World Non-Oil Primary Commodity Markets: A Medium-Term Framework of Analysis

An earlier paper by the authors analyzed demand-side factors underlying the short-run fluctuation of non-oil primary commodity prices (Chu and Morrison (1984)). The analysis focused on the impact of economic activity, domestic prices, and the exchange rates of industrial countries on the short-run fluctuation of non-oil commodity prices. The earlier paper showed that a large part of that fluctuation could be explained by the fluctuations of these demand-side variables. The analysis, however, did not adequately take into account the role of supply-side factors: both those relevant for short-run fluctuations of commodity prices (for example, supply shocks) and those relevant for medium-term fluctuations (for example, supply-price dynamics).


An earlier paper by the authors analyzed demand-side factors underlying the short-run fluctuation of non-oil primary commodity prices (Chu and Morrison (1984)). The analysis focused on the impact of economic activity, domestic prices, and the exchange rates of industrial countries on the short-run fluctuation of non-oil commodity prices. The earlier paper showed that a large part of that fluctuation could be explained by the fluctuations of these demand-side variables. The analysis, however, did not adequately take into account the role of supply-side factors: both those relevant for short-run fluctuations of commodity prices (for example, supply shocks) and those relevant for medium-term fluctuations (for example, supply-price dynamics).

An earlier paper by the authors analyzed demand-side factors underlying the short-run fluctuation of non-oil primary commodity prices (Chu and Morrison (1984)). The analysis focused on the impact of economic activity, domestic prices, and the exchange rates of industrial countries on the short-run fluctuation of non-oil commodity prices. The earlier paper showed that a large part of that fluctuation could be explained by the fluctuations of these demand-side variables. The analysis, however, did not adequately take into account the role of supply-side factors: both those relevant for short-run fluctuations of commodity prices (for example, supply shocks) and those relevant for medium-term fluctuations (for example, supply-price dynamics).

This paper presents a medium-term framework of analysis in which supply-price dynamics—the effect of current prices on future supplies through the impact of current prices on investments in productive capacity, and the effect of capacity changes on commodity prices—are one of the building blocks. In addition, the analysis considers the role of supply shocks in the short-run fluctuations of commodity prices in a more systematic fashion than in the earlier paper.

The rest of the paper is organized as follows. In Section I the literature of supply-price relationships in primary commodities is briefly reviewed; in Section II a model of supply-price dynamics is specified; in Section III the model is estimated for commodity groups; and in Section IV the model is used to analyze the sources of commodity price fluctuations in the 1970s and early 1980s. Section V summarizes the analysis and results. The Appendix extends the discussion of a number of topics covered in the main part of the paper.

I. Survey of the Literature

Most previous studies on the subject have addressed either the influence of supply on price or the influence of price on supply and have for the most part analyzed these relationships for individual commodities. These studies can, however, provide a basis for the conceptual formulation of a more general model.

An early, rather simple model of supply-price dynamics was presented in the “cobweb theorem” (Ezekiel (1938)). In this model output is determined by the price in the previous period. When combined with a demand function, the model produces the “cobweb” pattern of dynamic price movements. The cobweb model, with various adaptations, has been applied to a number of agricultural commodities (Dean and Heady (1958), Harlow (1960), French and Bressler (1962)). Most subsequent studies, however, have expanded on one or the other of the two-way causation between supply and price.

Several studies have assessed the role of supply-side factors in commodity price fluctuations. Cooper and Lawrence (1975) discussed how supply-side factors tended to reinforce the commodity price boom of 1972-75—especially for food, but also for agricultural raw materials and metals. The supply-side factors (output, stocks, and refinery capacity) in equations that explain changes in metal prices relative to prices of manufactures were statistically significant, but their attempts to introduce supply variables for agricultural raw materials were not successful.

Bosworth and Lawrence (1982) built on the analysis of Cooper and Lawrence and in general obtained more significant results from supply-side factors. For groups of commodities a reduced-form equation derived from a rational expectations formulation related the commodity prices (deflated by prices of manufactures) to industrial production in market economies, to production and stocks of the commodity, and to a trend variable. Prices were shown to be most sensitive to production changes, with the elasticity in the range of -2 for food and of -4 for agricultural raw materials. For metals the coefficients of both stocks and capacity were shown to be significantly negative, with elasticities of -0.4 and -0.9, respectively.

Hwa (1979) developed a dynamic disequilibrium model of price adjustment in competitive markets. The resultant price equation related real commodity prices (deflated by prices of manufactures) to world consumption and production of the commodity, lagged price, and a trend variable. Hwa tested this equation for seven industrial commodities. The coefficients for both production and stocks were negative for all seven.

Although these studies support the hypothesis of a statistically significant, inverse relationship between commodity supply and price in the short run, they neither decomposed these supply changes into changes in capacity (long-run) and in capacity utilization (short-run) nor identified the determinants of supply changes. These relationships are addressed in the present study.

The literature on the determinants of supply has focused on both short-run and long-run supply responses to price changes. For annual crops, the short-run and long-run supply responses have been estimated with respect to total supply. For perennial crops and metals, variations in output have been decomposed into those attributable to changes in capacity utilization (short-run) and changes in the capacity itself (long-run).

For agricultural commodities, a large body of literature on supply response originated from a basic model developed by Nerlove (Nerlove (1958), Nerlove and Addison (1958), Nerlove and Bachman (1960)). Nerlove’s main contribution was to improve on the naive assumption of the cobweb model—that farmers’ price expectations are solely influenced by the current season’s price. He postulated (1958) that producers are influenced by their perception of what is a “normal” price and therefore used the adaptive expectations scheme developed earlier by Cagan (1956) for a study of demand for money. Desired output (or acreage) was specified as a function of the normal price and exogenous factors affecting supply, such as weather. To measure long-run supply elasticities, Nerlove used basically the same equation with distributed lags and added a time trend variable to represent factors such as technological change.

Many studies have since been conducted to estimate the supply responses of single commodities in individual countries. Several surveys of the studies have been compiled (Askari and Cummings (1976, 1977) and Bond (1983)). Askari and Cummings report more than 600 estimates of supply elasticities, based on Nerlove’s model, for different crops and countries. Although it is difficult to compare supply elasticities from different studies, the results in general conform to expectations: first, long-run supply elasticities tend to be greater than short-run elasticities, and most of the numerical values of the elasticities are in the 0.0-0.3 range; second, the next largest number of estimates falls in the range of 0.34-0.67. One study showed that long-run supply or demand can be less price elastic than short-run supply or demand if uncertainty is introduced (Hoel and Vislie (1983)).

Following Nerlove’s approach, many studies have included different variables to represent the influences of various nonmarket factors on supply. The most common of these variables is weather, represented by such measures as indices of rainfall, humidity, and frost. Parikh (1971) shows that nonmarket factors—such as rainfall, varietal improvements, expansion of irrigation, and total area under cultivation in the previous year—are important factors in determining output levels over time.

Risk or uncertainty is a factor that in general has been recognized to influence producers’ behavior but has been ignored in most empirical studies. Just (1977), however, included a risk variable to account for the influence of uncertainty about expected price on the supply of some annual crops in the United States. The study showed that risk is important in explaining acreage levels (a negative relationship), except for crops that experienced significant government intervention in the market process.

For perennial crops such as coffee and cocoa, rather major adaptations to the Nerlove model have been necessary (French and Matthews (1971), Wickens and Greenfield (1973), Bateman (1965)). As Wickens and Greenfield (1973) showed, the supply decisions for a perennial crop (coffee) can be divided into two parts: long-run decisions about potential production (or supply), and short-run decisions about the proportion of potential production to be harvested in the current season (utilization of potential supply). These authors derived their potential production function from investment theory, which yields a supply function similar to the basic adaptive expectations model of Nerlove, with a distributed lag of current and past prices as explanatory variables. The short-term harvesting decision was determined by recent price levels (a short distributed lag of price) and by the two-year bearing cycle of coffee. Wickens and Greenfield derived the total supply function by adding the short-term harvesting function to the long-run investment function.

For metals, the total supply function can be divided into two parts: exploration and investment in mines (potential supply) and the rate of extraction from existing mines (utilization of potential supply). The literature on the determinants of the rate of extraction of existing mines can be traced back to Gray (1914), who derived an optimal rate of extraction that was based on the maximization of average returns per unit cost. Gray’s study, recognizing the characteristic of exhaustible resources (that present extraction represents a cost in terms of sacrificed future returns), showed that a higher interest rate yields a higher rate of present extraction. Hotelling (1931) developed a dynamic optimization model in which the rate of extraction was determined by maximizing the present value of discounted future net profits. This approach has been used by many subsequent authors with various modifications. Carlisle (1954) focused on the decision concerning the optimal amount of the total deposit to be extracted, given that extraction costs first decrease, but then increase over time, and that the total amount of a deposit is never extracted.

Parish (1938) decomposed metal supply into the short-term rate of extraction and the long-term additions to capacity from investment in new mines. He showed that uncertainty and a high rate of interest encourage a higher current rate of extraction but discourage investment in new mines. Herfindahl (1955) introduced the exploration function into the dynamic theory of metal supply. In his model, the basic determinant of the rate of both exploration and investment was expected profits; given free entry and competition, prices tend to fluctuate around the long-run cost because of adjustments in investment, production, and price. Herfindahl later (1967) provided empirical support for his theory by finding a significantly positive relationship between the rate of exploration and expected profit in the U.S. petroleum industry.

A more recent study has provided a representative model of total metal supply that includes many of the relationships that had been developed previously (Anders and others (1980)). Several modifications have been made to this basic model to account for special situations such as monopoly, recycling, and the case of joint products. With respect to monopoly, Hotelling (1931) showed that monopolists tend to reduce the rate of extraction but not the rate of exploration. Various other studies (for example, Weinstein and Zeckhauser (1975), Sweeney (1977), Stiglitz (1976), and Kay and Mirrlees (1975)), however, have presented examples of monopolies that extract at the same rate as, or more than, perfect competitors. Recycled materials such as scrap copper have become important sources of supply for many metals, especially in locations where costs of pollution control are large. Schulze (1974) and Weinstein and Zeckhauser (1974) developed models whereby recycled supply is determined in a similar fashion to mined supply by maximizing the present value of expected profits. Finally, with respect to joint products from the same deposit, Herfindahl (1955) showed that this complication makes it impossible to maximize discounted profit for each metal simultaneously.

The present study applies the common underlying approaches to supply responses adopted in previous studies to a study of aggregate commodity prices. Non-oil primary commodities have been shown to be characterized by competitive market conditions. Supply of individual commodities has been shown to be determined primarily by producers’ profit expectations, represented by a variety of commodity price and production cost formulations.

II. The Model

With respect to the model presented, this section contains an overview of competitive price formation, a discussion of supply determinants, and an exposition of supply-price dynamics.

Overview of Price Formation

World non-oil commodity markets are in general competitive, although distortions do exist for some commodities. As in many other studies, therefore, the model introduced in this paper is a model of competitive price formation, rooted in the assumption that the market is cleared fairly rapidly and efficiently through adjustments in price. The analytical framework of the model is summarized in the following paragraphs.

Supply of commodities is determined as a result of producers’ profit maximization and is affected by shocks generated by non-economic factors (for example, adverse weather). Demand for commodities is determined by world economic activity and by the relative prices of commodities in relation to the prices of substitutes. In addition, long-run factors (for example, technological change and population growth) may be important but are not explicitly incorporated in the model. Supply may change as a result of changes either in potential supply (capacity) or in its utilization. Potential supply of commodities is comparable to capacity output in manufacturing; utilization of that supply is comparable to capacity utilization in manufacturing. The analogy is obvious in the case of mining. As Wickens and Greenfield (1973) have shown, however, similar comparisons may be made even for agricultural crops.1

The overall process of price determination may be summarized as follows. Prices are determined at the intersections of demand and supply schedules. In the short run, potential supplies are fixed, but the utilization can change. As shown later, it is useful to conceptualize an even shorter interval (less than a year); during this shorter period the supplies of annual agricultural crops are fixed, whereas those of some agricultural raw materials and metals are price elastic.2 In the medium run, both the utilization of potential supplies and potential supplies themselves can change. Dynamic interactions between supplies and prices in a world of uncertainty, together with shifts in demand, set the stage for the medium-term fluctuation of commodity prices that Herfindahl (1955) envisaged for metals. For example, low levels of supplies of commodities, reflecting capacity constraints caused by low levels of investments in earlier years, would raise prices above long-run average production costs; the high prices could encourage investment by affecting the expectations of future prices of existing, as well as new, producers, thus leading to an expansion of potential supply in later years that would reduce prices below the long-run average costs. This reduction of prices could then discourage investment and ultimately lead to capacity constraints. In such a world, prices would therefore tend to fluctuate around the long-run average production costs (including normal profits).

Determinants of Supply

The concept of potential supply, or of potential production, in the context of this discussion may be defined as the level of supply, or of production, that can be maintained with sufficient economic incentives for the given level of fixed inputs under normal production conditions.3

The intended utilization of potential production is determined as a result of producers’ short-run optimization. In addition, actual utilization may change as a result of noneconomic factors: unusually good or bad weather may cause actual production to rise above or to fall below the normal level, thus raising or reducing utilization; a labor strike or civil disturbance could also cause utilization to decline. In this paper, actual production may exceed potential production if production conditions (weather, for example) are more favorable than normal. (See the Appendix for further discussion of potential production.)

From the concepts discussed above, production is specified in the short run as a function of prices, supply shocks, and potential production:


where, in logarithms,

qts = world production of a commodity

ut = utilization of potential production

qct = potential production

pt = output price (the international price of commodities, in U.S. dollars)

est = exchange rates (in relation to the U.S. dollar) of the currencies of exporting countries

pst = domestic price levels in exporting countries

rpst = real price of commodities

sst = supply shocks,

and where α0, α1, α2, and α3 are coefficients (α1, α2, α3 ≥ 0).

Equation (1) defines world production of the commodity, whereas equation (2) specifies the utilization ratio as a function of a distributed lag of real (output) prices faced by producers and of supply shocks; equation (3) defines real prices faced by producers.

The change in potential production is specified as a function of average excess profits in recent years:


where rps¯ is the long-run average of rpst, β0 and β1 are coefficients (β1≥0), and k is a parameter (k > 0). Equations (4) and (5) specify the change in potential production as a function of the average excess profits (that is, the average real prices faced by producers in excess of their long-run average in recent years). Underlying this specification is the assumption that producers form their expectations of future prices on the basis of recent prices. As discussed earlier, it is assumed that, over a sufficiently long span of years, average real prices should approximate the long-run average costs, including normal profits. Equation (5) therefore defines excess profits (erpst). Average excess profits in recent years cause an increase in potential production (through expansion by existing producers or entry of new producers); negative average excess profits effect a contraction of potential production. An increase in potential production may result either from excess profits during the years of distant past (for example, from the maturing of new trees or the completion of construction of mines) or from excess profits during the years of the more immediate past (for example, from the reactivation of old trees or old mines). Coefficient (β0 indicates the long-run rate of expansion of potential production. A positive (β0 implies that potential production expands even without excess profits; this phenomenon can be explained by various factors (for example, technological changes or government subsidization) not accounted for by excess profit.

A Model of Supply-Price Dynamics

A complete system of equations, in change form, may be constructed by adding a demand equation to the supply equations introduced in the preceding subsection as follows, repeating equations (4) and (5).





Equilibrium condition


Equation (6), with the following notation, is a simplified form of the demand equation discussed in Chu and Morrison (1984, pp. 127−28):

qtd = quantity demanded

edt = exchange rates of importing countries (in relation to the U.S. dollar)

yt = industrial production (economic activity) in importing countries

pdt = domestic prices of importing countries,

and where γ0, γ1, γ2, are coefficients (γ1, γ2>0). The system above can explain both the short-run and the medium-term fluctuation of prices. The whole system captures the dynamic interactions between supply and prices over a number of years during which potential production changes in reaction to the fluctuation of real prices. With potential production given as exogenous, the subsystem consisting of equations (1a)(3a) and (6)(8) captures the short-run determination of prices.

III. Estimation Results

The model introduced in the previous section is sufficiently general to represent different types of primary commodities. In this paper the model is applied to non-oil primary commodities, which are classified in four broad groups: food, beverages, agricultural raw materials, and metals. The model is used to test several hypotheses about short-run and medium-run price formation in world commodity markets. (For a description of the structure of world non-oil primary commodity markets and of the statistical data used in the study, see the Appendix.)

Short-Run Fluctuations of Prices and Production

In the short run, producers accept predetermined levels of potential production; however, the utilization of potential production is determined as a result of producers’ short-run optimization. Equation (2), reproduced below, postulates that utilization depends on both current and lagged real prices, defined as the ratios of output (commodity) prices to domestic prices in producing countries (the latter being used as a proxy of input prices). The equation specifies utilization also as a function of supply shocks:


Different characteristics of production, however, yield different supply schedules. For food crops, consisting of food and beverages, production decisions are made at the beginning of a crop year, and it is not easy for producers to revise their production plans during the course of the year. For industrial raw materials, consisting of agricultural raw materials (for example, natural rubber, timber, and wool) and metals, production plans can be revised fairly easily.4 It is also true that, although supply shocks are a dominant factor in the production of food crops, they are relatively less important for industrial raw materials. The utilization equation may, therefore, have two different forms for the group of food crops and for the group of industrial raw materials, as follows.

Food crops:


Industrial raw materials:


Reflecting these characteristics of supply, annual fluctuations of the prices of food crops are determined in large part by the annual fluctuation of production as a result of lagged price changes and supply shocks. In contrast, the prices of industrial raw materials are determined mostly by shifts in demand. For the former group, therefore, the fluctuation of prices should correlate negatively with production fluctuations; for the latter group, price fluctuations, being primarily driven by shifts in demand, should correlate positively with production fluctuations.

This short-run price-production relationship may be empirically tested. In the results reported in Table 1, the rate of change in production is negatively correlated with changes in real prices in the case of food crops, but it is positively correlated for industrial raw materials (Part A). For both groups, the correlation is fairly low (an adjusted coefficient of determination, R¯2, of 0.121 for food crops and of 0.195 for industrial raw materials), but the relationships are statistically significant (at the 5 percent level for the former and at the 1 percent level for the latter).

Table 1.

Short-Run Price-Production Relationships

(Sample period: 1962−82)

article image
Note: The regression was based on pooled time-series and cross-sectional annual data; within each group (agricultural crop group and industrial raw material group), time series for subgroups (food group and beverages group for the former and agricultural raw material group and metal group for the latter) were pooled for the regression. R¯2 is the adjusted coefficient of determination; DW is the Durbin-Watson test statistic; and SEE is the standard error of estimate. One asterisk indicates significance at the 5 percent level, and two asterisks indicate significance at the 1 percent level; t-statistics appear in parentheses.

Includes beverages.

Consists of agricultural raw materials and metals.

The effects of lagged real prices on the utilization ratio are positive for both food crops and industrial raw materials, although for the latter group the coefficient is not statistically significant (Table 1, Part B). These results suggest the importance of the lagged real price variable in the determination of the utilization of potential production for the food crop group, while indicating that lagged real prices may not be a significant factor in the determination of the utilization of potential production for the industrial raw materials group. As mentioned earlier, the greater ability of producers of industrial raw materials to adjust utilization in the short run suggests that current real prices may be a more significant determinant of current production.

Determinants of Changes in Potential Production

In Section II the equations for changes in potential production5 were specified as


where k is an unspecified parameter that should reflect the gestation period of capital investments for potential production. Although it is known that the gestation periods are usually longer for beverages and metals than for food and agricultural raw materials, it is difficult to pinpoint the value of parameter k for each commodity group, particularly because each group consists of fairly heterogeneous commodities.6 Therefore, in this paper the value of k is empirically searched for each group. The estimated parameters for the four groups of commodities, together with the values of k, are reported in Table 2.

Table 2.

Determinants of Changes in Potential Production

article image
Note: R¯2 is the adjusted coefficient of determination; DW is the Durbin-Watson test statistic; and SEE is the standard error of estimate. One asterisk indicates significance at the 5 percent level, and two asterisks indicate significance at the 1 percent level; f-statistics appear in parentheses.

Starting points of the sample periods differ because of the differing values of k.

“CO” means Cochrane-Orcutt correction of the error term, with the order of the autoregressive equation for the error term indicated in parentheses.

The value of k is smaller for the commodity groups with relatively short gestation periods of capital investment (food and agricultural raw materials; four and five years, respectively) than for the commodity groups with relatively long gestation periods (beverages and metals; seven years for both), in confirmation of the hypothesis based on the nature of production technologies. The regressions also reveal strong long-run upward trends in potential production, probably as a result of the long-run expansion of demand for commodities. The long-run trend rate of annual increase in potential production ranges from 1.2 percent for beverages to 3.4 percent for food and agricultural raw materials.7 Around these long-run upward trends, the rate of change in potential production fluctuated as a result of the fluctuation of average past excess profits, as defined in equation (5). The excess profit variable is statistically significant for food, beverages, and metals; it is shown to have a positive, but statistically insignificant, effect for agricultural raw materials. The estimated coefficients range from 0.011 for agricultural raw materials to 0.169 for metals; the coefficients are 0.034 for food and 0.080 for beverages. The explanatory power of the equation is the highest for agricultural raw materials and the lowest for food. The R¯2 for the former is 0.842; for the latter, 0.341. For agricultural raw materials, average excess profits contribute little to the explanatory power; the second-degree polynominal time trend accounts for a large part of the variation in potential production (see footnote 7).

Since the early 1970s, the expansion of potential production decelerated significantly for agricultural raw materials and metals, whereas it slowed down only slightly or accelerated for food and beverages (Table 3). The annual rate of expansion of potential production was 3.9 percent and 1.2 percent for food and beverages during 1962–71; potential production for these two commodity groups continued to expand at similar rates during 1972–82. The rate of expansion of potential production, however, decelerated sharply for the other two groups—from 2.7 percent to 1.1 percent for agricultural raw materials, and from 4.3 percent to 1.0 percent for metals. The estimated equations suggest that the major factor underlying the disparate behavior of potential production was the different profile of excess profits among the various groups. Average excess profits changed from negative to positive for food and beverages between 1962–71 and 1972–82; they deteriorated by a small amount for agricultural raw materials and declined sharply for metals.8

Table 3.

Percentage Change in Potential Production and Excess Profit

article image

Determinants of Short-Run Equilibrium Prices

In the formation of short-run equilibrium prices, potential production is predetermined. In addition, for agricultural crops, even the utilization of potential production is determined by either exogenous (supply shocks) or other predetermined (lagged real prices) variables. Therefore, the reduced-form price equations derived from the system introduced in Section II may be written as follows.

Food crops:


with θ0 = γ01, θ2 = 1, θ3 = 1/γ1>0, and θ5 = γ21 > 0.

Industrial raw materials:


with θ0 = (γ0− α0)/(α1 + γ1), θ1 = α1/(α1 + γ1)>0, θ2 = γ1/(α1 + γ1)>0, θ4 = 1/(α1 + γ1)>0, and θ5 = γ2/(α1 + γ1)>0. In the classification above, food crops include beverages; for industrial raw materials the coefficient α2, for lagged real prices, in the demand equation is suppressed.

The estimated reduced-form price equations for the food crop groups are reported in Table 4 (Part A). In estimating equation (9), the constraint θ2 = 1 was imposed by transposing Δpdt − Δedt to the left-hand side of the equation and defining a new dependent variable, Δpt − (Δpdt − Δedt). In addition, two dummy variables (d1t, d2t) were included in the equation for beverages to account for persistent price increases that followed two consecutive production shortfalls in 1972−73 and 1975–76.9

Table 4.

Reduced-Form Price Equation

(Sample period: 1962–82)

article image
Note: R¯2 is the adjusted coefficient of determination; DW is the Durbin-Watson test statistic; and SEE is the standard error of estimate. One asterisk indicates significance at the 5 percent level, and two asterisks indicate significance at the 1 percent level; /-statistics appear in parentheses.

Note that Δpest = Δest, − Δest, and that Δpedt = Δpdt − Δedt.

Two dummy variables, d1t and d2t, were included in the equation for beverages to account for persistent price increases that followed two consecutive production shortfalls in 1972–73 and 1975–76. The values (and t-statistics) were: d1t, 0.240 (1.46); d2t, 0.705** (4.22).

As expected, the estimated price equations indicate that production is the dominant factor in annual fluctuations of food and beverage prices. Economic activity (industrial production) in consuming countries is shown to be fairly significant, although the r-ratios are lower than those for the production variables. It should be noted that underlying the specification of the equations is a maintained hypothesis that inflation (ΔpdtΔedt) in importing countries is a significant variable in the determination of commodity prices.

The reduced-form price equations for agricultural raw materials and metals have two variables that are highly correlated. Inflation in exporting countries (Δpest = Δpst − Δest) and that in importing countries (Δpedt = Δpdt − Δedt) are correlated for both groups of commodities because industrial countries dominate in both exports and imports of these commodities. Because of this multi-collinearity, to simply regress Δpt, on the two inflation variables and other explanatory variables would not yield valid statistical results for the two inflation variables. To test the separate effects of inflation in exporting and importing countries on commodity prices, the following procedures are used.

Note that θ1 and θ2 in equation (10) should sum to unity. The equation may therefore be rewritten as


Although estimating equation (11) does not give rise to the multicollinearity problem, the effect of the inflation variables (Δpst − Δest, Δpdt − Δedt) cannot be assessed separately. For such an assessment, the coefficients obtained from the estimated equation (11) are used to derive the following two equations:


The estimation results for equations (11), (12), and (13) are reported in Table 4 (Part B). The value of the coefficients θ1 suggested by the estimated equation (10) are 0.663 for the agricultural raw material group and 0.582 for the metal group. The equation being based on the redefined inflation variable, the statistical nonsignificance of the coefficient of the inflation variable is not a valid result for the impact of world inflation on commodity prices. In contrast, the significance of inflation in both exporting and importing countries for the commodity price fluctuations is well demonstrated by the second-step regressions of equations (12) and (13), for both agricultural raw materials and metals. Being based on small samples, the second-step regressions do not yield estimates of θ1 and θ2 that sum to unity. It is noteworthy, however, that for both groups the average of the θ^1 from equation (12) and (1 − θ^2) derived from equation (13) are close to estimates of θ1 based on equation (11). It is also noteworthy that the magnitudes of the impact of inflation in exporting countries on commodity prices are greater than those of importing countries for both agricultural raw materials and metals, thus suggesting the importance of the cost-push channel in commodity price changes.10

In Table 5, the estimated reduced-form coefficients of the variables for inflation and economic activity are aggregated on the basis of the export value shares for 1979–81. Constrained to sum to unity, the coefficients for the inflation terms are estimated at 0.267 for the inflation in exporting countries and 0.733 for the inflation in importing countries. The coefficient for the industrial production variable is estimated at 2.040. The earlier study by the authors (Chu and Morrison (1984, pp. 116–17)) suggested 1.2 for the coefficients of inflation in industrial countries and 2.0 for industrial production. Thus, the results of the present study are broadly comparable with those of the earlier study, but the present analysis delineates the channels through which inflation in industrial countries is transmitted to world commodity markets. Contrary to the implications of the earlier analysis, the present study suggests that only about 70 percent of the inflation in industrial countries is transmitted through the substitution channel, whereas the rest is transmitted through the cost-push channel. This result reflects the dominant position of the industrial countries as exporters of primary commodities.

Table 5.

Estimates of Coefficients for Reduced-Form and Structural Equations

(Sample period: 1962–82)

article image

Derived as weighted averages of the coefficients for individual groups.

See Table 4, note b, for values of dummy variables included in the reduced-form price equation for beverages.

IV. Supply-Price Dynamics and Commodity Price Fluctuation

In this section a model of supply-price dynamics is used to analyze how various supply and demand shocks give rise to commodity price fluctuations. The first subsection assesses the model’s capability to trace the historical paths of commodity prices. The second analyzes the sources of commodity price instability since 1969. The third subsection discusses how prices respond to a demand shock; the simulation thus shows indirectly how supplies and prices interact dynamically, either intensifying or mitigating the effects of short-run disturbances on commodity prices. The last subsection shows how the model can be used to simulate commodity prices during 1983–85.

The Model

The system estimated in Section III for four groups of commodities forms a larger system for the non-oil primary commodities as a whole. With such a model, total non-oil commodity prices may be simulated on the basis of the simulated prices for individual commodity groups.

The system is summarized in Table 6, and the summary statistics of a dynamic simulation for commodity prices based on the historical values of the exogenous variables, including the estimated supply shocks for food and beverage prices (explained later in this section), are reported in Table 7. In addition, the actual and simulated values of overall commodity prices are compared in Chart 1. The model is simple. A more sophisticated model— incorporating adequately the role of expectations and stocks and relying on improved estimates of potential production—would yield better estimates of parameters. Nevertheless, the system traces the movements of commodity prices fairly accurately.11

Table 6.

The Estimated Model

article image
article image
Note: The subscript t, for time, has been dropped in the notation of the equations.

The aggregation of group price indices was “arithmetic” for the logarithms of prices, or “geometric” for their antilogarithms; this methodology differs from the methodology (“arithmetic” for the antilogarithms) used to derive the current Fund index of non-oil commodity prices (see the Appendix).

Table 7.

Dynamic Simulation of Prices

article image
Chart 1.
Chart 1.

Actual and Simulated Commodity Prices, 1969–82

Citation: IMF Staff Papers 1986, 001; 10.5089/9781451956726.024.A006

Analysis of Sources of Commodity Price Instability

The model incorporates several sources of commodity price instability. On the supply side, supply shocks for food and beverages, and domestic prices and exchange rates of exporting countries for all four groups of commodities, are included. On the demand side, economic activity (industrial production) of importing countries as well as their domestic prices and exchange rates are included. To assess the extent of the contributions made by these variables to the instability of overall commodity prices, prices are simulated on the basis of the assumption that each of these variables, one at a time, had been completely stabilized at its long-term trend.

Such an assessment can be made in either of two ways. First, the simulation results based on the historical time path of an exogenous variable can be compared with those based on the assumption that the exogenous variable had been stable, with all the other exogenous variables assuming historical values but with the error terms of all the stochastic equations replaced by zeros. Second, a comparison of the simulation results can be made that is based on the same assumptions as above, except that the residuals of the stochastic equations (as estimates of the error terms) assume historical values. The second method is used in this section; therefore, “simulated prices” with historical values of exogenous variables are the same as the actual historical prices. The results are summarized in Table 8. In item (2a) of the table the extent of the impact of supply shocks on total non-oil commodity prices is assessed. As already indicated, supply shocks are incorporated only for food and beverages. The effects of the supply shocks are estimated by equation (2a’):

Table 8.

Sources of Commodity Price Instability

(Sample period: 1969–82)

article image

The standard deviations were estimated by regressing the logarithms of prices on the linear time trend.


which may be written with a stochastic error term e as follows:


Both zt and et are not observed. They should not, however, correlate with rps−1, so that regressing Δut on Δrpst−1 is statistically valid. The effects of supply shocks are estimated by


which may be a reasonable approximation of such effects on the utilization of potential production. The simulation results reported in item (2a) of Table 8 are based on supply shocks equal to their mean values during the simulation period (1969−82) for both food and beverages.

In item (2b) of Table 8 both domestic prices and exchange rates of exporting and importing countries are assumed to have been fully stable, whereas in items (2b.i) and (2b.ii) domestic prices and exchange rates are stabilized alternately. In item (2c), industrial production is assumed to have been fully stable. Finally, in item (2d) of the table all of these variables are assumed to have been stable around their long-term trends. The results show that the elimination of major sources of price instability (supply shocks, domestic prices and exchange rates in trading countries, and industrial production) reduces the commodity price variability by about 40 percent (from 0.420 to 0.254) for levels and by 60 percent (from 0.161 to 0.064) for rates of change. The results also show that the two dominant sources of the variability of U.S. dollar prices have been fluctuations of industrial production and exchange rates. Stabilization of industrial production would have reduced the standard deviation of changes in price by 23 percent (from 0.161 to 0.124) and that of prices by 10 percent (from 0.420 to 0.379). Stabilization of exchange rates would have reduced the standard deviation of changes in price by 12 percent (from 0.161 to 0.141) and that of prices by about 18 percent (from 0.420 to 0.346).

The simulations that provided the results summarized in Table 8 may be viewed from a different angle. Table 9 shows how the historical paths of prices during the two recent commodity price cycles (1973–77 and 1979–82) would have changed under alternative scenarios.12 The results indicate the contributions made by the various factors. First, supply shocks reinforced the demand-side factors during all phases of the two cycles; without supply shocks, the price fluctuations during the two cycles would have been significantly smaller. Supply shocks reinforced demand-side factors, particularly during 1973–74, 1976–77, and 1981–82. Second, the 1975 decline in prices was in large part a result of the decline in industrial production; without the decline in industrial production, prices would have increased. Third, the 1981-82 decline in prices can be attributed to both the appreciation of the U.S. dollar exchange rate and the decline in industrial production, with the former playing a larger role than the latter.

Table 9.

Sources of Commodity Price Fluctuations

(Cumulative percentage change)

article image

The 1983-85 period is analyzed in Section IV of the paper (under “Simulation of Commodity Prices During 1983-85”), and simulation results for the period are presented in Table 11.

Simulation of Supply-Price Dynamics

In the model an increase in domestic prices or exchange rates (nominal or real) in both exporting and importing countries does not trigger an interaction between supplies and commodity prices; domestic inflation or a change in real exchange rates (exchange rates deflated by domestic prices) is transmitted to commodity prices fully without affecting real prices of commodities for producers, whereas a change in nominal exchange rates accompanied by no change in real exchange rates does not affect commodity prices.

The dynamic interaction between supplies and prices, however, is triggered by any shocks that affect real prices of commodities for producers. For example, other things being equal, a real depreciation of the U.S. dollar in relation to all currencies of importing countries or an increase in demand for commodities resulting from an increase in economic activity results in an increase in real prices of commodities and, through the effect of such increases on expected future prices, triggers increases in production in subsequent years. These increases in production in turn put a downward pressure on prices that reduces production in subsequent years, triggering another round of price increases. In such a situation, therefore, a temporary increase in demand leads to long-lasting changes in commodity prices. To simulate this supply-price dynamic, the following experiment was conducted. Prices were simulated for all the commodities combined, as well as for the four standard groups: food, beverages, agricultural raw materials, and metals. Industrial production was assumed to be 5 percent higher than the actual historical value for 1970, but the same as historical values for all other years. The prices simulated on this assumption were compared with those resulting from historical values of industrial production. The results of this exercise are reported in Table 10.

Table 10.

Effects of a Demand Shock on Prices

article image
Note: The demand shock assumed was a one-time, 5 percent increase in industrial production.

The difference between the two paths of price series suggests how prices interact with supply. On the basis of the effects in the first year, the elasticities—ranging from 1.3 for food to 3.4 for agricultural raw materials—confirm the results that were reported in Table 5. These positive effects of an increase in industrial production on commodity prices partially erode in subsequent years. For food and beverages, high real prices trigger increases in the utilization of potential production in the second year. At the same time, for all commodity groups, potential production begins to respond to the high prices. The resultant decreases in prices peak three to four years after the initial shock for food and agricultural raw materials, and seven years after for beverages and metals. The simulation results suggest that the amplitudes of price oscillations are larger for agricultural raw materials and metals than for food and beverages. In particular, for metals the amplitude of the oscillation is so large that a substantial initial price increase as the result of a demand shock can erode completely in seven years. For all non-oil commodities the elasticity is reduced from 2.0 to 0.3 in 12 years; comparable long-run elasticities range from -1.0 for metals to 1.4 for agricultural raw materials.

The simulation results summarized in Tables 8-10 confirm the authors’ earlier analysis (Chu and Morrison (1984)) of the three demand-side factors underlying commodity price fluctuations since the early 1970s—fluctuations of economic activity in industrial countries, world inflation, and exchange rates. Unlike the earlier study, however, the present study shows how supply-price dynamics can intensify or reduce commodity price fluctuations by increasing (or decreasing) utilization in the short run and potential production in the medium run after initial increases (or decreases) in commodity prices due to demand shifts. Note that, as reported in Table 10, the supply responses that follow an initial demand shock appear to peak around the sixth year after the shock. It is therefore quite plausible that the 1973–74 commodity price boom might have softened somewhat another demand-induced commodity price boom that occurred about six years later, in 1979–80.

Simulation of Commodity Prices During 1983–85

In this subsection the model is used to simulate the changes in commodity prices during 1983–85. For this simulation the actual or estimated values for some exogenous variables (domestic prices, exchange rates, and industrial production) for 1983,1984, and 1985 were available, as were production estimates for food and beverages. Commodity prices increased by 10 percent during 1983-84 but declined by 12 percent in 1985. Commodity price changes for 1983-85 can be simulated on the basis of these exogenous variables and compared with the actual changes for the period. In Table 11 the simulated overall price changes for 1983–85 are compared with actual price changes, and the assumptions underlying the simulations are summarized.13 The overall price changes were simulated by aggregating simulated price changes for individual groups.

Table 11.

Simulation of Changes in Commodity Price Recovery During 1983–85

(Annual percentage change)

article image

Adjusted for exchange rate changes.

Not shown because they were not used as exogenous variables to simulate prices.

The simulations capture the recovery of commodity prices during 1983–84 and the downturn in 1985, although the magnitude of the downturn in 1985 is substantially underpredicted.14

The simulations of price increases for food, agricultural raw materials, and metals were conducted in a straightforward manner. The simulated prices in 1983 and 1984 for beverages, however, were obtained by utilizing the coefficient estimated for the dummy variable to explain the price increase in 1974 in the presence of a recovery in production. An increase in production was accompanied by a price increase during 1983–84. A number of explanations could be given. A major part of world supply of coffee—the dominant commodity in the beverage group—was under the control of the International Coffee Organization (ICO) through an export quota arrangement. The ICO also strengthened the effectiveness of the export control scheme during 1983–84. The droughts in West Africa lowered the quality of robusta coffee produced in that region, reducing the supplies of quality coffee of that brand. The effects of these developments, however, are difficult to quantify. The recovery of about 2 percent for beverage production in 1983–84 also followed a 14 percent decline in 1982. As the two episodes in 1974 and 1977 suggest, the 2 percent production recovery, following a major production failure, probably was not sufficient to reverse the upward pressure on prices.15

The model may be used to isolate the effects of some of the above factors. Three scenarios may be examined. First, had food production (excluding beverages) remained the same in 1983 as in 1982, instead of declining by 4 percent, food prices would have declined by 1 percent rather than increase by 9 percent. A 10 percentage point decline in food prices in 1983 would have lowered the rate of increase in overall commodity prices in 1983 by 4.5 percentage points, from 8.0 percent to 3.5 percent. Similarly, had food production in 1984 only recovered to the 1982 level (that is, increased by 4 percent), instead of increasing by 8 percent, overall commodity prices would have increased by 5 percent rather than by 2 percent. Although these rather simple calculations have to be refined by incorporating the role of stocks, they suggest the substantial impact of supply shocks on the short-run fluctuation of overall commodity prices. To what extent were the declines of food production in 1983 and of beverage production in 1982 attributable to the decline of real food prices in 1982 and of real beverage prices in 1981? Real food prices declined by 6 percent in 1982; real beverage prices by 32 percent in 1981. Estimated equation (2a’) for the change in utilization of potential production suggests an elasticity of utilization of 0.05 for real prices of food and of 0.16 for real prices of beverages. Therefore, although the impact of the decrease in real prices on food production in 1983 was negligible, the effect on beverages in 1982 was perhaps not insignificant.

Second, world metal production declined by 6 percent in 1982; this decline in production suggests a significant increase in the underutilization of productive capacity of metals in 1982. Perhaps the excess capacity of metal production acted to cushion against any increase in metal prices. In this connection, note that potential production data used in this study are estimated from the trend-through-peaks method (see the Appendix). In view of the nature of this method, potential production for agricultural raw materials and metals during the late 1970s and the early 1980s could have been underestimated. In such a case, the effects of excess capacity on agricultural raw materials and metal prices could have been larger. Also note that the model suggests 3 percent increases in potential production of food and beverages in 1983. This projection reflects excess profits estimated for the food and beverage sectors in previous years. Had it not been for this increase in potential production, the effects of adverse weather in 1982-83 on food and beverage prices could have been larger.16

Third, had industrial production of European countries increased in 1983–84 by 6 percent per annum rather than by 2 percent, the rate of increase in overall commodity prices might have been raised by 4 percentage points. This rather large simulated impact on commodity prices is explained by the fact that Europe accounts for about one half of total world imports of non-oil primary commodities.

In 1983–84, domestic prices in both exporting and importing countries declined in general after exchange rate adjustments with respect to the U.S. dollar, reflecting both appreciation of the dollar and (possibly) the efforts of some primary exporting developing countries to depreciate their currencies in real terms. In addition, industrial production grew only modestly in comparison with 1976–77.17 All these factors, except for the supply shocks, appear to have kept the recovery of primary commodity prices fairly modest in 1983–84.

Supply-price dynamics also provide a linkage between current prices and future prices through the effects of current excess profits on future changes in potential production. The low commodity prices and consequential negative excess profits during 1981-82 suggest possible deceleration in the growth of potential production of commodities during 1986-87 in the case of food and beverages and during 1988-89 in the case of agricultural raw materials and metals. If an expansionary phase of the world economic cycles should coincide with either of these periods, commodity prices could increase sharply.

V. Summary and Conclusions

In this paper an econometric model of world non-oil primary commodity markets focusing on supply-price dynamics has been presented. The model incorporates supply-side and demand-side factors, as well as the interactions between supplies and prices over the short term and medium term. The model decomposes production of commodities into potential production and the utilization rate; potential production is shown to respond to medium-term fluctuations of price, whereas the utilization rate responds to short-run fluctuations. The model was used to analyze and estimate the sources of commodity price instability; it was also used to analyze the slow pace of the 1983-84 recovery of commodity prices.

The analysis quantified the impact of supply-price dynamics on fluctuations of commodity prices. It showed that most of the price increase resulting from a demand shock can erode in the long run. For example, the elasticity of commodity prices with respect to industrial production changes from about 2.0 in the short run to 0.3 in the long run. This erosion results from supply responses to the initial price increase. The model traced historical movements of commodity prices fairly closely and confirmed the authors’ earlier findings (Chu and Morrison (1984)): that industrial production and exchange rate fluctuations have been major factors underlying commodity price fluctuations since the early 1970s. The simulations based on the model also showed that short-run supply shocks reinforced demand-side factors during the last two commodity price cycles (1972-77 and 1979-83), a result that also confirmed the earlier findings. Unlike the earlier study, however, the present analysis estimated the extent of the cost-push channel through which inflation in exporting countries is transmitted to world commodity prices. The study showed that, for agricultural raw materials and metals, the cost-push channel is more important than the substitution channel in the short run.

The analysis also showed that low world inflation (after adjusting for exchange rate changes) and a low pace of recovery in European economies were important reasons for the relatively slow recovery of commodity prices after the 1981-82 recession. The model suggests that, had the industrial production of European countries increased in 1983-84 by 6 percent per annum rather than by 2 percent, the recovery in overall commodity prices might have been 4 percentage points higher. On the supply side, simulations showed that changes in world food production in 1983 and 1984 were a major factor affecting food prices.

APPENDIX: Notes on World Commodity Markets, Data, Methodology, and Goodness of Fit

This Appendix discusses topics not adequately covered in the main text. The first section presents an overview of world trade in non-oil primary commodities; the second discusses the statistical data used in the paper; the third explains the methodology used to estimate potential production; and the final section discusses details of the statistics showing the goodness of fit of simulations of the model.

World Trade in Non-Oil Primary Commodities

The share of primary commodities in total merchandise exports remained at about 30 percent during 1979–81, the same as during 1968-70. The share of non-oil primary commodities, however, declined from 23 percent during 1968-70 to 16 percent during 1979–81. Food items accounted for about half of the total value of non-oil commodity trade, agricultural raw materials and metals accounted for 21 percent each, and beverages accounted for the remainder (7 percent) (Table 12). Industrial countries dominated both exports and imports of food, agricultural raw materials, and metals, accounting for about 70 percent of exports and 70–90 percent of imports; for beverages, developing countries accounted for the bulk (79 percent) of exports, whereas industrial countries accounted for the bulk of imports (88 percent). Among industrial countries, the United States and some of the smaller countries outside Europe and Japan were major exporters of food, agricultural raw materials, and metals, whereas Europe, Japan, and the United States were major importers. The United States and Europe were also major importers of beverages.

Table 12.

World Trade in Non-Oil Primary Commodities

(1979–81 average percentage share)

article image
Note: Country coverage roughly coincides with Fund membership.

Statistical Data

Statistical data used in the study were derived as follows.

Commodity Prices

Forty representative international price series of 34 primary commodities were included, with the weights as shown in Table 13. The individual commodity weights for the price indices of commodity groups (food, beverages, agricultural raw materials, and metals) are the same as the weights for the Fund’s old commodity price index, based on the export shares during 1968–70; the commodity group weights for the overall non-oil price index reflect the world export values of these commodities during 1979-81. The Fund’s new commodity price index, in effect as of March 1986, reflects the export shares during 1979–81. The group indices based on the 1968–70 and the 1979–81 shares are closely correlated. The individual prices and aggregate indices were obtained from the Fund’s International Financial Statistics (Washington, various issues).

Table 13.

Weights Used to Aggregate Commodity Prices and Production

article image
Note: Weights are based on world exports; country coverage roughly coincides with Fund membership.

Annual production series from the World Bank data bank were used. Production series in this data bank are, for agricultural products, from the Food and Agriculture Organization’s Production Yearbook (Rome) computer tapes and, for other commodities, from the World Bank’s Commodity Data Bank (Washington, Economic Analysis and Projection Department). Production series were not available for a few commodities in each group, and the aggregation is based only on the commodities for which production series were available. Potential production series were derived as explained in the next section of this Appendix.

Economic Activity, World Inflation, and Exchange Rates

The series used for the variables were economic activity (industrial production), world inflation (exporting and importing countries’ wholesale price indices or consumer price indices if wholesale price indices were not available), and average exchange rates in relation to the U.S. dollar.

To simplify aggregation, major exporters and importers (accounting for at least 2 percent of world total) for each of the four commodity groups (food, beverages, agricultural raw materials, and metals) were included. For each of the variables (industrial production in importing countries, and domestic prices and exchange rates with respect to the U.S. dollar in exporting and importing countries), data for the individual countries were aggregated with the appropriate weights for the sample countries. The structure of weights is summarized in Table 14.

Table 14.

Average Percentage Shares of Exports and Imports

article image

Potential Production

The methodology used to derive potential production is the trend-through-peaks method developed by Klein and Summers (1966). In the absence of relevant statistical data, more rigorous methods such as those employed by Artus (1977) could not be used.18 In addition to the data problem and the usual problems associated with the trend-through-peaks method, applying the idea of potential production to world production of primary commodities gives rise to several conceptual problems. It may not be meaningful to define potential production for a single commodity in some cases. For example, a certain area of land may be used alternately for two competing crops (for example, jute and rice). Therefore, connecting peaks may yield overestimated potential production series for a group of commodities. In these cases, more relevant potential production series may be derived only on the basis of aggregate production series of two competing crops, rather than on the basis of only one crop. Deriving potential production series on the basis of aggregate world production series could also yield biased results unless production peaks coincide in their timing in all producing countries. There is also the question of whether low production because of adverse weather should be regarded as underutilization of potential production or as low potential production.

Although the estimate of potential production as used in this paper needs refinement and improvement, its concept and estimation technique nevertheless represent a reasonable approximation for analyzing supply-price dynamics.

The series of potential production for each of the four commodity groups were derived as follows. First, for each commodity for which relevant production series were available, the years in which world production reached medium-term peaks were identified. Of those years, years in which production peaks occurred because of unusually favorable production conditions (for example, good weather) were excluded. For the remaining years, changes in actual production were compared with changes in prices in the current year or neighboring years. These comparisons proved useful in ensuring that the years chosen were the years during which high demand for commodities led to full or near-full utilization of potential production. Second, for the beginning and the ending years, judgments were made on the possible behavior of potential production for each commodity. Third, the potential production series thus derived for the sample commodities were normalized to 1979-81 (= 100) and were aggregated on the basis of the weights summarized in Table 13.

Goodness of Fit of the Simulations

In the main text of this paper (Section IV), the model’s capability to trace the historical movements of actual overall prices was reported. In Table 15 more detailed results are shown. The model traces both price levels and changes in prices fairly closely for all commodity groups. It also traces the movements of real prices, average excess profits, and potential production fairly well, except for the average excess profits of agricultural raw materials. The poor performance of the model in this respect does not, however, yield equally poor results for potential production because for agricultural raw materials time trend is a more important explanatory variable for changes in potential production.

Table 15.

Correlation Between Actual and Simulated Endogenous Variables

(Sample period: 1969–82)

article image


Islamic Interest-Free Banking: A Theoretical Analysis—mohsin s. khan (pages 1-27)

This paper begins by describing the basic concepts of Islamic banking, focusing in particular on the issue of elimination of the rate of interest from the system. Islam expressly prohibits a fixed or predetermined return on financial transactions but allows uncertain rates of return deriving from risk-taking activities. Consequently, a banking structure in which the return for the use of money fluctuates according to actual profits made from such use would be consistent with the precepts of Islam. The view that profits are legitimate, even though interest rates are not, has provided the basic foundation for the development and implementation of Islamic banking. In such a system profits and losses are expected to be shared between banks and economic agents according to certain predefined rules. The depositor is treated as if he were a shareholder of the bank and is thus entitled to a share of the profits made by the bank. He is not, however, guaranteed the nominal value of his deposit or a predetermined rate of return on the deposit. The system is symmetrical, so that if the bank incurs losses, the depositor is expected to share in these as well, and the nominal value of his deposit would be reduced. On the other side of the balance sheet the bank is also not permitted to charge a fixed rate of interest on loans; instead it must engage in some type of a profit- or loss-sharing arrangement with borrowers. Broadly speaking, therefore, the Islamic banking system can be regarded as an equity-based, rather than an interest-based, system.

On the basis of this concept of equity participation, a relatively simple theoretical model is developed in the paper to examine the workings of the Islamic banking system. It is shown that the Islamic system may well turn out to be better suited than the interest-based, or traditional, banking system to adjust to shocks that can lead to banking crises. In an equity-based system, shocks to asset positions are immediately absorbed by changes in the nominal values of shares (deposits) held by the public in the bank. Therefore, the real values of assets and liabilities of banks would be equal at all points in time. In the traditional banking system, since the nominal value of deposits is guaranteed, such shocks can cause a divergence between real assets and real liabilities, and it is not clear how this disequilibrium would be corrected and how long the process of adjustment would take. The paper concludes that from an economic standpoint the principal difference between the Islamic and the traditional banking systems is not that one allows interest payments and the other does not. The more relevant distinction is that the Islamic system treats deposits as shares and accordingly does not guarantee their nominal value, whereas in the traditional system such deposits are guaranteed either by the banks or by the government.

Long-Run Equilibrium in a Keynesian Model of a Small Open Economy—peter montiel (pages 28-59)

The analysis of stabilization policies in developing countries was greatly enhanced by the development of the monetary approach to the balance of payments. Many theoretical expositions of the monetary approach, however, employ “global monetarist” structural models that are designed to explain the domestic rate of inflation and the balance of payments under full employment. Short-run deviations of output from capacity can be incorporated into such models by postulating sluggish nominal wage adjustment. An earlier paper by the author (Staff Papers, Vol. 32 (June 1985), pp. 179-210) demonstrated that the resulting “Keynesian” model is consistent with the reserve-flow equation that is central to the monetary approach, although the short-run properties of the model are quite different from those of global monetarist versions.

The present paper analyzes the long-run properties of this Keynesian model and finds that many global monetarist results are restored, including the temporary nature of balance of payments deficits, an “offset coefficient” of −1 on changes in the stock of domestic credit, and the dependence of the effects of devaluation on the nature of the accompanying monetary policy. Thus there is no conflict between Keynesian and global monetarist models of small open economies in the long run. Rather, the key empirical question concerns the nature of the economy’s short-run dynamics. This issue is an important one because the social costs associated with a chosen set of policies—and hence with both their desirability from the point of view of the authorities and their credibility from the point of view of the foreign and domestic private sectors—will depend on the particular path the economy travels during the transition to a new long-run configuration.

Financial Crowding Out: Theory with an Application to Australia—Andrew feltenstein (pages 60-89)

The paper presents a theoretical model designed to analyze the extent to which spending by the public sector crowds out production and investment by the private sector. The model is dynamic with two periods, and all agents have perfect foresight. Private enterprises are constrained to cover current expenditures from current revenues, but investment is financed by the sale of bonds. The government, however, issues a combination of money and bonds to finance any deficit it may incur. Future returns on new investment must equal interest payments on debt incurred by the private investor, but government debt is rolled over from period to period.

The model is applied to Australia, and a benchmark solution is calculated for 1981–82, the last two years for which the Australian exchange rate was fixed. The solution is shown to yield a reasonably accurate approximation to the actual outcomes of the economy in those years. Two counterfactual simulations are then carried out. In the first, an increase in government expenditure brings about a slight increase in real income, with no significant change in the proportion of income going to private investment. In the second, an increase in the debt-financed proportion of the government’s deficit is found to have no impact on real income, although there is a decrease in both years in the proportion of private investment in total income. Thus, some crowding out does take place. Areas for future research would involve the introduction of an extended time horizon, as well as an endogenous capital account and floating exchange rate.

Interest Rates, Saving, and Investment in Developing Countries: A Re-examination of the McKinnon-Shaw Hypotheses—lazaros e. molho (pages 90-116)

The McKinnon-Shaw propositions concerning the effects of interest rates on saving, investment, and asset holding in developing countries are re-examined. Both McKinnon and Shaw have argued that raising interest rates could promote saving and investment in developing countries by alleviating financial repression—excess demand for loans and nonprice credit rationing. McKinnon, however, emphasized internal financing possibilities as the effective constraint on capital formation, whereas Shaw argued that debt finance was the relevant constraint. In part as a result of these differences, the McKinnon and Shaw hypotheses have often been viewed as mutually inconsistent.

The paper highlights the intertemporal aspects of the two propositions and shows that they constitute complementary rather than competing theories. In the spirit of McKinnon and Shaw, the paper develops a theoretical model of the individual’s investment-saving decision in a financially repressed economy. The model shows that, when all rates of return are certain, McKinnon’s complementarity hypothesis—that investment is positively related to deposit rates—may hold not only under pure self-finance, but also under partial debt finance. When the return to capital is assumed to be uncertain, however, the complementarity relationship may not hold, even for the case of pure self-finance. The model also shows that the complementarity relationship is inherently intertemporal, with current deposits serving as a conduit for future capital formation.

On the basis of results of the theoretical model, inferences are derived for the behavior of some key macroeconomic variables in developing countries. Aggregate saving, investment, and money holding are shown to be affected by interest rates, with a complex and possibly very long lag. Because most empirical studies on the complementarity hypothesis have failed to take this lag into account, their results may be of questionable value. Moreover, even if the lag in the effect of interest rates were properly accounted for, the quality of aggregate time series data in developing countries would limit the usefulness of regression analysis in estimating saving-investment relationships. Future research should thus place more emphasis on simulations of theoretical models that are based on micro-economic data, and less emphasis on econometric tests.

Wage Indexation and the Real Exchange Rate in Small Open Economies: A Study of the Effects of Fluctuations in Export Earnings—daniel gros (pages 117-38)

Many small countries that export primary commodities are subject to large and unpredictable fluctuations in their export earnings. Adjustment to these fluctuations in export earnings usually requires changes in relative prices such as the real wage rate and the real exchange rate. In many of these countries, however, such adjustments have not taken place because the real wage or the real exchange rate was fixed by indexation rules.

The paper adapts the analytical framework presented recently by Aizenman and Frenkel to the case of a small open economy that is subject to fluctuations in its export earnings and argues that the necessary adjustments in wages and the exchange rate can take place even in the presence of indexation rules. Although wages could still be adjusted to compensate for changes in the general price level, they should also be linked to permanent changes in export earnings. The link between changes in wages (or the exchange rate) and fluctuations in export earnings that would minimize unemployment is shown to depend on the nature of the shocks that affect export earnings and on the structure of the economy. For example, countries that export an exhaustible resource should react more to changes in export earnings than countries that export a renewable resource, since it can be expected that most shocks to the price of an exhaustible resource are permanent, whereas most shocks to the price or output of a renewable resource can be expected to be transitory. Moreover, the more open a country is, the stronger should wages react to changes in export earnings.

The analysis also examines the welfare and employment effects of different types of indexation rules. It is shown that a policy that fixes the real exchange rate is worse than a policy that fixes real wages. If fluctuations in export earnings are not taken into account in the adjustment of wages, then indexing wages only on the domestic component of the consumer price index (CPI) rather than on the full CPI also minimizes the unemployment effects associated with fluctuations in export earnings.

World Non-Oil Primary Commodity Markets: A Medium-Term Framework of Analysis—ke-young chu and thomas k. Morrison (pages 139-84)

In this paper the authors expand their earlier model of non-oil commodity price determination (Staff Papers, Vol. 31 (March 1984), pp. 93–140). The earlier study focused on the demand-side factors underlying the fluctuation of aggregate commodity prices in the short run. The present study discusses a medium-term framework for analysis of commodity price fluctuations. The study takes into account commodity production, which is decomposed into capacity and capacity utilization. Capacity is shown to respond to medium-term price fluctuations, whereas utilization responds to short-term fluctuations. The study shows how an initial price change caused by a change in demand can erode over time through the supply response to the initial price change. In this manner, the model captures the supply-price dynamics over the medium term.

The study also analyzes short-term supply-side factors—such as changes in domestic prices and exchange rates of exporting countries—as well as demand-side variables—such as changes in economic activity, domestic prices, and exchange rates of importing countries. The model traces the historical movements of commodity prices fairly closely. The simulations based on the model show that short-run supply changes reinforced demand-side factors during the two most recent commodity price cycles (1973–77 and 1979–82). The study measures the extent to which inflation in exporting countries is transmitted to world commodity prices through its effects on costs—with the prices of agricultural raw materials and metals affected to a greater extent by inflation in exporting countries (through cost-push effects) than by inflation in importing countries (through substitution effects); the converse holds for food and beverages. The study also measures the effects on commodity prices of recent changes in commodity production and of the relatively slow pace of the economic recovery in Europe after the 1981–82 recession.


Le système bancaire islamique: analyse théorique d’un système qui ne fait pas appel à l’intérêt—mohsin s. khan (pages 1-27)

L’auteur de la présente étude commence par décrire les concepts sur lesquels repose le système bancaire islamique en insistant plus particulièrement sur l’absence de taux d’intérêt dans ce système. Les règles de l’islam interdisent expressément d’assortir les transactions financières d’un taux de rentabilité fixe ou préétabli, mais les activités qui présentent un certain risque peuvent avoir un taux de rentabilité indéterminé. Une structure bancaire dans le cadre de laquelle la rentabilité de l’utilisation de la monnaie fluctue en fonction des bénéfices que son utilisation a procurés est donc compatible avec les préceptes de l’islam. L’opinion selon laquelle les bénéfices sont légitimes même si les taux d’intérêt ne le sont pas a servi de base à l’élaboration et à la mise en oeuvre du système bancaire islamique. Dans le cadre de ce système, bénéfices et pertes doivent être répartis entre les banques et les agents économiques en vertu de certaines règles préétablies. Le déposant est traité comme un des actionnaires de la banque et peut donc recevoir une partie des bénéfices qu’elle réalise. Il n’a toutefois pas la garantie que la valeur nominale de ses dépôts restera inchangée ou qu’il obtiendra un taux de rentabilité préétabli. Le système fonctionne de manière symétrique; autrement dit, si la banque subit des pertes, le déposant doit en assumer une part et la valeur nominale de ses dépôts s’en trouvera réduite d’autant. En contrepartie, la banque n’est pas autorisée non plus à percevoir un taux d’intérêt fixe sur les prêts. Elle doit, au contraire, se mettre d’accord avec ses emprunteurs sur un système de partage des bénéfices ou des pertes éventuels. De manière générale, le système bancaire islamique peut donc être considéré comme un système de prise de participation et non pas comme un système reposant sur l’intérêt.

En se fondant sur ce concept de participation au capital, l’auteur élabore, dans le cadre de la présente étude, un modèle théorique relativement simple pour examiner le fonctionnement du système bancaire islamique. Il montre qu’en définitive ce système se prête peut-être mieux que le système bancaire traditionnel, fondé sur les taux d’intérêt, aux ajustements qu’exigent les brusques changements qui risquent de provoquer des crises bancaires. Dans un système fondé sur une participation au capital, les chocs au niveau des avoirs sont immédiatement absorbés par une variation de la valeur nominale des parts (dépôts) détenues par le public dans les banques. La valeur réelle des avoirs et celle des engagements des banques sont donc égales à tout moment. Dans un système bancaire traditionnel, où la valeur nominale des dépôts est garantie, ces chocs risquent d’engendrer un écart entre avoirs réels et engagements réels et, à première vue, il n’est pas du tout clair comment un tel déséquilibre, s’il venait à se produire, pourrait être corrigé et combien de temps il faudrait pour opérer l’ajustement. L’auteur conclut que, d’un point de vue économique, la principale différence entre le système bancaire traditionnel et le système bancaire islamique ne réside pas dans le fait que le premier autorise les paiements au titre des intérêts tandis que le second ne le fait pas. Il existe entre les deux systèmes une distinction plus importante: le système islamique considère les dépôts comme des prises de participation et ne garantit donc pas leur valeur nominale, tandis que, dans le système traditionnel, la valeur de ces dépôts est garantie par la banque qui les accepte ou par l’Etat.

Equilibre à long terme dans un modèle keynésien décrivant une petite économie ouverte—peter montiel (pages 28-59)

L’analyse des politiques de stabilisation dans les pays en développement a grandement bénéficié de l’élaboration de l’approche monétaire de la balance des paiements il y a plus de dix ans. Bon nombre d’exposés théoriques de l’approche monétaire font toutefois appel à des modèles structurels “monétaristes globaux”, conçus pour expliquer le taux d’inflation intérieur et la balance des paiements en situation de plein emploi. Il est néanmoins possible de prendre en compte les écarts à court terme entre le niveau de production effectif et la capacité dans ces modèles en postulant que l’ajustement des salaires nominaux ne s’effectue que lentement. Comme l’a démontré une étude effectuée en 1985, le modèle “keynésien” qui en résulte est compatible avec l’équation de flux de réserves qui est l’élément central de l’approche monétaire, bien que les propriétés à court terme du modèle diffèrent grandement de celles des versions monétaristes globales.

L’auteur analyse les propriétés à long terme de ce modèle keynésien et conclut qu’il permet de retrouver bon nombre des résultats des études monétaristes globales; à cet égard, le modèle fait notamment ressortir que les déficits de balance des paiements sont temporaires, que les variations du stock de crédit intérieur sont affectées d’un “coefficient de compensation” égal à −1, et que les effets des dévaluations dépendent de la nature de la politique monétaire qui les accompagne. Il n’y a donc pas d’opposition entre le modèle keynésien et le modèle monétariste global décrivant de petits pays à économie ouverte dans l’optique du long terme. De fait, la principale question empirique qui se pose tient à la nature de l’évolution à court terme de l’économie. Il s’agit là d’une question importante, car les coûts sociaux associés aux trains de mesures retenus—et, partant, l’attrait qu’ils présentent pour les autorités et leur crédibilité pour le secteur privé intérieur et le secteur privé extérieur—dépendent du sentier suivi par l’économie pendant la période de transition qui la mènera vers une nouvelle configuration à long terme.

Éviction en matière de financement: théorie et application au cas de l’Australie—andrew feltenstein (pages 60-89)

La présente étude décrit un modèle théorique qui doit permettre de déterminer dans quelle mesure les dépenses du secteur public évincent la production et les investissements du secteur privé. Le modèle est un modèle dynamique sur deux périodes et tous les agents économiques prévoient parfaitement l’avenir. Les entreprises privées sont assujetties à une contrainte: elles doivent financer leurs dépenses de fonctionnement au moyen de leurs recettes de fonctionnement, et leurs investissements par la vente d’obligations. Cependant, l’Etat émet conjointement monnaie et obligations pour financer les déficits qui peuvent apparaître. Le rendement futur des nouveaux investissements doit être égal aux paiements des intérêts sur la dette contractée par l’investisseur privé, mais la dette publique est reconduite d’une période sur l’autre.

Le modèle est appliqué au cas de l’Australie, et il est procédé au calcul d’une solution de référence pour 1981-82, qui sont les deux dernières années pendant lesquelles le taux de change de la monnaie australienne est demeuré fixe. Il est démontré que la solution du modèle constitue une approximation relativement exacte des résultats enregistrés par l’économie au cours de ces années. Il est alors procédé à deux simulations théoriques. Dans le premier cas, un accroissement des dépenses publiques donne lieu à une légère augmentation du revenu réel, le pourcentage du revenu consacré aux investissements privés ne faisant apparaître aucune variation significative. Dans le deuxième cas, il ressort qu’une augmentation du pourcentage du déficit de l’Etat financé par l’endettement n’a aucun effet sur le revenu réel, bien que le pourcentage du revenu total représenté par les investissements privés diminue au cours des deux années. Il se produit donc un certain degré d’éviction. Les études qui seront consacrées à l’avenir à cette question pourraient faire intervenir un nombre de périodes plus élevé et introduire le solde des opérations en capital sous forme de variable endogène, ainsi qu’un taux de change flottant.

Taux d’intérêt, épargne et investissement dans les pays en développement: réexamen des hypothèses de McKinnon et de Shaw—lazaros e. molho (pages 90-116)

L’étude est consacrée à un nouvel examen des propositions énoncées par McKinnon et Shaw quant à l’effet des taux d’intérêt sur l’épargne, l’investissement et la détention d’actifs dans les pays en développement. McKinnon comme Shaw soutiennent qu’un accroissement des taux d’intérêt pourrait encourager l’épargne et l’investissement dans les pays en développement en remédiant à l’asphyxie financière dont ils souffrent et qui se caractérise par l’existence d’une demande excédentaire de prêts et un rationnement du crédit par des moyens autres que les prix. McKinnon insiste toutefois sur le fait que les possibilités d’autofinancement constituent vraiment la contrainte qui pèse sur la formation de capital; pour Shaw, en revanche, c’est le financement par endettement qui est la contrainte pertinente. D’ailleurs, c’est en partie du fait de ces différences que les hypothèses de McKinnon et de Shaw ont souvent été jugées incompatibles.

La présente étude met en relief les aspects “intertemporels” des deux propositions et montre que, loin de s’opposer, elles se complètent. En s’inspirant des démarches suivies par McKinnon et Shaw, l’auteur de la présente étude élabore un modèle théorique décrivant les décisions des particuliers en matière d’épargne et d’investissement dans une économie souffrant d’asphyxie financière. Le modèle montre que, lorsque tous les taux de rentabilité sont connus avec certitude, l’hypothèse de complémentarité de McKinnon—selon laquelle il existe une relation positive entre l’investissement et les taux servis sur les dépôts—peut demeurer valable non seulement en situation d’autofinancement pur, mais aussi lorsque le financement est partiellement assuré par l’endettement. Si l’on suppose toutefois que le taux de rentabilité du capital n’est pas connu avec certitude, il se peut que la relation de complémentarité ne soit pas valable, même en situation d’autofinancement pur. Le modèle montre en outre que la relation de complémentarité est essentiellement “intertemporelle”, les dépôts courants servant de véhicule pour la formation future de capital.

Certaines conclusions relatives au comportement de diverses variables macroéconomiques clés dans les pays en développement sont tirées des résultats du modèle théorique. Il en ressort que l’épargne globale, l’investissement et la détention de monnaie sont fonction des taux d’intérêt, dont l’effet se fait sentir avec un retard qui peut être très long et a une structure complexe. Parce que la plupart des études empiriques consacrées à l’hypothèse de complémentarité omettent de prendre ce décalage en compte, leurs conclusions sont entachées d’un certain doute. De surcroît, même si le retard avec lequel les taux d’intérêt font sentir leurs effets est correctement pris en compte, la qualité des séries de données globales relatives aux pays en développement limite l’utilité d’une analyse de régression pour l’estimation de la relation entre épargne et investissement. Les études qui seront consacrées à l’avenir à ce sujet devraient donc accorder une place plus importante à des simulations de modèles théoriques faisant appel à des données microéconomiques et accorder une place moins importante aux tests économétriques.

Indexation des salaires et taux de change réel dans de petits pays à économie ouverte: étude des effets des fluctuations des recettes d’exportation—Daniel gros (pages 117-38)

Les recettes d’exportation de nombreux petits pays qui exportent des produits primaires subissent des fluctuations profondes et imprévisibles. L’ajustement à ces fluctuations des recettes d’exportation passe généralement par une modification des prix relatifs tels que le taux de salaire réel et le taux de change réel. Dans un grand nombre de ces pays, l’ajustement en question n’a toutefois pas eu lieu parce que le taux de salaire réel ou le taux de change réel étaient fixés par des règles d’indexation.

La présente étude adapte le cadre analytique élaboré par Aizenman et Frenkel (1984,1985) au cas d’une petite économie ouverte dont les recettes d’exportation subissent des fluctuations et fait valoir qu’il est possible d’opérer les ajustements des salaires et du taux de change qui s’imposent même s’il y a des règles d’indexation. Bien qu’il soit possible d’ajuster les salaires de manière à compenser les variations du niveau général des prix, il importe d’établir un lien entre leur niveau et les changements permanents des recettes d’exportation. Il ressort de l’étude que le type de lien à établir entre les modifications des salaires (ou du taux de change) et les fluctuations des recettes d’exportation pour minimiser le chômage dépend de la nature des chocs qui influent sur les recettes d’exportation et de la structure de l’économie. Par exemple, les pays qui exportent des ressources non reconstituables doivent réagir plus fortement à des variations de leurs recettes d’exportation que les pays qui exportent des ressources renouvelables; on peut, en effet, supposer que les brusques changements de la situation économique qui entraînent une variation du prix d’une ressource non reconstituable sont, pour la plupart, permanents, tandis que la plupart de ceux qui provoquent une variation du prix ou de la production d’une ressource renouvelable peuvent être jugés temporaires. De surcroît, plus l’économie d’un pays est ouverte, plus la réaction des salaires à des fluctuations des recettes d’exportation devrait être forte.

La présente analyse s’étend également aux effets de divers types de règles d’indexation sur le bien-être et l’emploi. Elle fait ressortir qu’une politique qui fixe le taux de change réel est plus préjudiciable qu’une politique qui fixe les salaires réels. Si les fluctuations des recettes d’exportation ne sont pas prises en compte lors de l’ajustement des salaires, l’indexation des salaires uniquement sur la composante intérieure de l’indice des prix à la consommation plutôt que sur l’indice global des prix à la consommation permet de minimiser les effets que les fluctuations des recettes d’exportation ont sur le chômage.

Marchés internationaux de produits primaires non pétroliers: cadre d’analyse à moyen terme—ke-young chu et thomas k. morrison (pages 139-84)

Dans la présente étude, les auteurs élargissent le champ du modèle qu’ils avaient antérieurement élaboré pour déterminer le prix des produits non pétroliers (Staff Papers, volume 31 (mars 1984), pages 93 à 140). Dans le cadre de leur première étude, ils mettaient l’accent sur les facteurs relatifs à la demande qui sont à l’origine des fluctuations à court terme du prix global des produits. La présente étude replace l’analyse des fluctuations des prix des produits dans une optique à moyen terme. Il y est tenu compte de la production, qui est décomposée en capacité et utilisation de la capacité. Les auteurs démontrent que la capacité de production réagit aux fluctuations des prix à moyen terme, et l’utilisation de la capacité, aux fluctuations à court terme. Ils décrivent la manière dont une variation initiale des prix due à une modification de la demande peut s’éroder peu à peu sous l’effet de l’évolution de l’offre engendrée par la variation initiale des prix. Le modèle prend donc ainsi en compte la dynamique offre-prix à moyen terme.

Les auteurs analysent en outre les facteurs relatifs à l’offre qui s’exercent à court terme (variations des prix intérieurs et des taux de change des pays exportateurs, par exemple), ainsi que les variables relatives à la demande (variations de l’activité économique, des prix intérieurs et des taux de change des pays importateurs, par exemple). Le modèle retrace assez fidèlement l’évolution des prix des produits dans le passé. Les simulations effectuées à partir de ce modèle montrent que les fluctuations de l’offre à court terme ont renforcé les effets des facteurs relatifs à la demande au cours des deux cycles les plus récents des prix des produits (1973 à 1977 et 1979 à 1982). L’analyse permet d’évaluer dans quelle mesure l’inflation qui sévit dans les pays exportateurs se répercute sur les cours mondiaux des produits par le biais de son incidence sur les coûts; à cet égard, les prix des matières premières agricoles et des métaux sont plus grave