The World Model of Merchandise Trade: Simulation Applications
Author:
Duncan Ripley
Search for other papers by Duncan Ripley in
Current site
Google Scholar Close
Publication Date:
01 Jan 1980
eISBN:
9781451946864
Language:
English
Keywords:
SP; mixed-citation publication-type; terms of trade effect; competitor price; spot commodity price; model price elasticity; trade balance effect; Manufacturing; Terms of trade; Imports
Download PDF (1.8 MB)

The world model of merchandise trade as presently specified1 focuses primarily on the impact of levels of economic activity in the industrial countries on their foreign trade balances, although the effects of changes in relative prices are also considered. Raw material and fuel imports and transactions in manufactured goods are especially important in the trade of these countries, and the determinants of these flows, as well as those of agricultural goods, are quite distinct. As a result, the model distinguishes between these four types of commodity flows but pays particular attention to the determinants of trade in fuels, raw materials, and manufactured goods.

Abstract

The world model of merchandise trade as presently specified1 focuses primarily on the impact of levels of economic activity in the industrial countries on their foreign trade balances, although the effects of changes in relative prices are also considered. Raw material and fuel imports and transactions in manufactured goods are especially important in the trade of these countries, and the determinants of these flows, as well as those of agricultural goods, are quite distinct. As a result, the model distinguishes between these four types of commodity flows but pays particular attention to the determinants of trade in fuels, raw materials, and manufactured goods.

The world model of merchandise trade as presently specified1 focuses primarily on the impact of levels of economic activity in the industrial countries on their foreign trade balances, although the effects of changes in relative prices are also considered. Raw material and fuel imports and transactions in manufactured goods are especially important in the trade of these countries, and the determinants of these flows, as well as those of agricultural goods, are quite distinct. As a result, the model distinguishes between these four types of commodity flows but pays particular attention to the determinants of trade in fuels, raw materials, and manufactured goods.

The model thus far has been used mainly for simulations. The data requirements for simulations are typically smaller than for forecasts, since estimates for all exogenous variables for the forecast period are not required. Simulation does not deal with what is likely to happen, but rather with what would happen under alternative sets of exogenous model inputs. It is useful in analyzing the “normal” effect of certain policy changes or certain exogenous disturbances on the endogenous variables in the model, since it is possible to hold other variables constant. The magnitude of these impacts can be studied to see whether they appear realistic, and partial multiplier tables can be constructed for simple model applications. Because of their relative simplicity compared with forecasting applications, and because of the insight they provide to the reasonableness of the model’s structure and coefficients, this paper focuses on the results of model simulations. In Section I the structure of the model is reviewed, and certain new relationships that are needed for simulations and forecasting are given. Simulations of the effects of changes in activity levels are described in Section II, and simulations of terms of trade effects are presented in Section III. Finally, in Section IV the paper is summarized and suggestions for further work to enhance the simulation capabilities of the model are considered.

I. General Model Structure

The world model of merchandise trade is a semiannual model. The relationships explaining the behavior of export and import unit value indices and export and import volumes, which were fully described in the earlier paper (see footnote1), are very briefly reviewed in Section II, below.

A number of additional relationships are needed, however, if the model is to provide useful simulation results or be used in a forecasting mode. These relationships were alluded to in the first presentation of the model but are described in detail in Section II, below. Some of these are required for such simulations as the relationships between sport commodity prices and the level of economic activity in the industrial countries; others are needed both for simulations and for forecasting variables for which exogenous forecasts are not otherwise readily available. Finally, sets of relationships are necessary for “accounting” purposes, that is, to convert trade data from a customs basis to a balance of payments basis.

Review of Basic Model

The model of merchandise trade is disaggregated by country or country grouping and by commodity. The individual countries that are covered are 14 industrial countries, Austria, Belgium-Luxembourg, Canada, Denmark, France, the Federal Republic of Germany, Italy, Japan, the Netherlands, Norway, Sweden, Switzerland, the United Kingdom, and the United States. The model is closed in that all countries are covered either individually or within one of four country groupings—other developed countries, centrally planned economies, major oil exporting countries, and other developing countries. 2

Trade volumes and prices are disaggregated into four commodity classes: foods, raw materials, fuels, and manufactures. Disaggregation is felt to be desirable because the determinants of trade volumes and prices are thought to vary by commodity class. For example, most raw material imports are used as intermediate inputs, whereas many manufactured imports enter directly into final demand. Disaggregation also offers the advantage that exogenous information obtained from country and commodity specialists in the International Monetary Fund can be more easily incorporated in forecasting applications of the model. The benefits arising from disaggregation have to be weighed against the disadvantages of manipulating a much larger model and the difficulties of collecting data at a disaggregative level. In view of this trade-off, the commodity disaggregation is limited to the four classes given above.

The model can be thought of as consisting of three principal sets of relationships. Of major importance are the sets of relationships explaining movements in export and import unit values and those explaining real trade flows for the industrial countries. Of lesser interest, given the focus and specification of the model, are the unit value and volume relationships for the four regions.

In the first set of relationships, export and import unit values are determined for each industrial country and for each commodity class, excluding fuels. Movements in export unit values for manufactures for a particular industrial country, expressed in its own currency, are explained by the domestic cost of labor (the cyclical and noncyclical elements of unit labor costs), 3 by the costs of raw materials to domestic producers, and by the state of demand abroad represented by competitor prices for manufactures in foreign markets (expressed in the currency of the exporting country). Competitor prices, domestic raw material prices, and the cyclical element of unit labor costs are exogenous to the individual equations, but are endogenous to the model as a whole, while normalized unit labor costs are exogenous. Movements in import unit values for manufactures are explained by a weighted average of export unit values of partner countries, with weights reflecting the relative importance of imports from a particular country in the base year, 1970.

The basic exogenous data used to explain movements in unit value indices for nonmanufactures are 35 time series representing spot quotations in the world market for precisely defined primary commodities. These data are aggregated to obtain spot price indices for agricultural goods and raw materials, with compositions as similar as possible to those of the export or import trade of the particular country. These spot indices are the principal exogenous explanatory variables for export unit values for agricultural goods and raw materials, although an additional variable representing the domestic costs of processing these commodities is also included. For import unit values, two price series enter as explanatory variables—the import weighted spot price index, representing the price of imports of these commodities from the nonindustrial countries, and a weighted average of the export unit value indices for these commodities. For forecasts, projected spot prices are obtained from the commodity experts at the Fund. For simulations, however, a supplementary link has been developed to reflect the effect of variations in activity levels in the industrial countries on spot commodity prices. This supplementary link is described in Section II, below.

The activity level in the manufacturing sector plays an important role in determining real import demand; it enters in current and lagged form in the model’s specification of the equations for the volume of fuel and raw material imports, and as an important component in the demand for imports of manufactured goods. The raw material and fuel import equations reflect the view that imports of these commodities will be used, for the most part, as intermediate inputs in the production process. Because a substantial share of manufactured imports enters into final demand directly, the basic “demand” variable in the import volume equation for manufactures is constructed as a weighted average of final domestic demand for manufactures and of the output in the manufacturing sector, both expressed in index form, with weights reflecting the shares of manufactured imports going directly to final demand and to intermediate demand, respectively. The basic demand variable introduced in the equation explaining agricultural imports is real private consumption expenditure. Market variables in the export equations for agricultural goods and raw materials are weighted averages of the volume of imports of these goods in partner countries.

The price of manufactured goods produced domestically relative to the price of manufactured goods produced abroad plays an important role in determining the demand for imports of manufactured goods. In contrast, the export price of manufactured goods for a particular country relative to export prices in competitor countries plays an important role in explaining the volume of exports of manufactured goods. A price-relative, the gross national product (GNP) deflator divided by the import price for agricultural goods, proved significant in explaining agricultural imports for a number of industrial countries.

Import volumes for the regions are determined by the purchasing power of export receipts. Export receipts in turn are determined by manufacturing activity in the 14 industrial countries.

The model is estimated principally by ordinary least squares. The most frequently used period of estimation is the first half of 1964 through the first half of 1976.

Additional Model Relationships

In addition to the equations briefly described above, a number of other relationships are needed if the model is to be used for certain simulation or forecasting applications. Price relationships have been developed for three sets of prices which were assumed earlier to be exogenously determined—spot commodity prices, prices of raw materials on the domestic market, and prices of manufactured goods that are domestically produced. 4 No supplementary relationships have been developed to allow labor costs to respond to exchange rate and domestic price changes, so that these costs continue to be exogenously determined. It was not thought desirable to attempt to endogenize domestic wage formation within a model focusing on countries’ trade flows and, in any case, highly dependent on a wide range of exogenously determined domestic variables.

In the set of relationships for manufactures, the key exogenous variable is the final domestic demand for manufactures. Developments in final domestic demand for manufactures are related to developments in certain GNP components. This linkage is provided to facilitate the analysis of the trade balance implications of policy options formulated in terms of their impact on GNP components.

Finally, the model is based on customs data disaggregated by commodity class rather than on aggregate balance of payments data. Therefore, a translation of customs data into balance of payments data is needed to obtain estimates of model simulations or forecasts in terms of their implications for a country’s trade balance.

Price relationships

For most purposes spot commodity prices are treated as exogenous. However, in simulating income effects it is necessary to capture insofar as possible the effects of fluctuations in activity in the industrial countries on these prices. To this end, highly simplified partial relationships are specified between changes (in logs) in the individual spot commodity prices, on the one hand, and, on the other, between changes (in logs) in a world export price index for manufactures representing the cost of producing these commodities and changes in the level of capacity utilization in the industrial countries. Dummy variables are also included to allow for the more rapid adjustment of spot commodity prices than for export unit values of manufactured goods to changes in fuel costs. While this specification neglects a large number of variables that are important in the determination of commodity prices, it is assumed that movements in these variables will not tend to be correlated with changes in demand pressure in the industrial countries, so that the coefficient estimates for the capacity utilization variable will be unbiased. Because of the simplistic equation specification, a correction for serial correlation is automatically included.

The explanatory power of these equations is low. The coefficients for the variable reflecting the activity level in the industrial countries are given in Table 1. They vary widely across commodities, and the pattern is sometimes unexpected. For example, it seems somewhat surprising that no activity level effect was found for the spot price of iron. Clearly, the large negative coefficients on activity variables for the three sugar prices are the result of spurious correlation between decreases in the level of economic activity in the industrial countries and widespread failures of sugar crops. If the activity coefficients for sugar are set to zero, while the activity coefficients for the other commodities are aggregated according to the importance of the commodity in world trade, average capacity utilization coefficients of 1.3 for foods and 2.0 for raw materials are obtained. These coefficients appear reasonable. The individual commodity series are not used by themselves; rather, they are used as inputs in the construction of aggregate commodity indices. It is assumed that, as in the calculations discussed above, the aggregation procedure will prevent any undue variability in the commodity indices for a particular country.

Table 1.

Effects of 1 Per Cent Change in Level of Activity in Fourteen Industrial Countries on Spot Commodity Prices Expressed in U. S. Dollars 1

(In per cent)

article image

The t-statistics are in parentheses.

The activity coefficients for sugar are assumed to be equal to zero. The average based on all food series except sugar is 1.6; when the estimated coefficients for sugar are included, it is 0.8.

Domestic prices of raw materials enter into the determination of export prices for manufactures. Thus, in simulating the effects of changes in activity levels on countries’ trade balances, it is also necessary to consider the impact that changes in world spot prices have on domestic raw material prices. Further, forecasting relationships for these prices are required.

To reflect the linkage with the spot price indices, weighted averages of the spot commodity price indices are constructed for the 14 industrial countries, with the weights reflecting the approximate importance of these commodities as intermediate inputs for the manufacturing sector. Changes in domestic raw material prices are then related to changes in domestic wages and to the current and lagged changes in these weighted indices of spot prices. Because the domestic price indices, other than the index for the United States, cover fuel prices as well as other raw material prices, variables representing the current and lagged values of changes in the price of oil are included in most of the equations. The final regression results are given in Table 2. Because the wage terms were significant with the correct sign for only 3 of the 14 countries, they were omitted from the equations. 5 The oil price coefficients for Denmark and Italy, when unconstrained, appeared to be too high at 0.4 and 0.5 and, therefore, were set at 0.25.

Table 2.

Determination of Domestic Prices of Raw Materials in Local Currency 1

article image

These relationships are expressed in change in log form and are based on semiannual data for the period 1960–76. The t-statistics are in parentheses.

The explanatory power of the equations for the domestic price of raw materials varies widely among countries, with strong performances for Belgium-Luxembourg, France, the Federal Republic of Germany, Japan, the Netherlands, and the United Kingdom and weak performances for Denmark, Italy, and Norway. 6 The coefficients for current or lagged spot price averages, or both, were significant for 11 of the 14 countries, and were close to significance for 2 others. Without information about the precise coverage and construction of the domestic raw material price indices, it is difficult to judge whether the estimated spot price coefficients are reasonable. It would appear that the sum of these coefficients for a particular country should fall short of unity because of the inclusion of oil in the index, because of the share of raw material imports purchased under long-term contract, and because of the domestic value added that is included in the domestic price.

To simulate income effects it is also necessary to capture the impact of world spot price changes on the price of domestically produced manufactures. To measure this impact, and also to obtain a relationship needed for forecasting purposes, an equation closely akin to that used for export prices of manufactures was specified. Changes in prices of domestically produced manufactures are explained by changes in domestic costs of production and changes in the price of competing goods. Domestic costs are represented by unit labor costs and the costs of raw materials. Prices of competing goods are represented by the import unit value for manufactures. The estimation results are presented in Table 3. 7 As in the export unit value equations for manufactures, the coefficients of the domestic price of raw materials tend to be highly volatile and frequently exceed values suggested by countries’ structures of production. Consequently, the constraints obtained from input-output tables and used in the estimation of the export unit value equations for manufactures are also imposed on the coefficients for the domestic price of raw materials in the estimation of the equations for the domestic price of manufactures. A further similarity with the export unit value equations is the importance of short-run fluctuations in the explanatory variables in the determination of the parameter estimates. Thus, the effects of movements in the prices of manufactured imports tend to dominate domestic price formation for manufactures, depending on the openness of the economy, although cost factors must clearly be the determining influence in the longer term.

Table 3.

Determination of Domestic Prices of Manufactures in Local Currency 1

article image

For all countries other than Austria, the sample period is first half 1964-first half 1976; for Austria it begins in 1967. All variables other than dummies are expressed in local currency in changes in logs. The t-statistics are in parentheses.

For Canada, the dummy takes the value 1 for first half 1974 and zero elsewhere, capturing the especially large impact of fuel price increases on domestic price formation.

For Denmark, the dummy takes the value 1 for first half 1973, reflecting a very large domestic wage settlement and some recovery in profit margins.

For the Federal Republic of Germany, the dummy takes the value 1 for first half 1968 and -1 for second half 1968, reflecting a temporary change in the response pattern of price changes, perhaps resulting from a change in the value-added tax.

For Italy, the dummy takes the value 1 for first half 1975. For the Netherlands, the dummy takes the value 1 for first half 1969. In both cases, these dummies eliminate extreme observations that would otherwise bias the estimated coefficients. For Italy, the dummy may reflect the effect of restrictive domestic price policies, while for the Netherlands it may reflect the effects of the introduction of a value-added tax on January 1 and certain other tax modifications.

The R¯2 statistic relates to the explanatory power of the nonconstrained variables. The explanatory power of the relationship including the effects of the constrained variables tends to be higher in most instances, and the correlation between the actual and predicted value using the full relationship is also shown.

A memorandum item in Table 3 indicates the percentage of final domestic demand for manufactures in 1970 accounted for by manufactured imports. These figures must be viewed as approximate, since they are based on a combination of customs data, national accounts data, and fixed-share coefficients obtained from input-output tables. When the import share in final domestic demand is large, one would expect the coefficient of the import unit value for manufactures also to be large and, over the short run, the coefficient of the unit labor cost variable to be less important. These import shares were used as rough guides in assessing the reasonableness of the estimated coefficients for the import price of manufactures. According to the relative magnitude of these shares, the unconstrained import price coefficients appeared too low for Austria, Denmark, the Netherlands, and Switzerland at 0.14, 0.29, 0.16, and 0, while those for Japan, Sweden, and the United States, at 0.17, 0.86, and 0.20, appeared too large. Constraints for these parameters were selected on the basis of import shares and sample information. 8

Domestic labor costs appear to play virtually no role in the relationships for Belgium-Luxembourg, Denmark, and Norway. According to the share calculations, these are among the more “open” manufacturing sectors. For Austria, Canada, Italy, the Netherlands, and Sweden, about one third of the changes in unit labor costs is passed through to domestic prices for manufactures; for France, the Federal Republic of Germany, Japan, and the United Kingdom the calculated pass-through ranges from 0.6 to 0.8. The coefficients for Switzerland and the United States sum to more than unity, which is somewhat surprising. However, the deviations from unity are not statistically significant.

Linkage between GNP components and final domestic demand for manufactures

Supplementary links to translate changes in GNP components into changes in final domestic demand for manufactures are needed both for simulation and to obtain forecasts for this variable. These relationships are given in Table 4.

Table 4.

Determination of Real Final Domestic Demand for Manufactures 1

article image

The variables are expressed in changes in billions of 1970 U. S. dollars. For Austria, Belgium-Luxembourg, Denmark, the Netherlands, Norway, Sweden, and Switzerland, the relationships are based on annual data; for the other seven countries, semiannual data from 1964 to first half 1976 were used. For six of the smaller countries, data for 1960–75 were used; for Belgium-Luxembourg, the estimation period started in 1962. The t-statistics are in parentheses.

For Italy, government expenditures and changes in stocks are included in the consumption variable.

Initially, ordinary least squares were used to estimate the relationship between changes in the level of real GNP components and changes in the level of real final domestic demand for manufactures. The inclusion of the changes in inventory variable in the relationship precluded a log-linear specification, since this variable can be positive or negative. The specification in terms of levels was desirable in any case, in that it facilitated the assessment of the reasonableness of the estimated coefficients and the incorporation of outside information on the magnitude of these coefficients. 9 Because of a high degree of collinearity among the explanatory variables, the estimated coefficients for a number of the variables were excessively large and positive, while those for others were negative. Thus, it was necessary to incorporate prior information obtained from input-output tables on the size of these coefficients in the regression equations. 10 The technique used to incorporate this prior information was F-class estimation, which requires point estimates for the individual coefficients as well as the variance-covariance matrix of these estimates. 11 The point estimates for private consumption and investment expenditure and for government expenditure were taken from national input-output tables when available; the Austrian input-output matrix was used as a proxy for the Swiss input-output matrix. In the estimation a heavy weight was given to the point estimates (expressed as proportions) by specifying the standard deviation about these estimates to be 0.03. The standard deviation for the point estimates for Switzerland was increased to 0.1 because of the proxy nature of the input-output coefficients used for Switzerland. In no instance could one reject conclusively the hypothesis that the prior information was compatible with the sample data, although in a number of instances the goodness-of-fit was reduced substantially by the introduction of prior information.

Linkage between balance of payments and customs data

A final set of relationships is needed to translate customs data into balance of payments data. Customs data may differ substantially from balance of payments data for reasons that vary from country to country. In general, customs data, in contrast to balance of payments data, do not include transactions in non- monetary gold or electrical current, but do include military goods shipments; a large number of additional adjustments are made to customs data to arrive at balance of payments data, and these are given in summary form by country in the Fund’s Balance of Payments Yearbook. The mean values of the ratio of customs to balance of payments flows indicating the net magnitude of these adjustments for exports and imports are given in Tables 5 and 6, respectively. 12

Table 5.

Relationships Between Balance of Payments Data and Customs Data: Exports 1

article image

The ratio of balance of payments exports to customs exports was regressed on a constant term, certain trend elements, and, in a number of instances, shift or other dummy variables. For all countries other than Canada and Switzerland, semiannual data for 1970-first half 1976 were used. Because of a discontinuity in the Canadian data, the sample was shortened to 1971-first half 1976. For Switzerland, it was necessary to use annual data for the period 1967–75. The t-statistics are in parentheses.

Dummies were included to eliminate extreme observations to prevent them from distorting the coefficient estimates, or to reflect shifts in recording practices. For Austria, dummy I = 1 starting in first half 1973. For Belgium-Luxembourg, dummy I = 1 in second half 1974 and zero elsewhere; dummy II = 1 in first half 1975. For Denmark, the dummies take the value 1 for second half 1973, first half 1974, and second half 1974 and zero elsewhere. For the Federal Republic of Germany and Switzerland, dummy I = 1 starting in first half 1974. For the United Kingdom, dummy I takes the value 1 starting in second half 1974. For the United States, dummy I takes the value 1 starting in second half 1975.

The R¯2 coefficient reflects the power of the trend variables and dummies in explaining movements in the ratio of balance of payments to customs data. The correlation coefficient reflects the extent to which the movement in the balance of payments data is explained by the estimated relationships used in conjunction with customs data. When the adjusted R¯2 was negative it was not reported.

Table 6.

Relationships Between Balance of Payments Data and Customs Data: Imports 1

article image

The ratio of balance of payments imports to customs imports was regressed on a constant term, certain trend elements, and, in a number of instances, shift or other dummy variables. For all countries other than Canada and Switzerland, semiannual data for 1970-first half 1976 were used. Because of a discontinuity in the Canadian data, the sample was shortened to 1971-first half 1976. For Switzerland, it was necessary to use annual data for the period 1967–75. The t-statistics are in parentheses.

Dummies were included to eliminate extreme observations to prevent them from distorting the coefficient estimate, or to reflect shifts in recording practices. For Austria, dummy I = 1 in second half 1973 and zero elsewhere; dummy II = 1 in first half 1974. For Denmark, the dummies took the value 1 for second half 1973, first half 1974, and second half 1974 and zero elsewhere. For France, dummy I = 1 starting in first half 1975. For Italy, dummy I = 1 starting in second half 1972. For Japan, dummy I = 1 starting in first half 1973. For Norway, dummy I = 1 starting in second half 1973.

The R¯2 coefficient reflects the power of the trend variables and dummies in explaining movements in the ratio of balance of payments to customs data. The correlation coefficient reflects the extent to which the movement in the balance of payments data is explained by the estimated relationships used in conjunction with customs data. When the adjusted R¯2 was negative it was not reported.

To provide a simple technique for moving from data on a customs basis to data on a balance of payments basis, the ratio of exports (imports) on a balance of payments basis to exports (imports) on a customs basis was regressed on a constant term and certain trend elements; the residuals were inspected for discontinuities reflecting changes in recording procedures or shifts in adjustment factors and for extreme observations that might otherwise bias the coefficient estimates. In a number of instances, shift dummies for particular observations were included in the estimated equations. In general, the time trends did little to explain the movements in these ratios. Shift dummies were important in explaining the movement of the export ratios for Austria, the Federal Republic of Germany, Switzerland, and the United States and the movement of import ratios for France, Italy, Japan, and Norway. The R¯2 statistic relates only to the ability of trend and dummy variables to explain the movement in the ratios of data on a balance of payments basis to data on a customs basis; the correlation coefficients calculated between actual balance of payments data and estimated data obtained by using the predicted ratio and actual customs data are given in Tables 5 and 6. These coefficients are very close to one.

II. Simulations of Income Effects

The measurement of the impact of changes in activity or income on countries’ trade balances is important, since in the short to medium term income effects have a large impact on countries’ trade balances. Simulations of trade balance effects of a 1 per cent increase in aggregate domestic demand for a given industrial country, or a 1 per cent reduction in its output gap, provide useful information for policy formation on the magnitude of income effects. To the extent that simulations are based on the short-term to medium-term growth targets of authorities in the different countries, the information they provide about trade balance effects may suggest whether these targets are consistent with satisfactory external positions. Finally, it may be possible to derive sets of growth targets that satisfy certain external criteria while contributing to a greater absorption of unused resources.

In this section two simple multiplier-type calculations for each of the 14 industrial countries taken individually are presented: (1) the trade balance effects at the level of 1977 trade flows of a domestic demand expansion that would lead to a 1 per cent increase in manufacturing output in a particular country, with the activity levels in all other countries held constant; and (2) the trade balance effects at the level of 1977 trade flows of a 1 per cent increase in the level of domestic demand for a given country, with the level of domestic demand in all other countries held constant. These multiplier tables are useful in obtaining rough estimates of the effects of different growth paths on countries’ external positions, given certain restrictive assumptions that will be explored below.

A number of more complicated calculations concerning income effects are also of interest. Three sets of calculations are presented. The first set shows the extent to which additional unused resources could have been absorbed in certain industrial countries without causing a deterioration in their trade balances if the capacity utilization rates in manufacturing (ratios of actual to potential output) in the Federal Republic of Germany, Japan, and the United States had been about 5 per cent higher. Such calculations can be useful in assessing the dependence of economic growth in countries with balance of payments constraints that are not responsive to exchange rate movements, at least in the short run, on activity levels in partner countries. A second set of calculations shows the pattern of trade balances in industrial countries that would have prevailed if all industrial countries had been at “full-employment.” 13 (In these calculations, production capacities, money wage rates, and exchange rates are assumed to be exogenous.) Finally, while full-employment calculations are of interest, they may not provide a meaningful picture of the “underlying” external balance of payments positions of the various countries to the extent that full-employment activity levels are not achieved over the medium term. In such instances it may be more useful to consider what the external position of a country would be if the country and its trading partners were experiencing activity levels (in terms of utilization rates) expected to prevail two or three years hence. Various growth scenarios are, of course, possible, and each of them has different balance of payments implications. The “yield” of such an analysis is, therefore, not a forecast of what is expected to happen but an assessment of what could happen as a result of various growth patterns. In the third set of calculations presented here, two separate growth scenarios are introduced, and the corresponding trade balance implications for the industrial countries are derived.

Trade Balance Effects of Changes in Level of Manufacturing Output or Final Domestic Demand for Manufactures

It is important to note what assumptions underlie the calculations in Tables 7 and 8 and how they should be interpreted. By definition, the change in the activity level has no effect on the model’s exogenous variables, which include potential output, the exchange rate, and nominal domestic wages. The association of higher activity levels with higher levels of net capital formation and higher levels of potential output may be of some importance in projecting the strength of countries’ export performances over the medium term, but it is of little significance over a time span of one to two years. To the extent that countries continue to have high levels of unemployment, it may also not be unreasonable to assume that wages are unaffected. Finally, one of the purposes of the model is to look at countries’ trade balances, assuming no exchange rate change, to see whether such an assumption appears sustainable.

Table 7.

Effects of 1 Per Cent Increase in Level of Manufacturing Output on Countries’ Trade Balances 1

(In billions of U.S. dollars)

article image

At the level of 1977 trade flows.

Table 8.

Effects of 1 Per Cent Increase in Level of Final Domestic Demand for Manufactures on Countries’ Trade Balances1

(In billions of U.S. dollars)

article image

At the level of 1977 trade flows.

In Table 7, the actual level of manufacturing output in 1977 for a single country, say, France, is increased by 1 per cent, with the result that the output gap in manufacturing is reduced by approximately 1 per cent, while the levels of output in all other industrial countries remain unchanged. The volumes block of the model has essentially three sets of relationships—export volume equations, import volume equations, and an identity—relating imports, exports, final demand, and manufacturing output. Thus, when the levels of manufacturing output are exogenously determined, the model then solves for the levels of real exports and imports of manufactures and final domestic demand for manufactures that are consistent with the exogenously determined levels of manufacturing output. When French manufacturing output is increased, then final domestic demand in France must also increase. Because domestic demand increases, French manufactured imports will also increase. This may translate into an increase in German manufactured exports, but, since German output is constant by assumption, these additional resources must be made available for export by a reduction in German domestic demand. The calculations in Table 7 assume that appropriate policies to reduce domestic demand in partner countries, as described above, would have been implemented. In Table 8, the actual level of final domestic demand for manufactures in 1977 for a single country is increased by 1 per cent, while the levels of final domestic demand for manufactures in the other countries remain unchanged. This increase in demand increases the country’s imports of manufactures and encourages additional production at home and abroad.

It is somewhat difficult to use Table 8 for analyzing the trade balance impact of certain policy options because the options tend to be formulated in terms of their effects on more aggregative demand variables, such as GNP components, rather than in terms of the domestic demand for manufactures. To facilitate its use, the coefficients reported in Table 4 were combined with the levels data in real terms for 1977 for the various GNP components and the demand for manufactures to obtain “elasticities” for translating percentage changes in these components into percentage changes in final domestic demand for manufactures and finally into trade balance effects. The trade balance effects of a 1 per cent change in private consumption, private investment, or government expenditure on a country’s trade balance at the level of 1977 trade flows are given in Table 9. They should be used with caution since they depend heavily on the national accounts and trade balances for 1977.

Table 9.

Effects of 1 Per Cent Increase in Private Consumption, Private Investment, and Government Expenditure on Country’s Trade Balance 1

(In billions of U.S. dollars)

article image

At the level of 1977 trade flows.

Government and change in stocks are included in the consumption variable.

The calculations for Tables 7, 8, and 9 assume that changes in the level of production or demand would have no impact on domestic wages, exchange rates, or productive capacity, although provision is made for some impact on prices, particularly commodity prices, and on regional trade flows through additional export earnings. The figures are based on the scale of trade flows in 1977 and the size of the manufacturing sector in 1977. Thus, they cannot be used directly to forecast countries’ trade balances. Such forecasts would need to take into account the growth of actual output and the productive capacity of the economy, as well as the effects of changes in relative prices and factors reflected in trend terms, whereas these factors are held constant in the calculations presented here.

Table 7 shows that the “own” trade balance deterioration in U.S. dollars caused by a 1 per cent increase in manufacturing output in the Federal Republic of Germany would be approximately equivalent to the own trade balance deterioration caused by a 1 per cent increase in manufacturing output in the United States. The own trade balance deterioration caused by an expansion in Japanese manufacturing output is equal to the deterioration that a similar expansion in France would entail and less than half of the deterioration that the United States would experience. However, the trade balance effects of expansion in all industrial countries, as represented by the row sums, suggest a relatively strong deterioration in the U.K. and U.S. trade balances. 14

The trade balance effects of a 1 per cent increase in final domestic demand for manufactures, as shown in Table 8, are lower in all instances, in some cases by more than one half, than those relating to a 1 per cent increase in manufacturing output. A given change in final demand in a particular country translates into a smaller change in domestic manufacturing output: some portion of the change in final demand is always satisfied by imports, so that domestic manufacturing output has to increase by a lesser amount. Conversely, a given change in domestic output, with output in all other countries held constant, will be absorbed by an increase in domestic demand, but, since an increase in domestic demand also entails an increase in imports, this increase will be larger than the increase in domestic output. In the absence of an increase in output abroad, the increased demand for imports must involve a decrease in foreign demand, and thus of the exports of the particular country under consideration. The extent to which a change in final demand will be translated into a change in manufacturing output will vary from country to country depending on the propensities to import. More importantly, however, final domestic demand for manufactures may be substantially smaller than final output of the manufacturing sector if a large share of the manufactured products tends to be exported. The ratios of own trade balance effects of changes in aggregate demand to those of changes in manufacturing output range from a high of 0.8 for the United States to 0.3 for Belgium-Luxembourg, Denmark, and the Netherlands, with most of the remaining countries clustering around 0.5.

Table 10 presents estimates of how much more the output gap in manufacturing in certain industrial countries could have been reduced in 1977 (or by how much more their output could have been increased) without causing a deterioration in their trade balances if the output level in the Federal Republic of Germany, Japan, and the United States had been 5 per cent higher. As in Table 7, the calculations in Table 10 assume that the requisite demand management policies were followed in the industrial countries to assure the appropriate demand. The calculations ignore any feedback on domestic wages or exchange rates that higher levels of manufacturing output (or lower output gaps) might have.

Table 10.

Increased Output in Manufacturing in Selected Industrial Countries 1

(In per cent)

article image

Consistent with a 5 per cent increase in manufacturing output in the Federal Republic of Germany, Japan, and the United States and no deterioration in the trade balances of the industrial countries considered; at the level of 1977 trade flows.

First, the permissible increases in activity levels in all other industrial countries taken together that are consistent with no change in their external balances are considered in column 1. Those that would benefit most from such a scenario are Belgium- Luxembourg and France. Their advantage is large because of their pattern of trade, and because their own trade balance effects of domestic expansion tend to be small relative to the beneficial effects to be derived from the projected expansion abroad. In the second column it is assumed that the non-EEC countries do not respond to the external stimulus by absorbing unused capacity, with the result that they sustain trade balance improvements. The trade balance deterioration for the Federal Republic of Germany that would result from expansionary measures taken in this instance would be moderated significantly by the expansionary measures taken in its EEC partner countries to assure that their trade balances remained unchanged. In the third column it is assumed that expansionary policies are taken only in countries with high levels of unutilized capacity and with external positions that do not constitute significant policy constraints, while the trade balances of the other industrial countries strengthen. Thus, on the basis of the situation in 1977, the effects of higher activity levels in Canada, France, and Italy are explored. These estimates, suggesting the impact of higher activity levels on countries’ trade balances, clearly demonstrate that the more widespread the pursuit of expansionary policies the more countries are able to absorb unused resources without incurring a deterioration in their external accounts.

Trade Balance Effects of Changes in Activity Levels in Industrial Countries

As explained above, two sets of income simulations are presented here. The first set, in columns 1 and 4 of Table 11, explores the implications for the trade balances of the industrial countries of a simultaneous elimination of their output gaps. The estimated trade balance implications of this gap closing appear large at about $6 billion for the Federal Republic of Germany, $17 billion for the United States, and minus $12 billion for Japan, but are subject to very large margins of errors.

Table 11.

Trade Balance Effects of Changes in Activity Levels 1

article image

At the level of 1977 trade flows.

This figure reflects the amount by which the level of capacity utilization, as represented by the ratio of actual to potential output in manufacturing, is increased over its 1977 level to obtain the estimates presented here.

Calculations based on more limited absorption of unused capacity are presented in columns 2 and 3 and 5 and 6. Alternative A, reflected in columns 2 and 5, assumes substantial growth in the Federal Republic of Germany and, especially, Japan over the medium term and higher levels of unused capacity in France, Italy, and Sweden. Alternative B, in columns 3 and 6, assumes higher projected growth rates for Canada and the United States than under Alternative A, and lower growth rates for the other countries, especially the Federal Republic of Germany and Japan. Under Alternative A, expansion foreseen for the medium term would contribute to reductions in the German and Japanese balances of $3–5 billion, while the U.S. trade deficit would also increase; the external positions of France and Italy would strengthen by about $1.5 billion. Under Alternative B, the projected changes in utilization rates could contribute to a deterioration of $6 billion in the U.S. trade balance, with no change in the Japanese balance.

III. Simulation of Terms of Trade Effects

Although income variables in the industrial countries play a dominant role in the model, prices are also important in explaining developments in countries’ trade balances. Because of the rather short-run and partial nature of the model, it would not be realistic to use the model price elasticities to simulate volume effects of changes in relative prices two to three years in the future. However, equations explaining the movement of export and import unit values may be used to derive the time path of terms of trade effects over a period of 6 to 12 months.

The price equations used here are identical to those presented in the first paper on the world trade model, with the exceptions noted below. For three countries, the export unit value equations reported in the original paper did not appear satisfactory in that the coefficient on the competitor price variable was unreasonably low. 15 These relationships were re-estimated on the basis of revised data through the second half of 1977. The coefficients used in the simulations for Belgium-Luxembourg, Canada, and the Netherlands for the export unit value relationships for manufactures are given in Table 12.

Table 12.

Revised Export Unit Value Relationships for Manufactures 1

article image

The t-statistics are in parentheses. The period of estimation is second half 1962-second half 1977.

Constrained coefficients.

For Belgium-Luxembourg, first half 1965 = -1, second half 1965 = 0.5, first half 1970 = -1, and second half 1970 = 1. For the Netherlands, first half 1974 = 1.

In a few instances the labor cost coefficients and competitor price coefficients were scaled down proportionally so that the sum of these coefficients and the domestic price of raw materials coefficient would not exceed unity. These modifications tended to be minor for most countries.

Manufactured import unit values, by definition, are a simple transformation of countries’ export unit values for manufactures, although measurement errors can affect the estimated coefficients relating these two magnitudes. For most countries the estimated coefficient for the import unit value relationship for manufactures was close to one and was used in the simulations. For the Federal Republic of Germany, it was unreasonably low and was constrained to one. For a few other countries it was reduced marginally so as not to exceed one.

Simulations of partial trade balance effects based on the coefficients described above and relating to two sources of price change are presented here. These sources are changes in the nominal exchange rate of a particular country and changes in the domestic costs of production resulting from an exogenously given change in labor costs. Calculations of the static terms of trade effects of a nominal 10 per cent exchange rate devaluation at the scale of 1977 flows are given in Table 13. Simulations of the terms of trade implications of higher labor costs (by 10 per cent) in the industrial countries taken one at a time are presented in Table 14. The terms of trade effects are reported for the period in which the change occurs, and after a lapse of 12 months.

Table 13.

Terms of Trade Effects in Domestic Currency of Nominal 10 Per Cent Devaluation of Exchange Rate 1

(In per cent)

article image

The figures reflect the extent to which the per cent change in export prices in domestic currency exceeded the per cent change in import prices in the domestic currency.

Table 14.

Terms of Trade Effects in Domestic Currency of 10 Per Cent Increase in Nominal Wages 1

(In per cent)

article image

The figures reflect the extent to which the per cent change in export prices in domestic currency exceeded the per cent change in import prices in the domestic currency.

Austria and Sweden are not included in this table because labor cost variables were not included in the export unit value equations.

These estimates are presented to facilitate the short-run analysis of price effects but should be used with caution. They reflect terms of trade effects but not volume effects. Further, they reflect the working out over the short term of a once-and-for-all change in a country’s nominal wage or exchange rate. To the extent that the impact of exchange rate changes on domestic wages is significant within a short period of time, as it may well be especially in the smaller more open economies, the simulation results for these countries may not be useful.

Clearly, there are strong similarities, especially for the simulations based on wage changes, between the terms of trade effects for manufactures and the terms of trade effects for the aggregate trade balance in that the flow of manufactured goods tends to dominate exports and, to a somewhat lesser degree, imports of the industrial countries. For simulations based on exchange rate changes, the similarities between these two types of terms of trade effects are reduced somewhat because provision is made in these calculations for certain exchange rate changes also to affect world spot commodity prices. 16 These prices play a larger role in the determination of export and import unit values for raw materials and agricultural goods than in the determination of prices of manufactures. In addition, wages tend to be more important in the determination of the prices for manufactured goods. In the simulations presented here it is assumed that the price of oil in U. S. dollars remains constant.

The countries showing the smallest terms of trade deterioration from an exchange rate change (less than 1 per cent after one year), suggesting that at least for manufactures they are price takers, are Austria, Norway, and Sweden. The ones with the largest deteriorations by far are Switzerland and the United States—for Switzerland, because of the small size of the coefficient of competitor prices and because of the small weight for Switzerland in the calculation of competitiveness indices for partner countries and, for the United States, because of the omission of the competitor price variable from the price equation.

For the United States, the aggregate terms of trade effect differs substantially from the terms of trade effect for manufactures because of the postulated relationship between exchange rates and spot prices, the composition of U. S. trade flows, and, especially, the assumed stability of the price of oil in U. S. dollars. The Federal Republic of Germany, France, and Switzerland have the largest deterioration in the aggregate terms of trade.

For all countries other than the United States, the immediate terms of trade deterioration for manufactures alone of a 10 per cent devaluation exceeds the effect after one year because of the lagged adjustment of competitor prices to the export price of the devaluing country, and the further adjustment of the exporting country’s prices to those of its competitors. For the United States, the lagged adjustment of competitor prices is more than offset by the impact of the U. S. dollar devaluation on the spot prices of commodities, resulting in a further deterioration in the terms of trade for manufactures.

The aggregate terms of trade effect for 11 of the 14 countries moderated or remained virtually unchanged over the 12-month period considered. Because of the assumed effect of Japanese yen and U. S. dollar devaluations on spot commodity prices, and the relatively small lagged effects of competitor prices for Japan, the further deterioration in the aggregate terms of trade for these countries is not surprising.

The effect of a 10 per cent increase in nominal wages on the terms of trade for manufactures tends to be very similar to its terms of trade effect for aggregate trade flows. For most countries, the aggregate terms of trade effect after one year is smaller than the effect for manufacturing. For Denmark, the aggregate terms of trade effect is greater because of the relative importance of wages in the determination of export unit values for agricultural goods.

The absolute size of the terms of trade effect for the United States is almost twice as large for a 10 per cent change in wages as it is for a 10 per cent change in the exchange rate, and at least twice as large as the effect of a 10 per cent change in wages in other industrial countries. Again, this extreme result for the United States reflects the absence of a competitor price variable in the export unit value equation for manufactures and the relative importance of labor costs. The aggregate terms of trade effect is only slightly smaller than the effect for manufactures alone, in contrast to the pattern of effects shown in Table 13. This is because wage movements, unlike exchange rate movements, are not projected to have any impact on world spot prices.

IV. Conclusion

In the preceding sections, the basic structure of the world model of merchandise trade, as well as minimum extensions to the structure needed for simulations, has been described. Calculations showing trade balance effects of alternative activity and demand situations have been presented in some detail, since the model focuses particularly on the effects of economic activity on trade flows. Estimates were derived by changing the level of manufacturing output, or, alternatively, final domestic demand for manufactures, in a single country while holding constant the levels of manufacturing output, or final domestic demand, in partner countries. These estimates were presented in the form of multiplier tables. Alternatively, scenarios were constructed according to which activity levels in a number of countries, or even in all industrial countries simultaneously, were adjusted, and their trade balance implications were analyzed. The simulations based on variations in activity levels show that, while trade balances in particular countries may respond especially strongly to changes in activity levels, nonetheless cyclical factors tend to be extremely important in determining the evolution of trade balances in all of the industrial countries.

The price simulations presented in this paper abstract from the volume effects of price changes, which are realized with a substantial lag, and focus only on short-run terms of trade effects. This is because wages are assumed to be unaffected by changes in activity levels or exchange rates in the model, an assumption that can be realistic only for a short period of time. Further, cost variables play too small a role in the price determination process for realistic medium-term simulation results to be obtained.

The terms of trade simulations indicate that the effects of a 10 per cent change in the exchange rate, or a 10 per cent change in labor costs, may differ sharply between countries. These differences reflect, in large part, the different roles played by competitor prices in the formation of export prices of manufactures. Differences in terms of trade effects also stem from variations in the relative importance of commodity trade for the industrial countries.

In terms of improving the specification of the model, and especially of enhancing the usefulness of the price block for other than very short-run calculations, further attention needs to be given to supply constraints and the role of domestic production costs in the formulation of prices. Realism increasingly demands the inclusion of a mechanism for transmitting the impact of changes in import prices on prices of domestic factors of production.

*

Ms. Ripley, Senior Economist in the Central European Division of the European Department, is a graduate of Wellesley College and Harvard University.

1

See Michael C. Deppler and Duncan M. Ripley, “The World Trade Model: Merchandise Trade,” International Monetary Fund, Staff Papers, Vol. 25 (March 1978), pp. 147–206.

2

The first, third, and fourth groupings correspond to those now used in the International Monetary Fund’s International Financial Statistics.

3

The noncyclical element of unit labor costs is represented by normal unit labor costs, calculated as the ratio of labor costs to normal output per man hour; the cyclical demand is represented by the ratio of normal to actual unit labor costs.

4

Because of its small role in the model in explaining the demand for agricultural imports, no attempt has been made to endogenize the GNP deflator.

5

One reason for this general lack of significance may be the importance of imports of raw materials and fuels in the domestic market for raw materials.

6

The goodness-of-fit shown in Table 2 relates to the ability of the variables in the equations, whose coefficients are not constrained, to explain the movement of the dependent variable adjusted for the movement in the constrained variables. If the explanatory power of all independent variables with respect to the unadjusted dependent variable is taken into account, R2s for Denmark and Italy would increase to 0.5 and 0.4. When no constraints were imposed, the R2s for these two countries were 0.8 and 0.9.

7

As in the export unit value equations, labor cost was decomposed into its “normal” and “cyclical” components. The parameter estimates for the cyclical component were generally unsatisfactory, and, therefore, it was omitted from the final specification.

8

Constraints of 0.3 to 0.4 were introduced for Austria, but the results were highly unsatisfactory in that the standard error became quite large and the coefficients of labor costs became unrealistic; therefore, the original coefficients were retained. The difficulty here may be in large measurement errors in the import price variable for Austria.

9

The selection of the levels rather than logs specification made it necessary to standardize the variables in these relationships as regards magnitude and base period. Since the demand for manufactures series were constructed using 1970 as the base period for both exchange rates and prices, and since the base period for real GNP components varied widely across countries, these components were converted to index form and then multiplied by their 1970 value, in U.S. dollars, to achieve the needed reference period consistency.

10

Input-output coefficients reflect the proportion of the final expenditure category allocated directly to manufactured goods in the base period but necessarily ignore the indirect effects certain types of expenditures may have on the demand for manufactures; for example, government expenditure on services may boost private purchases of manufactures. Thus, it is particularly important that the GNP components be consistent, so that both direct and indirect effects are properly represented.

11

For a description of the technique, see Henri Theil, “On the Use of Incomplete Prior Information in Regression Analysis,” Journal of the American Statistical Association, Vol. 58 (June 1963), pp. 401–14.

12

The discrepancy between customs and balance of payments data for Belgium is particularly large. This is because Belgian balance of payments data, in contrast to balance of payments data for most other countries, are based on the exchange record rather than on trade returns, and the valuation and timing of trade flows differs between these two sources of trade information.

13

Full employment is defined here as actual manufacturing output being equal to potential manufacturing output. The approach used to estimate the potential output series is described in Jacques R. Artus, “Measures of Potential Output in Manufacturing for Eight Industrial Countries, 1955–78,” International Monetary Fund, Staff Papers, Vol. 24 (March 1977), pp. 1–35. The series coverage has since been extended to ten countries, and the series have been updated.

14

The effects of simultaneous expansion on countries’ trade balances can only be approximated by the application of multipliers presented here because of certain consistency adjustments in the model and because of nonlinearities.

15

The equation for Switzerland was re-estimated because of the unexpectedly small role played by competitor prices in this equation. The results based on revised data showed little change.

16

According to the relationships used in the simulations, a 10 per cent devaluation of the U.S. dollar results in a 3 per cent increase in all spot commodity price indices expressed in U.S. dollars; a 10 per cent devaluation of the Japanese yen and the deutsche mark both result in a 2 per cent fall in the U.S. dollar spot price of commodities. Devaluations of 10 per cent for the French franc, Italian lire, and pound sterling translate into decreases in U.S. dollar spot commodity prices of 1 per cent.

Same Series

Other IMF Content

Other Publishers

  • Collapse
  • Expand
Cover IMF Staff Papers
IMF Staff papers: Volume 27 No. 2
Author:
International Monetary Fund. Research Dept.
Volume/Issue:
Volume 1980: Issue 002
Publisher:
International Monetary Fund
ISBN:
9781451946864
ISSN:
1020-7635
Pages:
250
DOI:
https://doi.org/10.5089/9781451946864.024