The occurrence of a recession-like decline in commodity prices in 1984–86 during the upswing of the business cycle has raised concerns about the nature and causes of the decline—in particular, whether the causes may be related more to long-term structural factors than to short-term, reversible factors. For example, it has been recently stated that “the primary-products economy has come ‘uncoupled’ from the industrial economy” (Drucker (1986)) and that “economic growth is no longer accompanied by increased consumption of basic materials” (Larson, Ross, and Williams (1986)). The purpose of this paper is to provide an analysis of the causes of the recent decline in commodity prices in order to develop a better understanding of its underlying nature.
The cumulative commodity price decline of 26 percent in U.S. dollar terms during 1984–86 was the largest of the four previous declines that occurred since 1970, including those experienced in the 1975 and 1981–82 recessions, and it comprised declines across a broad range of commodities. The decline in real terms1 (-36 percent) and in terms of SDRs (-35 percent) was even larger. With the 1984–86 decline exacerbating the general weakness that had already characterized commodity markets throughout most of the 1980s, real commodity prices in 1986 reached their lowest levels since at least the 1930s.
In contrast to the two other recent major downturns in commodity prices in 1975 and 1981–82, during which economic activity in the industrial countries also declined, the 1984–86 decline was accompanied by an increase in economic activity. The unusual character of the large 1984–86 commodity price decline was underlined by our finding that an earlier, demand-oriented model (Chu and Morrison (1984)), which had predicted fairly well the four previous commodity price cycles since 1970, predicted a price increase during 1984–86. In the present study, a model (Chu and Morrison (1986)) that includes both supply and demand factors is adapted to explain the 1984–86 decline in commodity prices. In this more fully specified model, commodity prices are determined both by short-term factors influencing capacity utilization and by medium-term factors influencing production capacity. Thus determined, commodity supply, along with the demand factors, then influences current commodity prices.
The supply characteristics of food commodities (including beverages) are different from those of agricultural raw materials and metals because production of the latter groups of commodities can be more easily adjusted to short-run market conditions. Food production, in contrast, is virtually fixed in the short run. Conversely, for agricultural raw materials and metals, production capacity is less easily adjusted in the short-run than is capacity of food production. The reduced-form price equations, therefore, suggested an inverse relationship with current production for food prices, and an inverse relationship with production capacity for prices of agricultural raw materials and metals. The reduced-form equation for food commodities was2
for agricultural raw materials and metals it was
The variables were defined in logarithms as follows:
The price equation for all commodities was simply a weighted aggregation of estimated equations of the four commodity groups.
Equation (2) cannot be estimated directly because two independent variables (inflation in the exporting and importing countries) are highly correlated, since industrial countries dominate in both exports and imports. To solve this problem and obtain valid results for the inflation terms, we again followed the procedure used in Chu and Morrison (1986); that is, by noting that since θ1 and θ2 in equation (2) should sum to unity, one can rewrite equation (2) as
Although this procedure solves the multicollinearity problem, it does not yield separate inflation effects for the producing and consuming countries. For such a disaggregation, the estimated coefficients from equation (3) were used to obtain the following two equations:
Several modifications were introduced in re-estimating the 1986 model with more recent data for this study. Weights used in aggregating the price, production, production capacity, and supply variables were updated and expanded—from average relative shares in developing country exports in 1968–70 to shares in world exports in 1979–81—to reflect better the current structure of world trade in non-oil primary commodities. Country coverage on the demand side was increased from 7 to 13 countries. Geometric weighting procedures were used to give more appropriate weight to extreme observations, particularly in the case of real exchange rate changes in high-inflation countries. Finally, a recently developed index of total supply, which includes beginning stocks plus current production, was used in the present study as the supply variable (Δst) in place of production (Δqt) in the food and beverage equations. Consequently, dummy variables to account for the effects of large stock movements in certain years were no longer necessary.
The estimated reduced-form price equations for the food crops (food and beverages) are reported in Table 1 (upper panel). In estimating equation (1) we followed the method of Chu and Morrison (1986) of imposing the constraint θ2 = 1 by transposing (Δpdt - Δedt) to the left-hand side of the equation and defining a new dependent variable. The results of the estimation show that the dominant factor in yearly variations in food and beverage prices is variations in supply, which, of course, is inversely related to the corresponding price movements. It is also noteworthy that economic activity in consuming countries is significant. This specification incorporates a maintained hypothesis that inflation in importing countries [(Δpdt-Δedt)] is a significant variable in determining commodity prices. The opposite is true of price changes in producing countries, which is another way of saying that we have assumed a short-run price elasticity of supply of zero for the food crops.
The estimation results for industrial raw materials (agricultural raw materials and metals) are reported in the lower panel of Table 1. Industrial production is statistically the most significant variable, with an elasticity of 1.8. The results suggest that inflation only in the consuming countries is important in influencing commodity prices. With respect to the variable for production capacity, the results seem to suggest that the medium-term supply response through expansions in output capacity could be quantitatively important for agricultural raw materials (although the coefficient is not significant at conventional levels of confidence). Finally, the estimated constant term for all commodities shows a downward drift in commodity prices of 5 percent a year that is probably associated with structural shifts that tend over time to reduce demand (for example, substitution of synthetic materials for natural materials) and to increased supply (for example, technological change embodied in new high-yielding crop varieties).
|Commodity||Overall||Equation||Constant||Δpest - Δpedt||Δpest||Δpedt||Δst||Δqct||Δyt|
|Industrial Raw Materials|
The model does predict the beginning of a new cycle in 1984–85, with overall commodity prices predicted to decline by 1.9 percent in 1984 and by 6.6 percent in 1985 (Table 2). In 1986, however, an increase of 10.6 percent is predicted, primarily on the basis of the large depreciation of the U.S. dollar. Demand contributes positively to commodity prices during 1984–86, although at a decelerating rate. Supply factors are the main reason for the predicted decline in commodity prices in 1984–85.
Over the two-year period 1984–85. the negative contributions of food and beverage supplies to overall commodity price movements approximately doubled compared with 1983. Over this period, the supplies of almost all agricultural commodities increased substantially because of generally favorable worldwide weather conditions and policies in producing countries that encouraged continued growth of production, even in the face of falling world prices. Production capacity of agricultural raw materials, and to a lesser extent of metals, exerted a rather constant depressing influence on prices throughout the 1980s until 1986. when it appears that expansion of production capacity finally leveled off. probably in response to the persistently low prices. This is consistent with the mean estimated lags of response of production capacity to real commodity prices that were estimated at five years for agricultural raw materials and at seven years for metals (Chu and Morrison (1986)). Real metal prices began to fall in the late 1970s, and real prices of agricultural raw materials began to fall in the early 1980s.
Not only does the model predict a commodity price decline in 1984–85, but it also reflects the considerable weakness that has characterized primary commodity markets throughout the 1980s. Because of the difficulty in matching the supply variables precisely with annual calendar year data for the other variables, this model is probably better at predicting changes over a number of years than changes year to year. The actual cumulative commodity price decline of 27 percent in 1980–86 is fairly accurately predicted, at 23 percent, by the model. The major influences shown to be depressing commodity prices in the 1980s are the trend term, with a cumulative contribution of -35 percent; the supply of food and beverages, with a cumulative contribution of -19 percent; and production capacity of agricultural raw materials and metals, with a cumulative contribution of -11 percent. These were partially offset by positive contributions of 19 percent for economic activity and 16 percent for inflation (adjusted by exchange rate changes).
Although the annual model does predict a decline in commodity prices during 1984–86, the extent of the actual decline is underpredicted. This still leaves room for the explanations advanced by some observers that structural factors that depress prices and that are difficult to quantify have intensified in the 1980s. These include a reduced intensity of commodity use in industrial countries attributable to the shift away from heavy industry and increased substitution toward lighter and new materials. Because the model predicts fairly well the weakness in commodity prices throughout the 1980s, however, the underprediction of the 1984-86 price decline may be due simply to special factors. It is noteworthy that the weakness in metal prices, which is most frequently associated with the structural explanation, was fairly well predicted by the model in the 1980s.
The underprediction of the 1984–86 commodity price decline (Table 3) occurred largely in 1986 with respect to the food commodities, which at 42.9 percent have the largest weight in the total commodity basket. Although food prices were predicted to increase by 9.7 percent in 1986. they actually declined by 13.9 percent. The reason for the prediction of a recovery in food prices in 1986 was the strong positive contribution from the large depreciation of the U.S. dollar. Two somewhat related reasons may help to explain this prediction error. First, the enactment of the U.S. Farm Bill at the beginning of 1986 had a strong negative influence on food prices that was not captured in the model. This change in agricultural policy of the world’s largest food-exporting country significantly lowered support prices (for example, by about 25 percent for cereals) and resulted in greater supplies for export being made available to the market at sharply lower prices. Second, in part because of the U.S. Farm Bill but also because of the cumulative effects of several years of abundant world food production, price competition in food export markets intensified significantly in 1986, largely negating the positive contribution of the depreciation of the U.S. dollar.
The analysis of contributing factors shows that the main source of the commodity price decline in 1984–86, aside from the downward trend, was the augmentation of supply. It may be appropriate, however, to look at the contribution of demand from a different viewpoint. Economic activity in the industrial countries must grow by over 3 percent per year to offset the negative contribution of the secular decline in prices captured in the trend term. In this sense, the large deceleration in economic activity from 1984 to 1986 represented a depressing influence on commodity prices. The contribution of economic activity net of the trend term declined from 4.8 percent in 1984, to −0.1 percent in 1985, and to −2.7 percent in 1986. Although the depressive impact of the supply variables showed signs of beginning to moderate in 1986 in response to the low commodity prices, the sharply lower economic growth rates more than offset the improved supply situation.
In conclusion, this analysis has shown that, although the large 1984–86 commodity price decline was unusual in the sense that it occurred during an upswing of the business cycle and was caused more by supply than demand factors, the decline should still be seen as largely cyclical in nature and not the beginning of a structural decline. The main causes of the decline that have been identified are short-term and reversible, rather than long-term and irreversible. The fairly strong recovery in commodity prices that began in 1987 is consistent with this interpretation.
Chu, Ke-young. and Thomas K.Morrison, “The 1981–82 Recession and Non-Oil Primary Commodity Prices,”Staff Papers, International Monetary Fund (Washington), Vol. 31 (March1984), pp. 93–140.
Chu, Ke-young. and Thomas K.Morrison, “World Non-Oil Primary Commodity Markets: A Medium-Term Framework of Analysis,”Staff Papers, International Monetary Fund (Washington), Vol. 33 (March1986), pp. 139–84.
Drucker, Peter F., “The Changed World Economy,”Foreign Affairs (New York), Vol. 64 (Spring1986), pp. 768–92.
Larson, Eric,MarcRoss, and RobertWilliams.“Beyond the Era of Materials,”Scientific American (New York). Vol. 254 (June1986). pp. 34–41.
Mr. Morrison is a senior economist in the African Department and is a graduate of the University of Maryland.
Mr. Wattleworth is a senior economist in the Research Department and is a graduate of the University of California, Berkeley.
A longer version of this paper (Morrison and Wattleworth (1987)). which reports alternative model tests, is available from the authors, who would like to thank their colleagues in the Fund for helpful comments.
Deflated by the change in export unit values of manufactures of industrial countries.
The system of structural equations corresponding to the reduced-form equations is fully described in Chu and Morrison (1986).