Article

The 1981–82 Recession and Non-Oil Primary Commodity Prices

Author(s):
International Monetary Fund. Research Dept.
Published Date:
January 1984
Share
  • ShareShare
Show Summary Details

The depressed condition of non-oil primary commodity markets during 1981–82 was the worst since World War II. The overall index of annual average prices of primary commodities (other than gold and petroleum) declined by 12 percent in 1982 (in dollar terms), following a 15 percent decline in 1981. The cumulative two-year decline of 25 percent was the largest and longest in more than three decades. During the last three decades, the largest annual decline occurred during the 1975 recession, when primary commodity prices fell by 19 percent; thereafter, they quickly recovered, increasing by 15 percent in 1976 and by 21 percent more in 1977. Commodity prices in real terms, estimated by deflating nominal prices by the United Nations price index of manufactured exports of developed countries, fell by 20 percent in 1981–82 to a postwar low. Commodity prices increased by 6 percent in 1983, but the aggregate index in 1983 was 20 percent below the previous peak attained in 1980 and 6 percent below the average index for 1977–83.

The 1981–82 price decline has had a severe impact on the external balance of primary commodity-exporting countries. After declining by about 10 percent during 1978–80, primarily because of the sharp rise in oil prices, the external terms of trade of non-oil developing countries declined by a further 8 percent in 1981–82, notwithstanding the relative stability of oil prices during this period. Combined with a decline in the volume of non-oil exports, these price developments resulted in large aggregate current account deficits of $109 billion in 1981 and $82 billion in 1982, the average of which was almost twice the average annual deficit during 1977–80.1 The low-income developing countries were most adversely affected because of their higher dependence on primary commodities for export earnings.

The sharp decline in commodity prices during 1981–82 was only the most recent indication of a pattern that was first visible in the early 1970s. Although nominal commodity prices sharply accelerated and there were intermittent surges in real prices, the long- term downward trend in real prices from 1972 to 1982 was more than twice as steep as the like price trend from 1957 to 1971. In addition, the price instability2 of non-oil primary commodity prices from 1972 to 1982 was more than three times greater than it was from 1957 to 1971, reflecting the significantly more unstable economic environment—as seen in the behavior of industrial production, world inflation, exchange rates, and interest rates—in the later period. Commodity price instability was one of the major causes of export instability for a large number of primary commodity exporting countries3; export instability, in turn, may have been a major cause of instability in countries’ external balances and domestic economies.4

This paper analyzes the depressed state of primary commodity markets during 1981–82 in the context of developments over a historical period. The causes of primary commodity price movements are investigated, along with the relatively high price in-stability in recent years.

The organization of the paper is as follows: Section I describes the historical movements of non-oil primary commodity prices, focusing on the long-term movements of broadly aggregated prices vis-à-vis the long-term movements of the prices of manufactures and petroleum, and on their short-run fluctuations; Section II analyzes the determinants of commodity prices, including the major causes of commodity price fluctuations over various phases of commodity price cycles; and Section III presents conclusions. Appendix I shows the derivation of the price equation; Appendix II describes the behavior of the variables underlying commodity price movements; Appendix III includes a list of the commodities; and Appendix IV presents some econometric results not reported in the main text.

I. Historical Perspective

long-term developments

Nominal prices

The behavior of primary commodity prices has undergone a significant change since the early 1970s. After exhibiting a high degree of stability—as seen in the small changes in nominal prices—during the two preceding decades, commodity prices have since 1972 exhibited marked cyclical behavior at significantly higher nominal price levels (see Chart 1 and Table 1).5 During 1973 and 1974, the overall commodity price index approximately doubled, with virtually all primary commodities participating in this sharp upward movement. When the world experienced a major recession in 1975, commodity prices fell by 19 percent from the record high level of the previous year. Although this was the largest one-year decline in the last three decades, prices were still 60 percent higher in nominal terms than they had been in 1972.

Chart 1.Non-Oil Commodity Prices: Long-Term Developments, 1957–83

(1975 = 100)
Table 1.Indices of Non-Oil Primary Commodity Prices, 1957–831
Nominal Prices2Deflated Prices3
YearAllFoodBeveragesAgricultural

raw materials
MetalsAllFoodBeveragesAgricultural

raw materials
Metals
Indices
(1975 = 100)
1957574077745912790171163131
1958533877615511884172138123
1959523865685311785147154121
1960523760715411582134158121
1961503657665410979124145119
1962493756635310680121138115
1963524454655211395118142114
1964554262656511890134139139
1965543957637111382118132149
1966563959657711480121133158
1967524059596510681119119132
1968523960576710678121116137
1969564262617411082121119145
1970584570557810884129103145
19715646655567978111296118
1972635371726810185113116109
197397828712999134113121178136
1974124131104125124141148118143142
1975100100100100100100100100100100
19761158419212410611484188123105
19771398333212811412776303117104
1978133%2411381201067619110995
197915510925516815610876111117109
198016914122417517210689140110108
1981144122174158148958111510598
19821279717813613586661219392
1983135105192149135957413410594
Annual Rate of Change
(In percent)
1958–8–7–1–17–7–7–60–16–6
1959–10–1611–301–1512–2
19600–1–742–2–3–820
1961–4–3–6–70–5–4–7–8–1
1962–22–2–4–2–21–3–5–3
1963619–22–1619–23–1
19647–3150244–513–222
1965–2–7–9–310–4–9–12–57
1966304381–2216
1967–620–9–15–70–2–11–16
1968–1–31–230–31–23
196978461045026
19703513–95–2–17–14–0
1971–52–8–1–14–11–4–13–6–19
19721314931055121–8
19735355237947323365525
1974286019–325531–2–204
1975–19–23–4–20–19–28–32–15–30–29
197615–169224614–1688235
197721–2733712–1061–6–1
1978–416–2785–170–37–7–9
19791613622302–0–7814
1980930–12411–217–21–6–0
1981–15–14–22–10–14–10–9–18–5–9
1982–12–212–14–9–10–196–12–7
19836989091210123

Peaks and troughs in deflated prices for the sample period are indicated by bold face type and italic type, respectively.

Expressed in dollar terms.

Deflated by the United Nations index of prices of manufactured exports of developed countries.

Peaks and troughs in deflated prices for the sample period are indicated by bold face type and italic type, respectively.

Expressed in dollar terms.

Deflated by the United Nations index of prices of manufactured exports of developed countries.

As the world economy recovered in 1976 and 1977, commodity prices rose at annual rates of 15 percent and 21 percent, respectively. This rise was temporarily reversed by a 4 percent decline in 1978, which was due entirely to a 37 percent fall in beverage prices; the prices of food, agricultural raw materials, and metals all increased that year. With a recovery in beverages and continuing increases in other commodities, overall commodity prices rose by 16 percent in 1979 and by 9 percent in 1980.

In 1981, with the world entering another recession, commodity prices declined by 15 percent, with every major commodity group participating in this decline. The recession continued into 1982, and commodity prices fell a further 12 percent, in what turned out to be the largest continuous decline in the last three decades.6 Because the dollar appreciated over 1981–82, the cumulative two-year decline in overall commodity prices in SDR terms was about 12 percent, compared with 25 percent in dollar terms.

Since 1957, non-oil primary commodity prices have increased at a rate of 5.1 percent per year (see Table 2). This long-term period, however, is comprised of two rather distinct subperiods: from 1957 to 1971, when prices increased by 0.4 percent annually, and from 1972 to 1982, when prices increased 6.5 percent annually.7 Virtually all commodity groups followed a similar pattern, with the rate of change in food prices increasing from 1.0 percent during 1957–71 to 4.6 percent during 1972–82, in beverage prices from –0.6 percent to 11.1 percent, in agricultural raw materials prices from –1.6 percent to 5.8 percent, and in metal prices from 2.5 percent to 6.6 percent.

Table 2.Long-Term Price Trends, 1957–821
Annual Percentage Change
1957–821957–711972–82
Nominal Prices2
Primary commodity prices
Non-oil5.10.46.5
Food5.41.04.6
Beverages6.1–0.611.1
Agricultural raw materials4.2–1.65.8
Metals4.62.56.6
Oil13.3–0.328.3
Manufactures5.81.59.8
Real Prices3
Non-oil primary commodity prices deflated by:
Oil price4–7.40.7–16.7
Manufactures price5–0.6–1.1–2.7

The long-term trend is defined as the exponential trend estimated from a semi-log regression of quarterly price on time for each sample period; the rates of change are annualized. The percentage increase for the whole period could therefore be lower or higher than for either 1957–71 or 1972–82.

Expressed in dollar terms.

Deflated by the manufactures price index referred to in footnote 5.

Oil price is the weighted average of the official prices of Libya, Saudi Arabia, and Venezuela.

Manufactures price index is the United Nations index of prices of manufactured exports of developed countries.

The long-term trend is defined as the exponential trend estimated from a semi-log regression of quarterly price on time for each sample period; the rates of change are annualized. The percentage increase for the whole period could therefore be lower or higher than for either 1957–71 or 1972–82.

Expressed in dollar terms.

Deflated by the manufactures price index referred to in footnote 5.

Oil price is the weighted average of the official prices of Libya, Saudi Arabia, and Venezuela.

Manufactures price index is the United Nations index of prices of manufactured exports of developed countries.

Real prices

The average annual growth rate of 5.1 percent from 1957 to 1982 in non-oil primary commodity prices was slower than the rates for both oil prices and manufactures prices during the same period. Vis-à-vis oil prices, primary commodity prices fell by 7.4 percent per annum (86 percent cumulatively); vis-à-vis prices of manufactures, they fell by 0.6 percent per annum (14 percent cumulatively) (see Chart 2). The declines in relative prices were particularly notable during 1972–82.

Chart 2.Index of All Non-Oil Primary Commodity Prices: Nominal and Real1, 1957–83

(1975 = 100)

1 Deflated by the United Nations price index of manufactured exports of developed countries.

Because the short-run fluctuation of overall manufactures prices has been much milder than that of primary commodity prices, the short-run fluctuation of overall real commodity prices roughly mimicked that of nominal commodity prices. However, there were two notable divergences in their short-term movements: (i) during the 1975 recession, the decline in real prices was much smaller than the decline in nominal prices; and (ii) in 1979–80, although nominal prices increased substantially, real prices hardly rose because inflation of manufactures prices was high.

price instability

Overall price

Perhaps the most salient feature of commodity price behavior since the early 1970s, compared with that during the rest of the postwar period, is the marked increase in price instability. Although this instability is illustrated in Chart 1, it may be useful to investigate further its characteristics. This subsection describes fluctuations around both long-term and medium-term trends.8

In order to examine the behavior of commodity price fluctuations since 1957, the period has been divided into two subperiods, 1957–71 and 1972–82. Although the choice of these two periods is somewhat arbitrary, it is based on an examination of Chart 1 and, as explained later in the subsection entitled “estimation results” in Section II, on the recognition that 1971–72 was the approximate time at which major variables affecting commodity prices became much more unstable.

Primary commodity price instability, measured by the average percentage deviation of the overall index of quarterly prices from their long-term trends, increased more than threefold between 1957–71 and 1972–82 (see Table 3); the instability, measured as the average of the instabilities of individual prices, doubled; and similar conclusions are obtained if instability around the medium-term trend is examined. Oil prices underwent an even greater change in instability, which increased more than fivefold from 1957–71 to 1972–82 measured around long-term trends, and more than tenfold measured around medium-term trends. Prices of manufactures also experienced higher instability, increasing threefold measured around long-term trends and fivefold measured around medium-term trends.

Table 3.Instability of Primary Commodity Prices in Dollar Terms, 1957–82(In percent)
1957–821957–711972–82
(A)(B)(A)(B)(A)(B)
Around the Long-Term Trend1
Primary commodities
Non-oil: based on:
Aggregate index1721451320
Aggregation of individual instability indices2242811142227
Food1923561924
Beverages32369113136
Agricultural raw materials2025461418
Metals12169101216
SDR terms141745914
Real terms381045813
Oil45056572731
Manufactures15172369
Around the Medium-Term Trend
Primary commodities
Non-oil: based on:
Aggregate index7310
Aggregation of individual instability indices213818
Food9414
Beverages10423
Agricultural raw materials9410
Metals10710
SDR terms7310
Real terms37310
Oil410114
Manufactures315

Instability in column (A) is measured by the average of the absolute values of percentage deviations of the quarterly price from the trend: the long-term trend is estimated by the semi-log regression of the quarterly price on time, the medium-term trend by the 19-quarter moving average of the actual price. In estimating the instability around the medium-term trends, 1957–58 and 1981–82 are excluded from the sample years because the 19-quarter moving average of the prices could not be obtained for these years.

Instability in column (B) is the standard error of estimate (in percent) of the semi-log regression.

Figures represent the weighted average of instability indices of all individual non-oil primary commodities, which reflect offsetting price movements not reflected in the aggregate index.

Deflated by the United Nations index of prices of manufactured exports of developed countries.

Unlike other prices, oil prices in nominal terms were relatively stable except in 1973–74 and 1979–80, when they increased sharply; interpretation of the instability index should therefore take into account this particular behavior of oil prices.

Instability in column (A) is measured by the average of the absolute values of percentage deviations of the quarterly price from the trend: the long-term trend is estimated by the semi-log regression of the quarterly price on time, the medium-term trend by the 19-quarter moving average of the actual price. In estimating the instability around the medium-term trends, 1957–58 and 1981–82 are excluded from the sample years because the 19-quarter moving average of the prices could not be obtained for these years.

Instability in column (B) is the standard error of estimate (in percent) of the semi-log regression.

Figures represent the weighted average of instability indices of all individual non-oil primary commodities, which reflect offsetting price movements not reflected in the aggregate index.

Deflated by the United Nations index of prices of manufactured exports of developed countries.

Unlike other prices, oil prices in nominal terms were relatively stable except in 1973–74 and 1979–80, when they increased sharply; interpretation of the instability index should therefore take into account this particular behavior of oil prices.

The increase in the instability of primary commodity prices in 1972–82 reflects to some extent a sharp increase in the fluctuation of the dollar’s exchange rates vis-à-vis other major currencies. The conclusion that primary commodity prices were markedly more unstable in 1972–82 than in 1957–71, however, will be reached regardless of whether commodity prices are measured in dollar terms, SDR terms, or real terms (i.e., deflated by the United Nations index of prices of manufactured exports of developed countries), although the extent of the increase appears to be smaller when prices are measured in SDR terms or real terms.

Prices of commodity groups

Price instability around long-term trends increased for all major commodity groups after 1971. The index of price instability for food exhibited the largest rise (fourfold), and the subgroup most responsible for this sharp rise was cereals, whose instability more than quadrupled. Price instability for beverages and for agricultural raw materials approximately trebled after 1971, with coffee and cotton experiencing the largest increases. Metal price instability rose by only 30 percent.

Beverage prices were the most unstable in 1957–71 and retained this rank in 1972–82, when beverage price fluctuations were, on average, more than twice as great as the average fluctuations of all commodities. Metal prices, which had been nearly as unstable as beverage prices in 1957–71, were the least unstable of the four commodity groups in 1972–82. Price fluctuations of food and agricultural raw materials, which had been relatively small during 1957–71, both increased sharply after 1971 to occupy the second and third ranks, respectively.

The difference in the relative behavior of metal prices in the two periods is noteworthy. Relative to the other commodity groups, increases in both nominal prices and the degree of instability were significantly smaller after 1971. There were several reasons for these smaller increases. Metal prices’ average growth rate over 1957–71 increased most rapidly, at 2.5 percent, compared with negative or barely positive rates for the prices of other groups. Metal prices rose particularly fast in the early 1960s, encouraging a rapid expansion of capacity during the decade. During the 1960s, output of copper increased by 47 percent, nickel by 93 percent, and aluminum by 127 percent. When rates of metal consumption growth in the 1970s were lower than had been expected, excess capacity developed that limited price increases and contained price instability of metals relative to the other commodity groups that were more subject to supply constraints.

In summary, commodity price instability has been much more pronounced since the early 1970s than it was in the rest of the postwar period. This result holds for virtually all non-oil primary commodities, irrespective of whether the instability is measured around a long-term or medium-term trend, whether prices are denominated in dollars or SDRs, or whether prices are expressed in nominal or real terms.

II. Determinants of Primary Commodity Prices

The decline in non-oil primary commodity prices in 1981–82 reflected both cyclical and trend factors. These factors are analyzed in this section by examining the statistical relationships since 1957 between commodity prices and their (largely demand-side) determinants. The tentative nature of this analysis should be acknowledged, since commodity markets are influenced by a wide range of factors, many of which are not easily quantifiable. 9

price equation

A commodity price equation is derived and estimated for major groups of commodities for various time periods. The main determinants of commodity prices are then analyzed, with particular regard to their relative impacts during the 1975 and 1981–82 world recessions.

In the price equation to be estimated, the aggregate index of the prices of a commodity group is explained by a number of independent variables. In the logarithmic first-difference form, the equation is written as

where the θis (i = 1, 2, …, 5) denote lag polynomials of finite orders, with the variables defined in logarithms as follows:

p=non-oil commodity prices in dollars
y=economic activity
pd=domestic prices of substitutes in consuming countries
ed=exchange rates of the dollar vis-à-vis the currencies of consuming countries
i=interest rates
s=supply shocks
Δ=change

The hypotheses underlying the relationships of the above independent variables to commodity prices are as follows. Economic activity is the major demand-side variable that positively affects commodity prices through changes in real income or industrial demand in consuming countries. Inflation in importing countries, unless offset by exchange rate movements, positively influences commodity prices by raising the domestic price of substitutes and perhaps by increasing investor demand as a consequence of inflationary expectations. The movements of the exchange rates of the currencies of importing countries vis-à-vis the dollar affect commodity prices expressed in dollars because changes in relative prices (i.e., commodity prices in domestic currency relative to the prices of substitutes) affect the quantity demanded. Changes in the real rate of interest influence commodity prices by affecting the demand for stocks. Finally, supply shocks can have a significant impact on commodity prices, particularly prices of food and beverages, for which demand is relatively stable but annual supplies are unstable.

Equation (1) can be regarded as the transfer function for the price variable derived by eliminating all endogenous variables but the commodity price variable from a largely demand-oriented system of structural equations describing a competitive world commodity market. A model of the commodity market that would suggest the type of equation presented above is shown in Appendix I of this paper.

In estimating equation (1), this study determines empirically the exact orders of the lag polynomials on the basis of quarterly time series described in the next subsection; the orders of the lag polynomials having thus been determined, the equation is estimated both with and without an equality constraint for the coefficients of Δpdt and – Δedt.10

description of data

The variables in equation (1) are defined in logarithms as follows:

pt= aggregate index of non-oil primary commodity prices. All four commodity groups are included: food, beverages, agricultural raw materials, and metals. Equations for each of these commodity groups are estimated separately, and an equation is estimated for all four commodity groups combined.
yt= economic activity in industrial countries. Indices of industrial production for seven industrial countries (the United States, Japan, Canada, and four European countries: the Federal Republic of Germany, France, Italy, and the United Kingdom) are aggregated with their 1975 gross national products (GNPs) as weights. Real GNP was also tried as a proxy for economic activity, but the results obtained appeared inferior to those obtained using industrial production.
pdt= domestic prices. Ideally, the prices of substitutes should be used, but because of the paucity of information, the wholesale price indices (WPIs) of consuming countries are used as a proxy. The variable is defined in terms of domestic currencies. The same seven countries listed above for yt are used with the same weights.
edt= exchange rate. Weighted average of the exchange rates of the seven countries vis-à-vis the dollar. The weights are the same as those used for yt.
it= world interest rate. Proxied by the London three-month Eurodollar rate. This represents primarily the cost of holding stocks. As argued in Appendix I, the price equation has the change in the interest rate as an explanatory variable; the equation for the rate of change in price (Δpt) therefore has the second difference of the rate of interest (Δ2it) as an explanatory variable.
st= supply shocks. By examining the series of production of primary commodities and prices, the quarters in which supply shocks were unusually large were identified for food and beverages; for agricultural raw materials and metals, no such quarters were identified. Dummy variables were created for the food and beverage groups.
behavior of variables underlying commodity price movements

Important changes in the environment of commodity trade are summarized in Table 4 and illustrated in Chart 3. The first part of the table shows sharply higher average nominal commodity prices (122, with 1975 = 100) for 1972–82 relative to prices (53, with 1975 = 100) for 1957–71, but the gain for 1972–82 is more than offset by the inflation indicated by the rise in the general price level in industrial countries. Although the rates of increase in economic activity slowed down, the average rates of increase in non-oil primary commodity prices were sharply higher during 1972–82 than during 1957–71, partly as a result of sharply higher world inflation. The average annual rate of increase in industrial production declined from 5.9 percent during 1957–71 to 2.3 percent during 1972–82; but the annual increase in the average WPI for industrial countries rose from 1.4 percent during the former period to 9.5 percent during the latter, and the average annual increase in commodity prices rose from 0.4 percent to 6.5 percent.

Table 4.Behavior of Variables Affecting Commodity Prices, 1957–82
Average LevelAverage

Change Per Annum
Instability Around

Long-Term Trend1
1957–821957–711972–821957–821957–711972–821957–821957–711972–82
Primary commodity prices
All82531225.10.46.521520
Food6340975.41.04.623624
Beverages110631786.1–0.611.1361136
Agricultural raw materials92631324.2–1.65.825618
Metals87631224.62.56.6161016
Variables affecting commodity prices
Economic activity
Gross national product84661094.25.02.84.21.42.1
Industrial production87681144.65.92.37.12.64.3
Manufacturing output85661114.55.72.66.63.33.8
Capacity utilization952972932–0.2–0.4–1.04.02.63.6
Inflation (domestic currency)
Wholesale price index81571165.01.49.515.02.34.2
Consumer price index81551195.53.09.411.32.21.8
Oil price (in dollars)721614813.3–0.328.355.86.731.9
Exchange rate of U.S. dollar against:
Basket of 7 currencies106110101–0.50.10.24.20.94.8
SDR113121102–1.0–0.0–0.45.60.25.7
Interest rate
Eurodollar rate
(nominal)102741425.77.58.525.017.230.4

The standard error of estimate of the semi-log regression of price (quarterly data) on time, multiplied by 100.

Level of capacity utilization—not index (1975 = 100).

The standard error of estimate of the semi-log regression of price (quarterly data) on time, multiplied by 100.

Level of capacity utilization—not index (1975 = 100).

Chart 3.Non-Oil Commodity Prices and Major Determinants: Percentage Deviations from 19-Quarter Moving Average, 1959–83

(1975 = 100)

1 A decline indicates an appreciation of the dollar.

The data in the table also help to explain the sharp increase in the instability in commodity prices during 1972–82 compared with 1957–71. The instability of all major variables (economic activity, world inflation, exchange rates, and interest rates) affecting commodity prices was higher during 1972–82 than during 1957–71. See Appendix II for a summary of significant changes in major determinants of commodity prices during the sample period.

estimation results

Equations are estimated for each group of commodities (all commodities, food, beverages, agricultural raw materials, and metals) for three sample periods: from the first quarter of 1958 through the second quarter of 1982, the first quarter of 1958 through the fourth quarter of 1971, and the first quarter of 1972 through the second quarter of 1982. The end of 1971 is used as the point for dividing the whole sample period into two subperiods, because the fluctuation in commodity prices has become substantially greater since 1971.11 A number of important events took place around that time: (i) major currencies began to float in the last quarter of 1971 following the breakdown of the system of fixed exchange rates between major currencies; (ii) oil prices became more unstable beginning in 1972, a development that has had far-reaching effects on the world economy; and (iii) world inflation and economic activity have become sharply more unstable since the early 1970s.

Table 5 summarizes the results of estimation of the price equation in first-difference form for groups of commodities. Weak variables are suppressed, and the estimation is carried out with and without the equality constraint for the variables Δpdt and –Δedt.

Table 5.Price Equations I, 1958–8211
Inflation
Wholesale

price
Exchange

rate
Lagged

Price
Economic

Activity
Δpdt– Δedt
Sample

Period
ConstantΔpt-1ΔytΔyt-1ΔpdtΔedt
All Commodities
1958–82–0.028**0.979**0.680**1.412**
(–5.07)(3.66)(2.64)(6.37)
–0.032**1.005**0.655*1.794**–0.992**
(–5.63)(3.83)(2.59)(6.46)(–3.43)
1958–71–0.013*0.560*0.1061.054*
(–2.45)(2.31)(0.46)(2.17)
1972–82–0.030*1.179*1.044*1.385**
(–2.62)(2.28)(2.05)(3.83)
Food
1958–82–0.024**0.900*1.838**
(–2.85)(2.47)(5.15)
–0.030**–0.873*2.388**–1.230*
(–3.40)(2.43)(5.29)(–2.61)
1958–71–0.007*–0.3810.223
(0.99)(–1.32)(0.33)
1972–82–0.047*1.511*2.236**
(–2.57)(2.33)(3.86)
Beverages
1958–82–0.0110.229*1.115*0.255
(–0.94)(2.16)(2.25)(0.54)
–0.0180.2101.107*0.8430.414
(–1.46)(1.98)(2.25)(1.39)(0.65)
1958–71–0.030.188–0.0620.249
(–0.29)(1.31)(–0.13)(0.24)
1972–820.0050.1592.271*–0.232
(0.19)(0.95)(2.42)(-0.29)
Agricultural Raw Materials
1958–82–0.020**0.299**1.490**0.858**
(–2.81)(3.17)(4.65)(2.88)
–0.018*0.307**1.471**0.714–1.005**
(–2.32)(3.22)(4.55)(1.90)(–2.65)
1958–71–0.020**0.2101.305**1.102
(–2.93)(1.80)(3.17)(1.64)
1972–82–0.0110.2322.155**0.645
(–0.79)(1.44)(3.40)(1.48)
Metals
1958–82–0.028**1.533**1.797**
(–3.20)(3.97)(5.04)
–0.027**1.531**1.751**–1.852**
(–2.88)(3.94)(3.84)(–3.78)
1958–71–0.0180.9751.803
(–1.40)(1.60)(1.43)
1972–82–0.037*1.806**1.933**
(–2.46)(3.33)(4.18)

The independent variable in all equations is Δpt. The level of statistical significance is indicated by one (95 percent) or two (99 percent) asterisks. D–W denotes the Durbin-Watson statistic; SEE denotes the standard error of estimate; and the numbers in parentheses below the coefficients are t-statistics.

Supply Shocks
Sample

Period
Interest

Rate

Δ2t-1
Food

sft
Beverages

sbt
R2D–WSEE
All Commodities
1958–82–0.003–0.028–0.068*0.4731.720.038
(–1.56)(–1.37)(–2.51)
–0.004–0.023–0.067*0.4951.660.037
(–1.76)(–1.11)(–2.51)
1958–71–0.0080.1431.870.023
(–1.91)
1972–82–0.004–0.053–0.0650.5221.740.050
(–1.18)(–1.16)(–1.76)
Food
1958–82–0.0530.2802.090.062
(–1.63)
–0.0460.3011.900.061
(–1.41)
1958–711.890.033
1972–82–0.0620.3772.320.082
(–1.41)
Beverages
1958–82–0.134*0.1711.740.083
(–2.00)
–0.138*0.1831.750.083
(–2.07)
1958–71
1972–82–0.1360.2031.630.114
(–1.44)
Agricultural Raw Materials
1958–82–0.007*0.4161.820.047
(–2.54)
–0.007*0.4121.860.047
(–2.50)
1958–71–0.016*0.2331.850.032
(–2.54)
1972–82–0.0070.4611.670.060
(–1.93)
Metals
1958–82–0.007*0.2761.840.063
(–2.10)
–0.007*0.2691.830.063
(–2.07)
1958–71–0.0020.0301.860.062
(–0.16)
1972–82–0.008*0.4131.840.067
(–2.05)

The independent variable in all equations is Δpt. The level of statistical significance is indicated by one (95 percent) or two (99 percent) asterisks. D–W denotes the Durbin-Watson statistic; SEE denotes the standard error of estimate; and the numbers in parentheses below the coefficients are t-statistics.

The independent variable in all equations is Δpt. The level of statistical significance is indicated by one (95 percent) or two (99 percent) asterisks. D–W denotes the Durbin-Watson statistic; SEE denotes the standard error of estimate; and the numbers in parentheses below the coefficients are t-statistics.

Explanatory power of equations

The results are more satisfactory for 1972–82 than for 1957–71. The adjusted coefficients of determination are higher for 1972–82 than for 1957–71 for all commodity groups. The variables included in the equation explain a substantial proportion of the variations in the rates of change in prices for all commodity groups except beverages. The adjusted coefficients of determination for the whole sample period (1957–82) range from 0.171 for beverages to 0.473 for all commodities with the equality constraint for the coefficients for Δpdt and – Δedt; they range from 0.183 for beverages to 0.495 for all commodities without the constraint. These coefficients of determination are somewhat higher than those obtained in a previous study that used a similar dependent variable but excluded the interest rate and exchange rate as independent variables.12 For 1972–82, the coefficients obtained in the present study range from 0.203 for beverages to 0.522 for all commodities.13

The results are less satisfactory for 1957–71 than for 1972–82, probably because the fluctuations of the demand-side variables were substantially less during the former period than the latter, implying that, in relative terms, supply shocks were more important in determining the fluctuations of commodity prices during the former period than the latter.14 The results suggest that this was probably the case, particularly for the food and the beverage groups.

Effects of major explanatory variables on commodity prices

The model confirms the influences of the level of economic activity, world inflation, and exchange rates on commodity prices. For 1958–82, these variables are highly significant for all groups of commodities. The only exception is the coefficient of the inflation variable, which is not significant for beverage prices; this result may be due to the dominance of supply shocks in the fluctuation of beverage prices.

The elasticity of overall commodity prices with respect to industrial production is estimated at about 2 for 1972–82, higher than the estimated elasticity for 1957–71. As might be expected, the coefficients of the industrial production variables are larger for agricultural raw materials and metals than for food and beverages.

The elasticity of overall commodity prices with respect to inflation is somewhat larger than unity when the absolute values of the coefficients for the inflation (measured in domestic currencies) and exchange rate change variables are constrained to be equal.15 The coefficients vary substantially among commodity groups, ranging from not statistically different from 0 to close to 2.

The role of the dollar exchange rate in the determination of commodity prices expressed in dollars, as indicated in the results of the regressions, is not surprising. The results show that commodity prices have been sensitive to exchange rate changes during 1972–82. The elasticity of overall commodity prices with respect to the U.S. dollar exchange rate vis-à-vis major currencies is estimated to be somewhat greater than unity. The coefficients vary among commodity groups and are significant for all groups except beverages.

The effects of interest rate changes on commodity prices are not sufficiently robust; the results reported in the table are based on the interest rate variable lagged one quarter as an explanatory variable and suggest an inverse relationship, as indicated in the model introduced earlier. However, if the same equations are estimated with the current interest rate variable as an explanatory variable, the estimated coefficients become either positive or weakly negative. The results with a one-quarter lag are strong for agricultural raw materials and metals.16 Notwithstanding these rather inconclusive results, the unusually high real interest rates during 1981–82 may have contributed to the downward pressure on commodity prices by encouraging inventory reductions, just as the predominantly negative real rates of interest during the 1970s may have exerted upward pressure on nominal commodity prices.

The estimated coefficients for supply shocks all have the expected signs; however, the supply shocks, which are quantified as dummy variables that distinguish only those quarters when conspicuous effects of supply shocks occurred, are not dealt with adequately in the model.

The strong negative constant term estimated for all groups of commodities except beverages is also noteworthy. It reflects the effects of the secular drifts of the supply and the demand functions as specified in Appendix I. For example, the supply function specified in equation (5’) does not include variables accounting for long-term expansion of production capacity of primary commodities and the innovations in production technologies. Also, the demand function specified in equation (4) does not include variables accounting for secular drifts in the demand function that could have occurred because of long-term growth in production of synthetic substitutes. The estimated constant term measures the net effects of the secular trends of the omitted variables on commodity prices. In other words, the estimated coefficient suggests that if the expansion in production capacity, innovations in production technologies, and other secular factors occurred as they actually did, while the variables—such as economic activity and relative prices—that are included in the equations did not change, then commodity prices would have declined by more than 2 percent a quarter.17

Other results

The dynamic nature of the equations should be noted: in the equation for all commodities, both the coefficients for the current and lagged economic activity variables are significant; in the food and the beverage equations, only the lagged economic activity variable is significant; and in the beverage and the agricultural raw material equations, the lagged dependent variables are significant for some periods.18 There is some indication that for beverages and agricultural raw materials, the responses of commodity prices to changes in the explanatory variables are not instantaneous.

The results indicate that the values of some coefficients changed between the 1958–71 and the 1972–82 sample periods: in virtually all cases, the coefficients of economic activity variables became larger in absolute terms in the latter period than in the former, implying a greater fluctuation of commodity prices in the latter period in response to variations in economic activity.19 A possible explanation for the greater response of commodity prices to changes in economic activity in the 1970s may be that larger absolute upward fluctuations in economic activity result in capacity constraints being reached and larger downward fluctuations succeed in overcoming a ratchet effect that limits downward price movements in response to smaller declines in economic activity. The increased use of futures markets in the 1970s may also have resulted in greater responsiveness of commodity prices to underlying real and monetary variables.20

Possible shifts of the coefficients between the two subperiods (1958–71 and 1972–82) have been tested separately for each coefficient, with the maintained hypothesis that all other coefficients have remained the same during the entire sample period.21 The tests suggest that the coefficient of economic activity shifted after 1971 for all commodities, beverages, and agricultural raw materials. No strong evidence is obtained for any other coefficient (see Table 6).

Table 6.Price Equations II, 1958–821
Explanatory Variables
Economic Activity
ConstantLagged

Price

Δpt-1
ΔytΔyt-1Δyt. dtΔyt-1. Δdt-1
All Commodities
–0.022**0.896**0.1741.035*
(–3.75)(3.41)(0.53)(2.45)
Food
–0.024**0.900*
(–2.85)(2.47)
Beverages
–0.0010.168–0.1522.414**
(0.08)(1.58)(–0.22)(2.58)
Agricultural Raw Materials
–0.015*0.213*0.861*1.402*
(–2.19)(2.19)(2.19)(2.61)
Metals
–0.028**1.533**
(–3.20)(3.97)

The dependent variable in all equations is Δpt. The level of statistical significance is indicated by one (95 percent) or two (99 percent) asterisks. The variables dt and dt-1 are dummy variables that are assigned values of either 0 (for 1958–71) or 1 (for 1972–82). D-W denotes the Durbin-Watson statistic, and SEE denotes the standard error of estimate.

Explanatory Variables
InterestSupply Shocks
InflationRateFoodBeverages
Δpdt – ΔetΔ2it–1sftsbtR2D–WSEE
All Commodities
1.214**–0.004–0.026–0.061*0.5011.770.037
(5.27)(1.95)(–1.32)(–2.30)
Food
1.838**–0.0530.2802.090.062
(5.15)(–1.63)
Beverages
–0.134–0.1350.2191.670.081
(–0.28)(–2.07)
Agricultural Raw Materials
0.793**–0.008**0.4501.750.045
(2.73)(–2.95)
Metals
1.797**–0.007*0.2761.840.063
(5.04)(–2.10)

The dependent variable in all equations is Δpt. The level of statistical significance is indicated by one (95 percent) or two (99 percent) asterisks. The variables dt and dt-1 are dummy variables that are assigned values of either 0 (for 1958–71) or 1 (for 1972–82). D-W denotes the Durbin-Watson statistic, and SEE denotes the standard error of estimate.

The dependent variable in all equations is Δpt. The level of statistical significance is indicated by one (95 percent) or two (99 percent) asterisks. The variables dt and dt-1 are dummy variables that are assigned values of either 0 (for 1958–71) or 1 (for 1972–82). D-W denotes the Durbin-Watson statistic, and SEE denotes the standard error of estimate.

comparison with other studies

Other studies have developed and tested various models of commodity price determination. Like the present study, all of these studies analyze prices of groups of commodities except for the two studies by Hwa (1979 and 1981), which analyze six individual commodities. Unlike the present study, however, most of these studies use the relative price of commodities vis-à-vis manufactures as the dependent variable, except for Hwa (1979 and 1981) and Enoch and Panic (1981), who use changes in the nominal price of commodities. The positive relationship between world economic activity and non-oil primary commodity prices has been well established; all of these studies obtained significant results for this relationship. The price of manufactured exports was shown to be positively related to commodity prices in one study (Enoch and Panic (1981)), and the price of oil and oil price shock dummy variables had significantly positive coefficients in other studies (Enoch and Panic (1981) and Grilli and Yang (1981)). In all these studies, however, the exchange rate variable was not tested separately. Inflationary expectations were shown to be positively related to the prices of several individual commodities (Hwa (1979)). One study also found that the level of the interest rate—not the change in the interest rate that is used in the present study—is negatively related to commodity price movements, especially for metals and agricultural raw materials (Grilli and Yang (1981)), but one found insignificant results (Enoch and Panic (1981)). Supply-side variables (such as production of a commodity or group of commodities, and stocks) that were introduced in several studies showed the expected inverse relationship to commodity prices (Bosworth and Lawrence (1982), Cooper and Lawrence (1975), Hwa (1979 and 1981)). Finally, one study found that commodity prices are positively related to exchange rate variability (Grilli and Yang (1981)). The present study tests the level of the exchange rate, not its variability, as a determinant of commodity prices.

factors underlying instability of commodity prices

Table 7 illustrates how the historical movements of the major explanatory variables identified in this study can help trace the short-term fluctuations in commodity prices. 22 The variance of the rate of change in overall commodity prices is estimated at 6.4 for 1958–71; during this period, the variances of the three demand-side variables—industrial production, world inflation, and the exchange rate—are estimated, respectively, at 2.0, 0.2, and 0.2.23 Together with the estimates of the coefficients for these variables presented in Table 6, these estimates of the variances suggest that contributions made by the variations in these three variables to the variation in commodity prices are 1.6, 0.3, and 0.3, respectively, and that about a third of the variations in commodity prices during 1957–71 is accounted for by the variations in the three critical demand-side variables.24 Although the extent to which the variation in the rate of change in commodity prices was explained by the variations in the three variables remained the same (at about a third) during 1972–82, the degree of variation in these variables was sharply higher in 1972–82 than in 1958–71 (see Table 7). The variance of the rate of change in commodity prices was about nine times greater during 1972–82 than during 1958–71. The variances for all the demand-side variables were greater in the former period than the latter, but the variance for industrial production was twice as large, while the variance for the inflation rate was more than 12 times larger; the instability was greater for both wholesale prices and the exchange rate, but it was much greater for the latter. At the same time, the variance of the random disturbance term, reflecting, among other factors, production shocks that were not exceptional during 1957–71, also became much larger during 1972–82, though its relative contribution to commodity price variation declined substantially.

Table 7.Factors Underlying Fluctuation of Commodity Prices, 1958–82
1958–711972–82
(1) Variation in rate of change16.452.9
Explained by variations in:
(2) Three major variables: (a) + (b) + (c)22.217.2
(a) Industrial production1.67.9
(b) Wholesale price0.33.5
(c) Exchange rate0.35.8
(3) Error term36.326.6
(4) Residual: (1) - (2) - (3)4–2.19.1

Variance of the quarterly rate of change in percent.

Sum of the variances of the three variables multiplied by the respective coefficients from Table 6 squared.

The error term reflects the significant variables left out of the equation, including production shocks that were not exceptional.

The residual could be either positive or negative, because it reflects covariances between the explanatory variables included in the equation.

Variance of the quarterly rate of change in percent.

Sum of the variances of the three variables multiplied by the respective coefficients from Table 6 squared.

The error term reflects the significant variables left out of the equation, including production shocks that were not exceptional.

The residual could be either positive or negative, because it reflects covariances between the explanatory variables included in the equation.

Table 8.Indices of Commodity Price Instability and Underlying Factors
Instability Around

Long-Term Trend1
Variance of

Rate of Change2
1958–711972–82Ratio1958–711972–82Ratio
(1)(2)(3) = (2)(1)(4)(5)(6) = (5)(4)
Prices (in dollars)
All commodities25400166539
Food36576161110910
Beverages1211,29611221647
Agricultural raw materials36324914675
Metals100256338762
Underlying factors
Industrial production6.718.532.04.02
Inflation adjusted for exchange rate changes4.135.190.45.012
Wholesale price5.417.530.22.412
Exchange rate0.816.8210.23.920

For commodity prices, the instability indices are the standard errors reported in Table 3 (column (B) under the appropriate periods) squared; for underlying factors, the indices are derived in the same manner on the basis of the semi-log regression of a variable (quarterly) on time.

Variances of the quarterly rates of change (in percent).

For commodity prices, the instability indices are the standard errors reported in Table 3 (column (B) under the appropriate periods) squared; for underlying factors, the indices are derived in the same manner on the basis of the semi-log regression of a variable (quarterly) on time.

Variances of the quarterly rates of change (in percent).

The fact that the elasticity estimate of the exchange rate variable reported earlier is not far from unity suggests that the increase in instability of the dollar exchange rates vis-à-vis other major currencies was probably not a major cause of the increase in instability of commodity prices in SDRs, as reported in the subsection entitled “price instability” in Section I of this paper.25

The long-term upward trend of nominal commodity prices has been sharply steeper since the early 1970s, reflecting higher world inflation, while the longer-term downward trend of real commodity prices has also been steeper, reflecting the higher and sustained inflation of manufactures prices, as well as their downward rigidity. The variation in commodity prices has been sharply greater since the early 1970s as a result of greater instability in the economic environment. The preceding analysis has focused largely on the characteristics of commodity price behavior during 1972–82, compared with 1957–71. The analysis therefore has not focused specifically on price developments in very recent years. An examination of the price declines in 1975 and 1981–82 and the price increases preceding the declines will help illustrate how changes in the underlying and following factors studied in this paper influence commodity prices.

comparison of 1975 and 1981–82 recessions

In this section, the commodity price cycle of 1978–83 is analyzed and compared with the previous major price cycle of 1972–77. The commodity price declines of 1975 and 1981–82 were both preceded by relatively large commodity price increases, but the increase preceding 1975 was by far the larger. Although both increases occurred over seven quarters, the cumulative increase preceding the 1975 decline was 131 percent, compared with a 29 percent increase preceding the 1981–82 decline. Movements of the explanatory variables are consistent with the greater strength of the commodity price boom before 1975. Industrial production rose by a cumulative 19 percent in 1972–74, compared with 11 percent for 1979–80; and inflation by a cumulative 40 percent in the former period, compared with 30 percent in the latter. The dollar depreciated by a cumulative 10 percent prior to 1975, compared with no change prior to 1981–82; and the price of oil increased by 382 percent in the former period, compared with 149 percent in the latter (Table 9). The real interest rate (Eurodollar rate) decreased by 10 percentage points prior to 1975, compared with an increase of 11 points in 1979–80, an observation that also is consistent with a stronger commodity price boom before 1975.

Table 9.Commodity Prices and Their Major Determinants: Changes During the 1972–77 and 1978–83 Cycles I(In percent)
1972–77 Cycle1978–83 Cycle
Price increaseDecreaseRecoveryPrice increaseDecreaseRecovery1
Period21972 I–74 II1974 III–75 IV1976 I–77 II1978 IV–80 IV1981 I–82 IV1983 I–83 IV
Length (in quarters)1066984
Commodity Prices
Cumulative
All commodities: nominal prices1312458292816
Food161–26–269–4423
Beverages771246–17–016
Agricultural raw materials130–243225–2821
Metals129–361735–211
All commodities: real prices357314231814
prices in SDRs932260261418
Annual rate
All commodities401736121516
Food47–18–126–2523
Beverages251129–8–016
Agricultural raw materials40–162110–1521
Metals39–261114–111
All commodities: real prices32022261914
prices in SDRs30153711718
Major Determinants
Cumulative
Industrial production1912141199
United States20–13169–1115
Europe413–91110–122
Japan28–17152338
Wholesale price index40101230123
United States351693282
Europe44462028197
Japan5355283–3
Exchange rate10220204
Europe4–1729–44912
Japan–228–9923–10
Oil price3821261495–15
Annual rate
Industrial production789549
United States8–9104–415
Europe45–674–52
Japan10–12101018
Wholesale price index14781262
United States131061342
Europe4164131297
Japan1833122–3
Exchange rate3210104
Europe4–526–22212
Japan–76–6411–10
Oil price8884683–15

The analysis of the recovery phase is confined to 1983 because the latest data were available only through the fourth quarter of 1983 at the time of writing.

The periods for the commodity price indices. The Roman numerals following the years denote calendar quarters. For the explanatory variables, the period chosen for various phases of the two cycles are the same as for the commodity prices except for the following: (a) Industrial production, (i) 1975 recession: increase (1971 III-1973 IV); decrease (1974 1-1975 II); and recovery (1975 III-1976 IV). (ii) The 1981–82 recession: increase (1978 1-1980 I); decrease (1980 11–1982 IV). (b) Exchange rate. 1975 recession: increase (1971 11–1974 II).

Deflated by the United Nations index of prices of manufactured exports of developed countries.

Includes the United Kingdom, the Federal Republic of Germany, France, and Italy.

The analysis of the recovery phase is confined to 1983 because the latest data were available only through the fourth quarter of 1983 at the time of writing.

The periods for the commodity price indices. The Roman numerals following the years denote calendar quarters. For the explanatory variables, the period chosen for various phases of the two cycles are the same as for the commodity prices except for the following: (a) Industrial production, (i) 1975 recession: increase (1971 III-1973 IV); decrease (1974 1-1975 II); and recovery (1975 III-1976 IV). (ii) The 1981–82 recession: increase (1978 1-1980 I); decrease (1980 11–1982 IV). (b) Exchange rate. 1975 recession: increase (1971 11–1974 II).

Deflated by the United Nations index of prices of manufactured exports of developed countries.

Includes the United Kingdom, the Federal Republic of Germany, France, and Italy.

The period of commodity price decline was six quarters for 1975 and eight quarters for 1981–82. The cumulative decline was 24 percent for the former period and 28 percent for the latter. Industrial production declined by a cumulative 12 percent during the 1975 recession and 9 percent during 1981–82, indicating that economic activity was a more important factor in the 1975 decline. The wholesale price index increased by 10 percent in the earlier period, compared with 12 percent in the latter period. The dollar appreciated by 2 percent during the 1975 recession, while it appreciated by 20 percent during 1981–82, indicating a significantly greater impact on commodity prices expressed in dollars in the latter period. The real interest rate increased by 9 percentage points during the 1975 recession, while it has decreased by 5 percentage points during 1981–82. Real interest rates, however, were significantly higher in 1981–82 than in 1975, when they were actually negative.

The large role played by the appreciation of dollar exchange rates in the decline of dollar prices is reflected in the relatively smaller decline in commodity prices in SDR terms, compared with U.S. dollar terms, during 1981–82. The SDR prices declined by 14 percent in 1981–82, compared with a 28 percent decline in dollar prices (see Table 9). Real commodity prices declined less during 1981–82 (18 percent) than in 1975 (31 percent). It is, however, true that the 1981–82 decline brought real prices to a postwar low.

The commodity price recovery in 1976–77 lasted six quarters, during which prices increased by 58 percent. During this same period, industrial production increased by 14 percent, and the inflation rate was 12 percent. The dollar appreciated by 2 percent, and the real interest rate declined by 13 percentage points. The strong recovery in economic activity was accommodated by a rather expansionary monetary policy in the industrial countries.

In Table 10, major demand-side factors that have been identified as important determinants of commodity prices are considered in “simulations” of the changes in the aggregate index of commodity prices during the various phases of the two commodity price cycles using the price equation in Table 6.

Table 10Commodity Prices and Their Major Determinants: Changes During the 1972–77 and 1978–83 Cycles II
Demand-Side VariablesSupply Shocks
Percentage

change in

commodity prices
Contributions made

by major determinants
Excess or shortfall (-) in annual

percentage change in production1
Economic activityExchange rateAgricultural raw materials
ActualSimulated2InflationTrendFoodBeveragesMetal
1972–77 Cycle
Price increase13168304812–22–21–714
Decrease–24–26–2512–2–134–4–4–5
Recovery58282914–2–133–21–12
1978–83 Cycle
Price increase293923360–20–3–21–1
Decrease–28–44–1917–24–183132
Recovery169194–5–93

Actual rate of change in production above or below the long-term average rate of change in production. For the cycle around the 1975 recession, the periods used for the production figures are as follows: (i) 1972–73 for price increase; (ii) 1975 for price decrease; and (iii) 1976 for recovery. For the cycle around the 1981–82 recession, the periods are as follows: (i) 1979–80 for price increase; and (ii) 1981 for price decrease.

Based on the historical movements of the explanatory variables in Table 9 and the estimated coefficients summarized in Table 6. Simulated change in price is the sum of the contribution made by major factors. The contributions made by major factors are calculated as the product of the change in the variable and the associated coefficient.

Although complete data are not available, there is a strong indication that the recovery of agricultural products prices—including some foods, beverages, and agricultural raw materials—was facilitated by worldwide production shortfalls owing to adverse weather.

Actual rate of change in production above or below the long-term average rate of change in production. For the cycle around the 1975 recession, the periods used for the production figures are as follows: (i) 1972–73 for price increase; (ii) 1975 for price decrease; and (iii) 1976 for recovery. For the cycle around the 1981–82 recession, the periods are as follows: (i) 1979–80 for price increase; and (ii) 1981 for price decrease.

Based on the historical movements of the explanatory variables in Table 9 and the estimated coefficients summarized in Table 6. Simulated change in price is the sum of the contribution made by major factors. The contributions made by major factors are calculated as the product of the change in the variable and the associated coefficient.

Although complete data are not available, there is a strong indication that the recovery of agricultural products prices—including some foods, beverages, and agricultural raw materials—was facilitated by worldwide production shortfalls owing to adverse weather.

For the 1972–77 price cycle, the demand-side variables produce a path of commodity prices roughly similar to the historical path. The model based on the demand-side variables, however, substantially underpredicts the price increases during the period before the recession and during the recovery. This probably results from the fact that, as indicated in the right half of the table, supply shocks for beverages substantially reinforced the demand-side variables during these two phases of the cycle. During the entire cycle, economic activity and inflation dominated the exchange rate as causes of price movements; however, inflation was the more important determinant of price increases before the recession, while economic activity was the more important determinant of price increases after the recession. There is an indication that supply shocks for food were also responsible for the price increase before the recession. There are fairly strong indications that supply shocks for food reinforced the demand-side variables during the recession, and that similar shocks for beverages played the same role during the recovery.

For the 1978–83 cycle, the demand-side variables overpredict both the intitial price increase and the subsequent decrease during the recession. The simulated path, however, is fairly close to the actual path of prices. A notable aspect of the results is the role played by the exchange rate in the price decrease. In contrast to the 1975 recession, exchange rate movements seem to have dominated the other two demand-side variables in the 1981–82 decline of commodity prices.

A full comparison between the recovery phases of the two cycles recessions is not feasible at the time of this writing; the recovery of commodity prices that started at the beginning of 1983 is still under way. An analysis of the first four quarters shows that the pace of recovery has been slower than during the 1976 recovery. The rate of increase in the overall commodity price index was 16 percent during 1983, compared with an annualized rate of 34 percent during the 1976–77 recovery. The price recovery in 1983 has been facilitated by production shortfalls of some food and agricultural raw materials, while in 1976, it was facilitated by a production shortfall of beverages. The overall pace of recovery in industrial production in 1983 has been about the same in 1983 as in 1976 entirely because of the strength of the U.S. economic recovery; the recovery has been modest in Japan and Europe. World inflation has been much milder in 1983 than in 1976. The U.S. dollar appreciated more in 1983 than in 1976. The simulation underpredicts the recovery in prices; part of the reason appears to be the reinforcement of supply shocks that facilitated the price recovery in 1983.

III. Conclusions

The study shows that non-oil primary commodity prices underwent a significant change in the 1970s. Both nominal commodity prices and price instability increased markedly. The major variables identified in this study that are shown to have power in explaining this changed behavior are economic activity, world inflation, the dollar exchange rates vis-à-vis other major currencies, and supply shocks. There is also some evidence to support the hypothesis that changes in interest rates inversely affect commodity prices. The limitations of this study should be noted when considering these results. In particular, neither the short-run and long-run price-supply relationships nor the implications of commodity price fluctuations for world inflation have been dealt with in this paper.

During 1981–82, commodity prices declined further (25 percent) and for a longer period (8 quarters) than they have in the last three decades. In 1981, real commodity prices reached a postwar low; and in 1982, they declined a further 11 percent, to a level 16 percent below the trough reached in the 1975 recession. Compared with the commodity price decline during the 1975 recession, the 1981–82 decline appears to have been caused relatively more by exchange rate movements. Economic activity had a strong influence on commodity prices in both recessions, but this influence was somewhat greater in 1975. Interest rates appear to have exerted relatively more influence in 1981–82 than in 1975.

The sharp decline in commodity prices during 1981–82 is shown to be a culmination of a trend toward more unstable prices that began in the early 1970s. The long-term downward trend in real commodity prices from 1972 to 1982 has been more than twice the trend from 1957 to 1971. In addition, primary commodity prices during 1972–82 were more than three times as unstable as they were during 1957–71, while fluctuations in world economic activity, inflation, exchange rates, and interest rates were significantly more pronounced in 1972–82 than in 1957–71.

APPENDICES
I. Derivation of Price Equation
the model

First, a simple model is constructed that incorporates the fundamental features of a competitive international commodity market in which the commodity price is quoted in dollars; the model is then applied to broadly aggregated commodities. The model consists of demand and supply functions.

Demand

Demand for the commodity is specified as

where

QtC = quantity demanded

Pt = price of commodity (in dollars)

PDt = average of prices of substitutes in consuming countries (in domestic currency)

EDt = average of exchange rates between the dollar and the currencies of consuming countries (national currencies/dollar)

Yt = level of economic activity in consuming countries

A0, α1, α2, α3 = parameters

Equation (2) is a typical import demand function for a commodity that has substitutes. The meaning of PDt or Yt would depend on the nature of the commodity. For, say, natural rubber, PDt could denote the price of synthetic rubber, while for beef, it could denote the prices of other meats. For meats as a group, PDt could denote the prices of other food substitutes that contain similar nutritional elements. It should be noted that PDt is measured in the domestic currency of the consuming country. The meaning of Yt could be disposable income for food or beverages, but it could also be automobile production for natural rubber.

If commodities are storable, the transactional, or nonspeculative, demand for stocks of the commodity may be specified as26

where

StD = transactional, or nonspeculative, demand for stocks of the commodity

rt = real rate of interest

QtD = total flow demand for the commodity (including both utilization and increase in stocks)

In equation (3), the demand for stocks for transactional purposes is a positive function of total flow demand for the commodity and a negative function of the real interest rate; this specification is based on the recognition that consumers of commodities would attempt to reduce inventories as their holding cost rises. The equation may be written in first-difference form as follows:

where Δ denotes “change”—for example, ΔStD=StDSt1D. Equation (3’) suggests that the flow demand for stocks for transactional purposes would depend on the change in the real rate of interest and the change in the total flow demand for the commodity.

In principle, a price equation could be derived by equating total flow demand to supply; in practice, the nonlinearity of the total flow demand equation derived from equations (2) and (3’) would cause a serious problem in reducing the system for derivation of a price equation in relatively simple form. Therefore, an equation is postulated for total flow demand (consumption plus desired increase in transactional stocks) as follows:

where qt = ln Qt. Equation (4) does not straightforwardly combine equations (2) and (3’), but it captures fairly faithfully the essential features of the relationship, specified in those two equations, between the flow demand and the set of explanatory variables. A number of these features may be summarized as follows:

(1) The basic features of the multiplicative consumption function in equation (2) are retained in the new formulation.

(2) Equation (4) is reduced to equation (2) if Δrt=ΔqtD=0; the flow demand for stocks should be zero in such a case.

(3) In equation (3), a negative Δrt and a positive ΔqtD imply an increase in the transactional demand for stocks; equation (4) also captures this relationship.

In equation (4), the coefficients (α1, α2) of PtEDt and PDt are specified to be different; in empirical estimation, the case in which the two coefficients are constrained to be equal will also be examined.

Supply

Supply of the commodity may be specified as

where

Qt = supply (production) of the commodity

PSt = average of production costs in exporting countries

ESt = average of exchange rates between the dollar and currencies of exporting countries (national currencies/dollar)

St = exogenous supply shocks, such as unusual weather and strikes by production workers

B0, β1, β2, β3 = parameters

However, in view of the low short-run price elasticities of supply (β1, β2) of primary commodities, the equation may be approximated by a simpler form.27

Determinants of price

The system consisting of equations (4) and (5’) is closed by the equilibrium condition

and can be solved for the price28

where

In deriving equation (7) for price, the real rate of interest r, is defined as

Equation (7) identifies major variables affecting commodity prices. They include four demand-side variables (economic activity, yt; price of substitutes, pdt; change in nominal interest rate, Δit; and exchange rate, edt) and a supply-side variable (supply shocks, st). The equation is in a dynamic form, suggesting lagged responses of commodity prices to changes in explanatory variables. The equation would have a more complicated dynamic structure if both the supply and demand equations took more proper dynamic forms to enable one to interpret αiS (i = 1,2,…, 5) and β3 as truly short-run elasticities. The equation would have a still more complicated dynamic structure if recognition were given to the fact that the market may not always be in equilibrium.

In spite of all its simplicity, equation (7) includes the important variables that affect commodity prices and the parameters that determine the extent to which those variables affect commodity prices.

In analyzing the channels through which the major variables included in equation (7) affect commodity prices, it is extremely important to distinguish between autonomous and induced changes in these variables. For example, a depreciation of the consuming countries’ exchange rates vis-à-vis the dollar caused by an acceleration of the same magnitude in those countries’ domestic inflation may not affect commodity prices. Similarly, if worldwide inflation accompanied by restrictive monetary policies induced an accelerated increase in the nominal interest rate, the positive effects of the inflation on commodity prices could be more than offset by the negative effects of interest rate movements and their possible dampening effects on economic activity.

limitations of model

The following limitations of the model should be noted:

(1) The model does not capture the role of speculative demand in the determination of commodity prices; anticipation does not play any role in the model. Accordingly, the model does not fully recognize primary commodities as assets. In view of the growth of futures markets for a number of commodities in recent years and also of the importance of some primary commodities as assets, a comprehensive model should incorporate the role of anticipation and the rates of return on commodities relative to those on other assets (i.e., major currencies and financial assets). The interest rate variable included in the equation, however, may reflect, in an imperfect way, the speculative demand for commodities, since it partially represents the opportunity cost of holding commodities.29

(2) The model is essentially a short-run model, largely demand oriented. Therefore, it does not capture the longer-term dynamic interactions between price and supply. The coefficients in equation (7) should be interpreted as indicators of short-run effects. The long-term expansion of world production of primary commodities is regarded as exogenous, and its effect on commodity prices would be reflected in the constant term to be included in equation (7) in the first-difference form.

(3) The model is based on the assumption that the causation between world inflation and commodity prices is unidirectional, from world inflation to commodity prices. To the extent that causation also runs in the opposite direction, the conclusions would have to be qualified.

II. Behavior of Variables Underlying Commodity Price Movements
economic activity

Industrial production of seven major industrial countries expanded at a rather steady rate from 1957 to 1971 (see Chart 4). Thereafter, the trend became less steep, and fluctuations around the trend increased markedly. The sharp increase in industrial production from 1972 to 1974 and the subsequent decline in 1975 coincided with similar movements in commodity prices over the same period. Industrial production then rose rather steadily until 1980, when it declined by about 1 percent. During this period, commodity prices also rose, except during 1978, when a decline was brought about largely by a sharp drop in beverage prices owing mostly to supply factors. The decline that occurred in 1981 lagged the decline in industrial production by about three quarters.

Chart 4.Europe, Japan, and the United States: Industrial Production, 1957–83

(1975 = 100)

It has been suggested that one of the reasons for the sharp rise and fall of commodity prices during 1972–75 was the unusual degree of synchronization of the business cycles of the major industrial countries30. Compared with previous cycles that were often at least partially offsetting, the synchronized expansion of the industrial production from 1972 to 1974 placed great pressure on available commodity supplies. The subsequent decline in 1975, as shown in Chart 4, was also synchronized, placing equally strong downward pressure on commodity prices. Since 1975, however, the business cycles of the major industrial countries have not maintained this synchronization, although during 1981–82 they declined together (with the exception of Japan).

world inflation

The rate of world inflation increased sharply after 1972 (see Chart 5). The rate of increase in industrial countries’ wholesale prices reached its peak in 1973–74 at an average annual rate of 16 percent, coinciding with the peak in commodity price increases during the same period. The second highest wholesale price inflation occurred in 1979–80 and was also accompanied by relatively large increases in nominal commodity prices. The price of petroleum experienced its greatest surges in these same two periods (see Chart 6).31

Chart 5Europe, Japan, and the United States: Wholesale Price Index, 1957–83

(1975 = 100)

Chart 6Prices of Manufactures and Oil, 1957–831

(1975 = 100)

1 United Nations index of manufactured exports of developed countries and a weighted average of official oil prices of Libya, Saudi Arabia, and Venezuela.

The rapid growth of futures trading has led some to attribute part of the increased commodity price instability in the 1970s to a rise in speculative demand. 32 During the 1970s, both the volume of futures trading and the number of commodities traded increased sharply. Trading volume increased from 13.6 million contracts in 1970 to 92.1 million in 1980, and the number of commodities traded increased from 40 to 82 over the same period.

Although speculative activity indisputably increased during the 1970s, it is still questionable whether this increased speculation was a cause or an effect of increased commodity price instability. A recent study testing the direction of causation has lent support to the hypothesis that commodity price instability led to greater speculation, rather than vice versa.33 An extensive and inconclusive literature exists, however, on whether speculation is stabilizing or destabilizing, a question beyond the scope of this study.

exchange rate

As indicated earlier, the impact of an exchange rate change of importing countries on demand for a commodity depends on the relevant elasticities; an appreciation of the dollar would tend to decrease demand for commodities traded in dollars, whereas a depreciation should tend to increase demand.

Since the end of 1971, exchange rate fluctuations, particularly those of the dollar against other major currencies of importing countries, have been much greater than in the rest of the postwar period. For example, the dollar depreciated about 14 percent against the major currencies during 1971–74 (see Chart 7). It then appreciated somewhat in 1975 before depreciating by about 12 percent during 1976–79. Finally, it appreciated sharply during 1981–82, far exceeding relative price movements in the United States and other importing countries. These movements in the dollar exchange rate since 1971 appear to be inversely related to primary commodity price movements.

Chart 7Europe, Japan, and the United States: Exchange Rates, 1957–831

(1975 = 100)

1 Vis-à-vis the dollar. A decline indicates an appreciation of the dollar.

Exchange rate instability in the 1970s may have influenced speculative demand for primary commodities, with greater instability creating greater demand for primary commodities as hedges against exchange rate risk.34 This impact of exchange rate fluctuations on commodity prices, however, does not necessarily result in an inverse relationship between exchange rates and commodity prices.

interest rate

Interest rates in the industrial countries have experienced much greater fluctuations since the early 1970s than at any other time in the postwar period, and they have also reached postwar highs (in both nominal and real terms) during this period. Interest rates can influence commodity prices on both the demand and supply sides. High interest rates in the 1970s have greatly increased the cost of holding inventories, thereby resulting in a rundown of user stocks of raw materials that has put further downward pressure on demand for current production. Several recent studies have related changes in the real rate of interest to inventory changes and have drawn implications for commodity prices.35

The generally higher real interest rates that prevailed in the 1970s, compared with earlier years, have also resulted in producers holding smaller raw material inventories. For example, because of the increased cost of holding stocks in the 1970s, the major grain exporting countries moved from a policy of holding large stocks to a policy of production adjustments.36 This supply-side effect would not necessarily lead to an inverse relationship between interest rates and commodity prices, but it might well be one of the factors explaining greater commodity price instability in the 1970s. For example, it has been shown that in 1972, immediately preceding the commodity price boom, stocks as a percentage of trend production were lower than their 1965–70 averages across a wide range of storable raw materials.37

supply shocks

Except for the effect of interest rates on production costs, the determinants of commodity price behavior discussed so far have all been demand-related. For individual commodities, however—particularly food and beverages, whose supply often depends as much on the weather as on expected demand—supply-related determinants may be quite important. It has been documented, for example, that in 1973 unusually poor growing conditions in much of the world were a major factor in the subsequent sharp increase in food prices, and that the 1975 frost in Brazil contributed significantly to the subsequent rapid rise in coffee prices.

In explaining movements in overall commodity prices, however, these supply factors are difficult to analyze, as supply-related data are often not available. Moreover, supply factors are usually impossible to forecast, so that their usefulness for more than a very short-term outlook exercise is limited.

III. Commodities and Weights in International Monetary Fund Index of Non-Oil Primary Commodity Prices

(In percent)

CommoditiesWeights Used to

Aggregate Prices1
All100.0
Food31.4100.0
Oils and oilseeds27.1
Groundnut oil7.2
Copra7.1
Groundnut cake4.7
Fish meal4.5
Palm oil2.7
Soybean meal0.5
Soybeans0.4
Cereals25.7
Maize10.1
Wheat8.3
Rice7.3
Sugar21.1
United States10.1
Free market7.5
European Economic Community3.5
Meat18.5
Beef14.1
Lamb4.4
Bananas7.6
Beverages18.2100.0
Coffee67.1
Robusta34.8
Other milds32.3
Cocoa19.7
Tea13.2
Agricultural raw materials22.5100.0
Cotton34.2
Medium23.1
Long11.1
Wool28.7
Fine21.3
Coarse7.4
Rubber23.1
Hides8.0
Jute4.0
Sisal2.0
Metals27.9100.0
Copper48.6
Iron ore20.8
Tin11.0
Aluminum10.2
Zinc3.7
Nickel3.0
Lead2.7

Based on the shares in the value of exports by developing countries in 1968–70.

Based on the shares in the value of exports by developing countries in 1968–70.

IV. Determinants of Commodity Price Movements: Other Empirical Results, 1958–821
Changes in Explanatory Variables
Sample

Period
ConstantLagged

price
Economic

activity
Wholesale

price

index
Exchange

rate
Interst

rate

changes
Supply shocks
FoodBeveragesR2¯D-WSEE
All Commodities
1958–82–0.028**0.0851.299**1.730**–0.919**–0.004–0.017–0.061*
(-4.66)(0.90)(5.41)(5.45)(-2.92)(-1.75)(-0.80)(-2.16)0.4621.780.038
1958–71–0.013*0.0130.556*1.488*–0.492–0.009
(-2.58)(0.10)(2.35)(2.18)(-0.68)(-1.97)0.1411.970.023
1972–82–0.049**0.0341.719**2.325**–1.068*–0.003–0.0290.071
(-3.14)(0.23)(4.06)(3.79)(-2.37)(-1.06)(-0.97)(-1.81)0.5141.670.051
Food
1958–82–0.032**0.0750.939*2.399**–1.135*–0.000–0.047
(-3.39)(0.74)(2.46)(4.99)(-2.18)(-0.08)(-1.35)0.2931.930.061
1958–71–0.007–0.0280.5110.7320.294–0.007
(-0.96)(-0.20)(1.52)(0.78)(0.28)(-1.22)1.870.033
1972–82–0.091**–0.0470.7474.370**–1.716*0.003–0.064
(-3.77)(-0.30)(1.09)(4.60)(-2.26)(0.56)(-1.39)0.4081.950.080
Beverages
1958–82–0.0170.230*0.9010.9390.427–0.004–0.135*
(-1.32)(2.15)(1.74)(1.51)(0.65)(-0.86)(-2.00)0.1581.840.03
1958–710.0030.172–0.426–0.034–0.883–0.011
(-0.31)(1.21)(-0.88)(-0.03)(-0.60)(-1.27)0.0131.850.047
1972–82–0.0210.2011.8890.9830.472–0.005–0.138
(-0.62)(1.19)(1.94)(0.794)(0.492)(-0.739)(1.40)0.1471.900.118
Agricultural Raw Materials
1958–82–0.018*0.307**1.471**0.714–1.005**–0.007*
(-2.32)(3.22)(4.55)(1.90)(-2.65)(-2.50)0.4121.860.047
1958–71–0.020**0.2101.036**1.087–1.122–0.015*
(-2.83)(1.77)(3.11)(1.18)(-1.10)(-2.50)0.2181.850.033
1972–82–0.0050.2572.142**0.026–0.993–0.007
(0.23)(1.60)(3.48)(0.04)(-1.95)

Metals
(-1.99)0.4731.890.059
1958–82–0.926*0.0831.459**1.585**–1.809**–0.008*
(-2.60)(0.84)(3.67)(3.19)(-3.66)(-2.19)0.2661.960.063
1958–71–0.0210.0370.8752.983–0.313–0.003
(-1.56)(0.27)(1.41)(1.69)(-0.16)(-0.24)0.0161.970.062
1972–82–0.0320.0791.731**1.666–1.905**–0.009*
(-1.51)(0.51)(3.00)(2.03)(-3.38)(-2.07)0.3851.930.068

The dependent variable in all regressions is Δpt. These regressions differ from those in Table 5 in that the absolute values of the coefficients of the wholesale price index and the exchange rate are not constrained to be equal. The level of statistical significance is indicated by one (95 percent) or two (99 percent) asterisks. D-W denotes the Durbin-Watson statistic, and SEE denotes the standard error of estimate.

The dependent variable in all regressions is Δpt. These regressions differ from those in Table 5 in that the absolute values of the coefficients of the wholesale price index and the exchange rate are not constrained to be equal. The level of statistical significance is indicated by one (95 percent) or two (99 percent) asterisks. D-W denotes the Durbin-Watson statistic, and SEE denotes the standard error of estimate.

REFERENCES

    BosworthBarry and Robert Z.LawrenceCommodity Prices and the New Inflation (Washington: The Brookings Institution1982).

    ChuK.E.C.Hwa and K.Krishnamurty“Export Instability and Adjustments of Imports, Capital Inflows, and External Reserves: A Short-Run Dynamic Model,” in Exchange Rate and Trade Instability: Causes Consequences and Remediesed. byDavidBigman and TeizoTaya(Cambridge, Massachusetts: Ballinger1983) pp. 195214.

    CooperRichard N. and Robert Z.Lawrence“The 1972–75 Commodity Boom,”Brookings Papers on Economic Activity: 3 (1975) The Brookings Institution (Washington) pp. 671715.

    DarbyMichael R.“The Price of Oil and World Inflation and Recession,”American Economic Review (Nashville) Vol. 72 (September1982) pp. 73851.

    EnochC.A. and M.Panic“Commodity Prices in the 1970s,”Bank of EnglandQuarterly Bulletin (London) Vol. 1 (March1981) pp. 4253.

    GoreuxL.M.“Compensatory Financing: The Cyclical Pattern of Export Shortfalls,”Staff PapersInternational Monetary Fund (Washington) Vol. 24 (November1977) pp. 61341.

    GrilliEnzo R. and Mau-ChengYang“Real and Monetary Determinants of Non-Oil Primary Commodity Price Movements” (unpublishedWorld BankDecember1981).

    HwaE.C.“Price Determination in Several International Primary Commodity Markets: A Structural Analysis,”Staff PapersInternational Monetary Fund (Washington) Vol. 26 (March1979) pp. 15788.

    HwaE.C.“A Simultaneous Equation Model of Price and Quantity Adjustments in World Primary Commodity Markets” (unpublishedWorld BankOctober1981).

    LabysWalter C. and Harmon C.Thomas“Speculation, Hedging and Commodity Price Behaviour: An International Comparison,”Applied Economics (London) Vol. 7 (December1975) pp. 287301.

    McCallumB.T.“Competitive Price Adjustments: An Empirical Study,”American Economic Review (Nashville) Vol. 64 (March1974) pp. 5665.

    McKinnonR.An International Standard for Monetary Stabilization (Washington: Institute for International Economicsforthcoming in1984).

    MorrowDaniel T.The Economics of the International Stockholding of Wheat,International Food Policy Research InstituteResearch Report No. 18 (WashingtonSeptember1980).

    RangarajanC. and V.Sundararajan“Impact of Export Fluctuations on Income—A Cross-Country Analysis,”Review of Economics and Statistics (Cambridge, Massachusetts) Vol. 58 (August1976) pp. 36875.

    RutledgeD.J.S.“Trading Volume and Price Variability: New Evidence of the Price Effects of Speculation,”International Futures Trading Seminar: ProceedingsVol. 5 (Chicago Board of Trade1979) pp. 16174.

    Schiavo-CampoSalvatore“Instability of Developmental Imports and Economic Growth: A Theoretical Framework,”Weltwirtschaftliches Archiv (Kiel) Vol. 117 (September1981) pp. 56273.

    ZellnerArnold and FranzPalm“Time Series Analysis and Simultaneous Equation Econometric Models,”Journal of Econometrics (Amsterdam) Vol. 2 (May1974) pp. 1754.

Mr. Chu, Assistant Chief of the Commodities Division of the Research Department, is a graduate of Kyung Hee University (Seoul) and Columbia University.

Mr. Morrison, Assistant Chief of the Commodities Division, is a graduate of Washington University (St. Louis) and the University of Maryland.

Non-oil developing countries’ exports of primary commodities averaged about $120 billion annually during 1979–80. The 1981 and 1982 declines in commodity prices, therefore, accounted for reductions in their export earnings of approximately $20 billion and $15 billion, respectively. In other words, if 1980 commodity prices had been maintained, along with constant import values and export volumes, the aggregate current account deficit of the non-oil developing countries could well have been under $60 billion in 1982, compared with the actual deficit of $87 billion.

Unless indicated otherwise, instability of price is defined in this study as the average percentage deviation of the actual price from the trend for a certain period. (See the subsection entitled “price instability” in Section I for the exact definition.)

For studies of the possible effects of export instability, see Chu, Hwa, and Krishmamurty (1983); Rangarajan and Sundararajan (1976); and Schiavo-Campo (1981).

The commodity price index used in this study, unless indicated otherwise, is the International Monetary Fund’s Research Department’s Index of International Market Prices, in terms of U.S. dollars, for Primary Commodities Exported by Primary Producing Countries (1975 = 100), the same index reported in the Fund’s monthly publicaton, International Financial Statistics (with 1980 = 100). The All Commodities Index includes 35 international price series chosen as representative of the 30 commodities exported by primary producing countries. It excludes petroleum and gold. The commodity price indices are weighted by average export earnings during 1968 through 1970 in 98 countries that do not include industrial and major oil exporting countries. See Appendix III for the list of commodities.

A longer-term perspective on commodity price developments can be obtained by reference to the Economist index for all nonfuel commodities, which is available for all years since 1860. The largest annual decline in commodity prices was 33 percent from 1920 to 1921. The largest and longest cumulative decline occurred from 1924 to 1932, when commodity prices fell every year for a total decline of 47 percent. Cumulative price declines exceeding the 25 percent decline recorded during 1981–82 have occurred only four times in the last 120 years (28 percent from 1864 to 1869, 40 percent from 1920 to 1922, 47 percent from 1924 to 1932, and 26 percent from 1951 to 1953).

See the subsection entitled “estimation results” in Section II for reasons for this division of the sample period.

The long-term trend is defined as the exponential trend estimated from a semi-log regression of quarterly price on time; the medium-term trend is defined as the 19-quarter (approximately a five-year) average of quarterly prices. The marked instability of commodity prices during 1972–82 is also obtained even if the instability is measured in terms of SDRs or in real terms—that is, deflated by the United Nations price index of manufactures. Moreover, the instability during 1972–82 was relatively large even when viewed in historical perspective: for example, based on the Economist index of commodity prices, the instability of primary commodity prices during 1911–56 is estimated at 6.4 (percent of the trend), compared with 3.8 for 1957–71. The instability for 1911–56, during which the world witnessed three major wars (the First and Second World Wars and the Korean conflict) and the Great Depression, was, therefore, almost twice as great as that for 1957–71, while the instability for 1972–82 was more than three times as great as that for 1957–71.

See Appendix I for discussion of the limitations of the model.

See Appendix I for the implications of this equality constraint.

The coefficients of determination between actual and predicted prices are far higher than the adjusted coefficients of determination from the regression based on the rates of change for all groups of commodities, ranging from 0.973 for metals to 0.993 for all commodities; they are also higher for 1972–82 than for 1958–71.

See the analysis at the end of this subsection.

As indicated in Appendix I, the short-run coefficients of the inflation and exchange-rate-change variables may not be unity in the model, as specified in this paper, because of the interest rate variable’s presence; they may also differ from unity because of supply responses to price changes.

A lagged effect of interest rate changes on commodity prices is consistent with a lag associated with expectations and the lags in delivery of inventory items ordered in the previous period. In Grilli and Yang (1981), the interest rate variable is also significant with a one-quarter lag.

Interpretation of this coefficient requires caution. Had demand not grown because of long-term stagnation in economic activity, production capacity of primary commodities might not have grown as it did. Therefore, although the equation may be used for simulations and forecasts, all results should be interpreted bearing the complex causal relationships between supply and prices fully in mind.

See Appendix III for the results of the tests on the significance of lagged dependent variables.

These results are consistent with those obtained by Enoch and Panić (1981) using a different model specification but a similar breakdown of time periods.

A number of studies have tested various hypotheses on the structural shifts of the commodity markets. See Grilli and Yang (1981) and Hwa (1979).

The analysis in this section is based on the coefficients estimated with the assumption of equality between the coefficients of Δpd and – Δed. If the coefficients estimated without such constraint were used, the effects of exchange rate movements on commodity prices would become relatively smaller, but the thrust of conclusions would not change. Although the equation is designed to explain basically short-run fluctuations of commodity prices, an attempt has been made to use the equation to identify causes underlying the long-term development of prices. For that purpose, the average rates of increase in the overall index of commodity prices are compared with the rates simulated on the basis of the rates of increase in the three major variables and time trend. The simulations indicate that the dominant factor underlying the sharp acceleration in nominal commodity prices during 1972–82 was the acceleration in world inflation; the contribution made by the increase in industrial production during 1972–82 was slightly higher in absolute terms, but lower in relative terms, than during 1957–71. The increase in the contributions of the increase in industrial production was caused simply by an increase in the coefficient; the rate of increase in industrial production decelerated substantially during 1972–82.

It should be noted that, as indicated in Table 8, fluctuations in commodity prices and underlying factors, measured as the variances of the variables’ rates of change, are good indicators of the fluctuations in these variables, measured as the average deviations from trends.

The variance of the dependent variable (Δpt) in equation (1) can be expressed as the sum of the variances of the explanatory variables multiplied by their respective coefficients squared plus the variance of the disturbance term and a function of the covariances of the explanatory variables.

This assessment, however, needs to be qualified. The present study does not deal with other possible channels through which unstable exchange rates in the 1970s and 1980s could have contributed to the unstable prices of traded goods in general. For example, the fluctuation of U.S. dollar exchange rates could have aggravated the commodity price instability by triggering interventions (not fully sterilized) in foreign exchange markets by major industrial countries other than the United States, resulting in greater instability in the world money supply and in world economic activity. (See Chapter IV of R. McKinnon’s An International Standard for Monetary Stabilization (Washington: Institute for International Economics, forthcoming in 1984) for such a view.)

Stocks include those held in both exporting and importing countries. Although stocks are held for various reasons, two obvious influences on their size common to exporting and importing countries would be the volume of transactions in the commodity and the cost of holding stocks, which could be represented by the real rate of interest.

In this study, the long-run expansion of world production of primary commodities is regarded as exogenous and captured by a time-trend variable. The extreme assumption of zero elasticities, of course, is not realistic and may cause an overestimation of the impacts of demand variations on price. Although an examination of the relationship between price and production would be useful, it is left for a future study.

In closing the system in this manner, it is assumed that actual stocks are always equal to desired stocks. Equation (7) may be called a “transfer function,” in the sense used by Zellner and ralm (1974), rather than a “reduced form.” In the equation, all the current and lagged endogenous variables except price are eliminated. In a typical reduced form, any lagged endogenous variables in the structural system could be present as explanatory variables.

The coefficients for pdt and ed, in equation (7) would be equal if the parameters ai and a2 in the demand equation were equal. The value of the coefficients for pdt and edt might not be unity if the a4 coefficient were not zero, although in this case the long-run coefficient for pdt and edt should be unity.

Although the anticipation of future price movements does not play any role in the model, equation (7) is not incompatible with models in which such anticipation is explicitly incorporated. For example, a partial-adjustment commodity market model with a stock-demand equation based on an adaptive price- expectation scheme would yield a price equation similar to equation (7), although restrictions on the parameters would oe different (see McCallum (1974)). The demand and supply equations can also be specified as functions of real prices by deflating nominal prices with a series representing the world price level. Such a specification would be particularly appropriate for models dealing with individual commodities. In the present study, the model deals with a composite of non-oil primary commodities, and the distinction between the general price level in consuming countries and, say, the price of substitutes is not straightforward in practice. Ii the price of substitutes is specified as a function of the general price level, the resulting price equation would also be similar to equation (7), though it would have a more complicated lag structure.

Petroleum is a major component of the production costs of a wide range of primary commodities, in the form of energy costs for some (e.g., metals) and in the form of other inputs for others (e.g., fertilizers). Petroleum prices also strongly influence transportation costs, and, in addition, petroleum-based synthetics comprise some of the most important substitutes for agricultural raw materials. Finally, movements in petroleum prices over the last decade have perhaps been as good an indicator as any of the trend in inflation, leading major surges in prices in both the early and late 1970s. Petroleum price movements may also influence commodity prices indirectly through their impact on the business cycle, though their impact on the business cycle is still debatable. See Darby (1982).

Other Resources Citing This Publication