 611, 1977.

    • Search Google Scholar
    • Export Citation
  • Thakur, Subhash M. (1975), “A Note on the Concept of Effective Exchange Rate” (unpublished, International Monetary Fund, July 18, 1975).

    • Search Google Scholar
    • Export Citation
  • Yandle, Christopher A. (1977), “Using the Commodity Method for Analyzing the Effects of Exchange Rate Changes—Some Considerations,” in The Effects of Exchange Rate Adjustments, Proceedings of a Conference Sponsored by OASIA Research, Department of the Treasury, April 4 and 5, 1974, ed. by Peter B. Clark, Dennis E. Logue, and Richard James Sweeney (Washington, 1977), pp. 34154.

    • Search Google Scholar
    • Export Citation
*

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.

Mr. Goldstein, Assistant Chief of the Special Studies Division of the Research Department, is a graduate of Rutgers University and of New York University. He was formerly a Research Fellow in Economics at the Brookings Institution. He is currently on a leave of absence from the Fund at the Office of International Monetary Research, U.S. Treasury.

Ms. Schadler, economist in the Special Studies Division of the Research Department, holds degrees from Mount Holyoke College and the London School of Economics and Political Science.

1

To mention specific figures, the five primary commodities (copper, cobalt, iron ore, fish meal, and zinc) considered in this paper accounted for 97 per cent of Zambia’s total export earnings in 1976. The corresponding figures for Chile, Peru, and Zaïre were 64 per cent, 47 per cent, and 70 per cent, respectively.

2

For example, the (geometric) mean rate of consumer price inflation over the four years 1973–77 was 246 per cent in Chile, 27 per cent in Peru, 33 per cent in Zaïre, and 15 per cent in Zambia. The corresponding (GNP-weighted) inflation figures for the 14 industrial countries was 10 per cent. The link between inflation rates and money supply changes is strongly suggested by the analogous figures for money supply changes over this period, namely, 225 per cent for Chile, 25 per cent for Peru, 33 per cent for Zaïre, 15 per cent for Zambia, and 8 per cent for the industrial countries. These figures are taken from International Monetary Fund, International Financial Statistics (various issues).

3

The external terms of trade means the ratio of a country’s export prices to its import prices. In contrast, the internal terms of trade refers to the relationship between the price of tradable goods and the price of nontradable goods. Since for many primary producing countries the external terms of trade are fixed in the world market, exchange rate changes can have a real impact in the economy only by (temporarily) altering the country’s internal terms of trade, i.e., by altering the profitability of exporting relative to other activities in the economy.

4

The industrial country multilateral exchange rate model (hereafter referred to as industrial country MERM) developed by Artus and Rhomberg (1973) estimated the impact of exchange rate changes under the assumption that the authorities in each country adjusted monetary and fiscal policy so as to hold real output constant. Our model is similar in spirit to the Artus-Rhomberg model in that the effect of exchange rate changes can be estimated only when the path of monetary policy is known. It is, however, more general than the Artus-Rhomberg model, since by not restricting the growth of the money supply to only one value it permits the calculation of exchange rate effects under alternative assumptions about money supply growth (with the Keynesian-neutral, constant real output case as only one of many possibilities).

5

This stipulation about the medium run is important because in the long run all relative prices in the economy will be determined by real factors, and rates of change of exchange rates and money supplies will be jointly determined. Thus, while in the short run a country can set its exchange rate at some desired level and eliminate unwanted changes in its money supply by sterilizing inflows or outflows of foreign exchange, abnormal inflation rates will in the long run lead to either a depletion of foreign exchange reserves or a money supply backed entirely by foreign reserves. Ultimately, trade or capital restrictions will have to be introduced or intensified, or the exchange rate will have to be changed. At the other end of the spectrum, if the time period is too short, there will not be enough time for most consumers and producers to react to the relative price changes induced by exchange rate changes.

6

In the notation, the world price of commodities in terms of primary producing countries’ currencies is always expressed in real terms. For example, the world copper price in terms of U. S. dollars is converted into domestic PPC currency units (by the appropriate PPC currency/dollar exchange rate) and then deflated by the domestic PPC price level.

7

As noted earlier, another quite different argument for using the overall domestic price level as the deflator in the export supply function is that it permits exchange rate changes to affect at least one relative price in the economy even if the primary producing country’s external terms of trade are fixed in the world market. Specifically, because domestic prices include nontraded goods, an exchange rate change can affect trade flows by altering the relative price of traded to nontraded goods (i.e., the internal terms of trade).

8

In the case of Zambia, we did have access to an index of mining company costs for copper, albeit only for the period 1971–76. It was found that the simple correlation between this index and the consumer price index was about 0.9 (using annual observations for both series).

9

See, for example, Goldstein and Khan (1978) where trend real income, as well as relative export price, is included in the export supply equation.

10

Many models of import demand in developing countries ignore relative price effects since technological and institutional constraints are said to limit their importance. To make the model applicable to as wide a variety of countries as possible, however, a relative price argument has been included. Estimates of aggregate import demand functions for a number of less developed countries by Khan (1974) indicate a significant role for relative prices in many cases.

11

Import expenditure can, of course, vary positively with the increase in import prices if the elasticity of import demand with respect to relative import prices is inelastic (less than unity).

12

In simulation exercises it is unfortunately not possible to handle inflation determination symmetrically between primary producing countries and industrial countries. In the first place, the industrial country MERM employs far more restrictive assumptions about monetary policy than those applied here for primary producers (see footnote 4). Second, even if this was not the case, the estimation of money-demand equations for each of the industrial countries would be beyond the practical scope of this paper. In view of this constraint, the actual rate of inflation in industrial countries was used for the real exchange rate simulations that are reported in Section III. The reasoning was that it was less harmful to have a theoretical asymmetry in the explanation of inflation (i.e., exogenous and actual for industrial countries versus endogenous and estimated for the four primary producing countries) than to impose a restrictive set of assumptions about monetary policy in the four subject countries.

14

In equations (14)–(16), Mo is the nominal money stock, PE is the expected price level, the superscript D denotes demand, and Δ represents the first-difference operator.

15

Following Hirsch and Higgins (1970, p. 454), one can think of the effective exchange rate in general terms as the “total relationship between the given currency and all others.”

16

In addition to these problems, it should also be mentioned that the model (like its predecessors) is unable to distinguish expected from unexpected exchange rate changes, nor does it account for the possibility that the structural coefficients may change over time as producers and consumers alter their behavior in response to past policy actions of the authorities. Suffice it to say that both these problems are very difficult to deal with.

17

See Richard (1978) for a good discussion of these issues.

18

See Clark (1977) for an application of this approach, albeit to an industrial country.

19

Different methods of measuring foreign exchange receipts, as well as their role in the import demand function, are discussed in Hemphill (1974). Teigeiro (1977) found, however, that export earnings did as well or better than net foreign assets in explaining the demand for imports in the majority of 22 Latin American and Caribbean countries.

20

See, for example, Behrman’s (1975 a) work on Chile’s imports and exports.

21

For example, data for Zambia and Zaïre are generally available only for 1965–75, and then usually only on an annual basis.

22

The estimated demand-for-money equations that were used to generate estimates of the domestic rate of inflation are available upon request from the authors, whose address is Research Department, International Monetary Fund, Washington, D.C. 20431. For Peru, we used an equation developed by the Bank of Peru. For Zambia, we used the estimates developed by Paljarvi and Russo (1977). Since good equations did not seem to be available for Chile and Zaïre, we produced our own estimates based on the model described in equations (14)–(16).

23

The commodity studies used to select values for the parameters in Tables 1 and 2 included Askari and Cummings (1977), Banks (1974), Bélanger (1976), Charles River Associates (1969), Fisher, Cootner, and Baily (1972), Labys (1975), Richard (1978), and Segura (1973). The studies by Aghevli and Khan (1976), Behrman (1975), Bélanger (1976), Hemphill (1974), Khan (1974), and Teigeiro (1977) proved useful in selecting values for the behavioral parameters in Table 3.

24

If the value of imports is low relative to the value of exports in a given year, then a smaller change in the value of imports than of exports could still be consistent with a larger percentage change in import value.

25

Bélanger (1976) came to the same conclusion about the dominance of export receipts in the cases of Zaïre and Zambia.

27

Recall that ndi is defined to be negative, so that the denominator in equation (6) is positive.

28

See Artus and Rhomberg (1973) and Artus and McGuirk (1978) for an explanation of the effective exchange rates calculated from the industrial country MERM. See Lipschitz (1979) for an evaluation of alternative bilateral and multilateral weighting schemes.

29

These indices of effective exchange rates in Tables 15 and 16 are obviously subject to margins of error as wide as those associated with the estimated trade balance changes. For example, there is a specific source of error introduced into the real effective rate indices by the fact that the model uses estimated as opposed to actual inflation rates in the four copper producing countries.

30

Although a review of the literature is not included, the model can be viewed as a descendent of earlier work done at the Fund by Armington (1969), Artus and Rhomberg (1973), Bélanger (1976), and Ridler and Yandle (1972). It is also related to recent work done outside the Fund on the effect of exchange rate changes on trade flows by Branson (1972), Black (1976), Bautista (1977), Clark (1977), and Isard (1977).

31

The feedback effect of the balance of payments on the money supply is, of course, widely recognized in the literature; see, for example, Frenkel and Johnson (1976) and International Monetary Fund (1977). For good treatments of the tradable/nontradable distinction within the context of a monetary model of inflation and the balance of payments, see Aghevli and Rodriguez (1979) and Blejer (1977).