Information about Western Hemisphere Hemisferio Occidental

IV Global Effects of Changes in Imports in Program Countries

Morris Goldstein
Published Date:
March 1986
  • ShareShare
Information about Western Hemisphere Hemisferio Occidental
Show Summary Details

The broad characteristics of program countries that were reviewed in Section III are helpful as a indication of their potential to influence macroeconomic developments in the rest of the world. In order, however, to gain a more focused view of the global effects of Fund programs, it is useful to study such effects within the more formal framework of an econometric global trade model.

The more formal approach is enlightening for at least three reasons. First, while crude share calculations can provide estimates of the “impact” or “first-round” effects of changes in program countries on other country groups, they typically cannot provide estimates of the induced “later-round” effects; yet these later-round, more general equilibrium effects could be quantitatively quite significant. To take a specific example, suppose that program countries experience, on average, a 5 percent fall in the volume of their imports during the program year and that: (1) industrial countries, on average, send 5 percent of their total exports to program countries; and (2) exports average 15 percent of GNP for industrial countries.37 In this case, the impact effect of the lower imports by program countries on GNP in industrial countries would be to reduce the latter by 0.0375 percent (since 0.0375 = 5.0 – 0.05 – 0.15). But what happens next?

To estimate the full effect of the same fall in imports, one would want to know how the induced change in the GNP in industrial countries affects: (1) domestic consumption, investment, the demand for money, interest rates, and so on; (2) trade flows and ultimately real income again within the industrial countries as a group, as lower income induces lower imports from other industrial countries, and another round of exports, real income, and imports; and (3) subsequent trade flows between industrial countries and developing countries (including program countries), as lower income in industrial countries also reduces the volume and prices of exports from developing countries to that region, and in turn, the foreign exchange receipts, import volumes, and real incomes in developing countries.

Although it is difficult to generalize across global trade models, recent research seems to suggest that such later-round, linked effects can multiply the impact effect of the disturbance by two to three times.38 Returning to our example, this would mean that the full effect of the 5 percent fall in the imports by program countries on real GNP in industrial countries could be to reduce it by 0.075 to 0.1125 percent (compared with the impact effect of a decline of 0.0375 percent).

The second area where econometric models have a comparative advantage over rough share calculations is in estimating the timing of the effects of programs. It is one thing to assert that, say, a fall in the flow of financing to program countries will eventually affect their imports. It is quite another to identify the shortrun (one-year) and long-run (three-year) elasticities of import volume with respect to foreign exchange receipts, to say nothing of how these elasticities may differ across groups of developing countries (across low-absorbing oil-exporting developing countries, for instance, compared with low-income non-oil developing countries).39 Since most econometric global trade models explicitly allow for lagged effects in the determination of trade volumes and prices, they can tell us something about the speed with which disturbances in program countries might be transmitted to non-program countries.

Yet a third reason for turning to trade models is that marginal trade propensities may differ significantly from the average propensities captured in the share estimates, and it is changes at the margin that are most relevant for assessing the effects of Fund-supported programs. The share of imports and exports in the GNP of (industrial countries) has been on a clearly rising trend over the last thirty years, and marginal exceed average propensities when the latter are rising.40 Econometric trade models give estimates of the relevant marginal propensities (or elasticities) directly from the estimated coefficients.

In this section, three world trade models are used to simulate the effects on the trade and output of industrial countries of a hypothetical reduction of $20 billion in the value of imports of developing countries. Each model has strengths that at least partially compensate for the weaknesses of the others.

The OECD Interlink model contains a fully articulated income-expenditure model for each of 23 OECD countries. In these individual country models, blocks of equations determine the main components of demand, wages and prices, foreign trade price and volumes, the distribution of income, output and employment, and financial variables. The non-OECD regional models are much simpler, containing only reduced-form equations for import volume and export-pricing behavior. Because output in the OECD model is endogenously determined for the industrial countries, it is possible to calculate traditional foreign-trade-multiplier effects on real GDP in the industrial countries in response to exogenous foreign trade disturbances. The second attraction of the Interlink model is that it contains an up-to-date foreign trade matrix (for 1982) so that the direction of trade between industrial and developing countries, both in the aggregate and on an individual-country basis, is accurately reflected. However, the simulation properties of the Interlink model have to be gleaned from published results and these only consider the non-OECD area as a group.41 This means, for example, that program countries (the great majority of which are non-oil developing countries), are lumped together with, say, major oil exporting developing countries, as well as with non-member planned economies; to the extent that the trade behavior of these non-program countries differs from that of program countries, such aggregation into one “non-OECD region” could distort the results.42

The second model used in the simulations, the Fund’s World Trade Model, has the advantage that non-oil developing countries constitute a separate group in the model and its published results. Since the commodity composition and direction of trade of the (1983) program countries is apparently quite similar to those of all non-oil developing countries, aggregation problems are reduced.43 This model has two disadvantages for our purposes: industrial-country income is endogenously determined, so that only the effects on trade of a change in imports by developing countries can be studied; and the model’s foreign trade matrix relies on 1970 data for the direction of trade between industrial and developing countries.

The third model used is the LINK model of Project LINK. It has two clear advantages for our purposes: (1) real output is endogenous in both industrial and non-oil developing countries, with the latter depending not only on capital stock but also on non-fuel imports in non-oil developing countries; and (2) the published simulation results confine the import shock to non-oil developing countries rather than to all developing or non-OECD countries. The disadvantages of this model are that the published results consider the effects of a $20 billion change in transfers to non-oil developing countries rather than in their imports, and that both the underlying estimation results and the base-line scenario for computing impact and dynamic multipliers are based on pre-1977 data and forecasts. Of these deficiencies, probably only the outdated estimation period is potentially serious, for two reasons. The first is that the estimated import volume equations in the LINK model itself suggest that imports by non-oil developing countries respond to foreign exchange receipts with a one-year elasticity that approaches unity (so $20 billion less in transfers to developing countries results, within a year, in a $16.4 billion reduction in imports by developing countries). The second reason is that the multipliers in the model are apparently not very sensitive to the characteristics of the base-line or control solution.

Before turning to the simulation results, three caveats are appropriate about the inferences that can legitimately be drawn from them. To begin with, so long as import decisions in program countries are affected by foreign exchange receipts, changes in imports can reflect a wide variety of program (and non-program) influences—such as the effects of Fund programs on the exports and net capital inflows of program countries, as well as on overall public and private spending and investment. For example, the move from an overvalued to a more realistic exchange rate could simultaneously increase the production of exportables and reduce capital flight in a program country; on both counts, foreign exchange receipts would be altered and imports would change, and this quite apart from any changes induced by the program in, say, the government’s fiscal position. For this reason, the results of the import simulations should not be interpreted as implying that Fund-supported programs only work, and have international effects, by directly affecting program countries’ ability to import.

The second caveat arises from the earlier discussion of program effects in Section II; historical or observed changes in the imports of program countries should not be equated with the effects of Fund programs. It is the size of the cross-country multiplier that is of interest in the simulation exercise, not the sign of the initiating disturbance. The $20 billion fall in the imports of program countries used in the simulation exercises should be viewed without prejudice as to the sign of any transmission effects of programs on imports. If, for example, one assumed that without Fund programs the decline in program country imports would have been much larger in say, 1982–83, then the simulations could just as well be run using an increase in imports. In other words, what is of interest in the simulations is the size of the cross-country multipliers, not the sign of the initiating import disturbance.

The third caveat deals with the precision of the simulation results. Such exercises are apt to be subject to fairly wide margins of error because: (1) trade, and especially income determination in developing countries, remains the most primitive part of most global trade models (as an example, exports of developing countries in these models are independent of imported inputs); (2) such trade models are designed for, and best suited to, analyzing transmission from North to South rather than the other way around; and (3) the financial link among developed and developing countries (some of which may respond to the same program measures as imports) is not well developed in these models. The simulation results should, therefore, be viewed as indicative rather than conclusive.

Simulation Results

The effects on real output and trade of a hypothetical $20 billion exogenous reduction in the imports of the developing countries are given in Tables 17 and 18, which use the OECD Interlink Model and the Fund’s World Trade Model, respectively. In the former model, the import reduction applies to the whole non-OECD group, while in the latter, it occurs only for non-oil developing countries. Table 19 shows the results of a similar exercise, namely a $20 billion exogenous reduction in financial transfers to non-oil developing countries (implying a $16 billion fall in developing countries’ imports), using the Project LINK model.

Table 17.Effects on Real Output and the Trade Balance of a $20 Billion Decline in Imports by Non-OECD Regions: Interlink Model(In percentage change from baseline solution)
All OECD countriesFirst YearSecond YearThird Year
Real GDP−0.3−0.3−0.3
Exports of goods and services (in volume)−1.1−1.1−1.1
Imports of goods and services (in volume)−0.5−0.5−0.5
Total domestic demand deflator−0.1−0.2
Current balance (in billions of U.S. dollars)−9.0−8.6−9.1
Non-OECD region
Export volume−0.9−0.9−0.9
Import volume−1.8−1.8−1.8
Current balance (in billions of U.S. dollars)
Effects on selected countries (third year only)Real GDPVolume of Exports
United States−0.1−1.0
United Kingdom−0.3−1.2
OECD (Europe)−0.3−1.2
Source: Larsen et. al. (1983), Table A10.
Source: Larsen et. al. (1983), Table A10.
Table 18.Effects on Merchandise Trade Balances in Industrial Countries of a $20 Billion Decline in Imports by Non-Oil Developing Countries: IMF World Trade Model(In billions of U.S. dollars)
First YearSecond YearThird Year
Merchandise trade balance−10.1−11.4−11.2
Source: Fund staff estimates.
Source: Fund staff estimates.
Table 19.Effects on Real Output and Trade Balance of a $20 Billion Decline in Financial Transfers to Non-Oil Developing Countries: Project LINK Model(In percentage change from baseline solution)
OECD countriesFirst YearSecond YearThird Year
Real GDP−0.5−0.6−0.4
Implicit price deflator−0.1−0.1−0.3
Value of exports−3.1−3.1−2.3
Value of imports−1.3−1.6−1.5
Trade balance (in billions of U.S. dollars)−11.2−9.5−6.6
Developing countries
Real income−1.7−1.8−1.5
Value of exports−1.4−1.7−1.4
Value of imports−8.4−7.1−4.6
Trade balance (in billions of U.S. dollars)
Source: Weinberg (1979), Table IIf.
Source: Weinberg (1979), Table IIf.

Using the value of imports in 1983 as a base, a $20 billion reduction in imports would represent a 14.5 percent fall in the volumes of imports by Group A program countries—a fall that is twice as large as that actually recorded by these countries in 1983, and almost three times as large as the average for 1973–82. Even if some spillover from program countries to other non-oil developing countries is allowed, the size of the fall in imports in these simulations should be large enough to produce an upper-bound estimate of the global effects on output and trade associated with changes in the imports of program countries.

Four findings are apparent from the Interlink model results in Table 17. To start with, the transmission effect of the lower imports of non-OECD countries on real income (GDP) in the OECD group is rather small; the $20 billion import decline induces a 0.3 percent decline in real GDP in OECD countries. This is in part a reflection of the more general empirical conclusion that cross-country expenditure multipliers are much smaller than own-country ones, usually on the order of one-tenth to one-twentieth as large.44 It is also the basis for Hickman and Filatov’s conclusion for 13 OECD countries that:

… it remains true that the cross-multipliers, even in elasticity form, are generally low except for small countries that are close trading partners of larger ones. This implies that independent domestic shocks even in large countries are unlikely to lead to synchronized fluctuations in the industrialized world… .45

The chief implication of this finding is that one should expect the own effects of expenditure-changing policies in Fund programs to be more significant than the cross-country (or global) effects—even when there are some relatively large traders among program countries.

A second finding, implicit in Table 17, is that the final cross-country output effects of changes in the imports of developing countries appear to be about 2½ times as large as the “impact” effects. Because non-OECD countries account for approximately 30 percent of total OECD exports, and because total exports are roughly 20 percent of the GDP the OECD group, one would expect a 2 percent fall in non-OECD imports to generate an initial 0.12 percent fall in OECD GDP (–0.12 = – 2.0 × 0.3 × 0.2). But Table 17 indicates that after the induced domestic and foreign trade effects of this initial income decline are accounted for, the fall in OECD GDP will be 0.3 percent. This implies an international multiplier of about 2½ (–0.3/–0.12). At the same time, Table 17 also suggests that these multiplier effects die off very quickly after one year, with the effect on GDP in the OECD countries being identical in the first and third years. This quick decay in the transmission process reflects the rather short time lags in many of the behavioral relationships in the model (in which, for example, non-OECD regional groups reach their peak propensities to spend foreign exchange earnings on imports within one year), as well as the dampening influence of the moderate export openness (OECD exports/OECD GDP) and export destination (OECD exports going to non-OECD/total OECD exports) ratios. The key message is that any global effects associated with changed imports in program countries should have pretty much run their course within a year of the changes.

Third, Table 17 shows that in tracing out the global effects of changes in imports by developing countries, one also has to be aware of the linkages going in the opposite direction, namely, from induced lower OECD imports to lower non-OECD exports. Indeed, for country groups as broad as those covered by Table 17, the calculations suggest that non-OECD exports fall by half as much (0.9 percent) as non-OECD imports (1.8 percent). This means that if developing countries want a current account improvement of $10 billion, they would need to reduce their imports by approximately $20 billion. Since this “feedback effect” increases with the size of group initiating the import change, it is certainly true that the prospects for external adjustment are different when one country reduces imports on its own than when many countries do so simultaneously.

Fourth, the individual-country results shown in Table 17 reinforce the earlier conclusion from the simple export-destination ratios (Table 7) that changes in imports in program countries are likely to have quite a different impact across supplying countries. In particular, even when the transmission effects on exports are quite similar on supplying countries, the effects on their real output can be quite variable because of inter-country variations in export-GDP ratios. The main reason GDP in the United States is less affected by lower non-OECD imports than other industrial countries is that its export-GDP ratio is much lower.

The simulation results from the Fund’s World Trade Model in Table 18 show similar effects on the trade balance after the $20 billion reduction in imports of non-oil developing countries to those emerging from the OECD Interlink model. For example, the World Trade Model estimates that the trade balance in industrial countries would deteriorate by about $10 billion in the first year after the import shock, whereas the corresponding estimate in the Interlink model is roughly $9 billion.46 The induced effects on export and import volumes in industrial countries are also similar across the two models. In short, there is nothing in the World Trade Model numbers to contradict the earlier inference that a $20 billion fall in the imports of non-oil developing countries should not produce large trade and real output dislocations in the rest of the world.

Finally, we come to the simulation results for the Project LINK model. Because income in both OECD and developing countries is endogenous in this model, these are perhaps the most interesting estimates for the purposes of this study. Two aspects of the Project LINK results in Table 19 are worthy of explicit mention.

First, the cross-country effects of the lower transfers to developing countries on real output and the trade balance are larger in this model than in the other two, although they are still considerably below the magnitudes associated with “throwing the world into or pulling the world out of an existing recession.” More specifically, the $20 billion drop in transfers to non-oil developing countries reduces real GDP in the OECD group by 0.5 percent, export values by 3.1 percent, and import values by 1.3 percent—all within a year. As before, the transmission effects after three years are quite similar to those after one year.

Second, and consistent with the evidence on the size of domestic versus cross-country expenditure multipliers, the import reduction has a much larger (over three times larger) effect on real income in developing countries than on real income in industrial countries. This larger domestic effect arises because, as previously mentioned, the Project LINK model allows a direct role for (non-fuel) import volumes in explaining real income in developing countries.


The simulation experiments reviewed in this section show that changes in imports by program countries affect economic activity in the rest of the world, and in the expected direction. But, just as important, they strongly suggest that the size of such global transmission effects is small. Specifically, even a 7 percent (or $10 billion) fall in the value of imports by program countries (such as occurred in 1983) appears to be associated with only a 0.1–0.2 percent fall in real GNP in industrial countries. This is not the stuff of which global recessions are made, or ended.

The same simulation exercises also indicate: (1) that the lion’s share of these transmission effects on trade and output take place within one year of the import change; (2) that the full or final effect on real GNP in industrial countries, albeit small, is considerably larger countries (or program countries), and the share of (say, two to three times) than the “impact” effect; exports in GNP. Finally, the simulation results imply and (3) that even among the seven largest industrial that the effects of changes in imports of program countries, these induced effects on output differ across countries are likely to be much greater on their own countries because of inter-country differences in both real income and growth rates than on those of their the share of total exports going to non-oil developing trading partners.

    Other Resources Citing This Publication