Jump to ContentJump to Main Navigation
Commodity Prices and Development$

Roman Grynberg and Samantha Newton

Print publication date: 2007

Print ISBN-13: 9780199234707

Published to Oxford Scholarship Online: January 2008

DOI: 10.1093/acprof:oso/9780199234707.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2018. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in OSO for personal use (for details see www.oxfordscholarship.com/page/privacy-policy). Subscriber: null; date: 21 January 2019

Long‐Run Trend in the Relative Price: Empirical Estimation for Individual Commodities

Long‐Run Trend in the Relative Price: Empirical Estimation for Individual Commodities

Chapter:
(p.35) 3 Long‐Run Trend in the Relative Price: Empirical Estimation for Individual Commodities
Source:
Commodity Prices and Development
Author(s):

Mohammad A. Razzaque

Philip Osafa‐Kwaako

Roman Grynberg

Publisher:
Oxford University Press
DOI:10.1093/acprof:oso/9780199234707.003.0004

Abstract and Keywords

This chapter details an empirical investigation to quantify the trend decline in relative prices of a number of individual commodities, and the impact of declining commodity prices on the balance of payments of LDCs, SVS, and HIPCs. The empirical evidence presented points to the presence of a statistically significant declining trend in the relative price of most individual commodities, the results highlighting the magnitude of the deterioration of the relative commodity prices during the recent past.

Keywords:   balance of payments, commodity prices, declining trend, relative price

In this chapter we estimate the trend growth rate in relative prices for individual commodities. Most studies consider the aggregate or composite relative price index in order to examine the validity of the PS hypothesis. However, individual commodity prices rather than the composite price index are more important for countries in ascertaining their problems or prospects related to export earnings and balance of payments emanating from trends in commodity prices. Some commodities might be subject to much steeper declining rates than the overall relative price index, in which case the movement in the aggregate price index would hardly reveal the practical consequences for countries specializing in them. In fact, the Prebisch–Singer thesis can also be considered for each of the major commodity groups (such as food, agricultural raw materials, minerals, etc.) and for the individual products comprising the broad classifications. It might also be of interest to see whether the hypothesis holds for all commodities, and if not, whether some characteristic features can be identified for the commodities that do not experience deteriorating net barter terms of trade. Most importantly, any general policy conclusion can only be deduced if a similar trend is revealed for most individual commodities. The empirical results, as provided in Bleaney and Greenaway (1993), show that different broad categories of primary commodities appear to exhibit price behaviour which is different from the aggregate relative price index. If this is so, then the examination of price behaviour at the individual level should be the most appropriate way of evaluating trends in commodity prices.

(p.36) 3.1. Methodology

The literature review in the previous section included studies investigating individual commodities, usually with data taken from the work of Grilli and Yang (1988) or its updated version. However, these studies (e.g. Cuddington, 1992; Kellard and Wohar, 2002; and Leon and Soto, 1997) place too much emphasis on testing unit root properties in order to determine the appropriateness of trend stationary vis‐à‐vis difference stationary models for estimating the trend equation. The results of these studies are highly influenced by whether the relative price data (for any individual commodity) are to be considered as TSP or DSP (Cuddington, 1992) and how many break points are being explicitly tested for in the process of determining the time series property of the variable under consideration (Leon and Soto, 1997; Kellard and Wohar, 2002). The underlying econometric tests have low power, as well as methodological issues that are as yet unsettled.1 Therefore, using these estimation techniques is unlikely to be informative.

One alternative to avoiding the unit root testing procedure and the ensuing pitfalls is to follow the methodology used by Bleaney and Greenaway (1993) by constructing a general error correction model that encompasses both the trend and difference stationary models. Instead of prior testing of time series properties of the data, this methodology aims at minimizing the possibility of uncovering a spurious trend by appropriately allowing for possible dynamics involved in the determination of the trend rate. Despite the standard practice in modern applied time series econometrics of testing for integrating orders of variables before running a regression, the use of such a framework that does not require prior testing for unit roots may be appropriate, given very recent developments in the field. In fact, Pesaran et al. (2001) have devised a new approach to testing for the existence of a valid long‐run relationship between variables which is applicable irrespective of whether the underlying variables are stationary, integrated of order 1 or mutually cointegrated. It has been argued that using this procedure it is unnecessary to establish the order of integration of the variables prior to estimation of the long‐run relationship and that therefore, unlike typical applications of cointegration analysis, this method is not subject to the well‐known shortcomings associated with the pre‐testing techniques. The recent development thus supports the methodology employed by Bleaney and Greenaway (1993), especially when it has been demonstrated that the determination of unit root properties for commodity prices series with violent fluctuations is anything but straightforward. Further, the Bleaney‐Greenaway approach happens to be a special case in the Pesaran et al. framework. In the following we outline the methodology (p.37) adopted by Bleaney and Greenaway (1993) and relate this to the framework of Pesaran et al. (2001).

C o n s i d e r t h e s t a n d a r d t r e n d e q u a t i o n : l n R P = a + b t + u
(1)

where RP is the relative price and all other variables are as defined in the previous section. According to Cuddington and Urzua (and all others follow them), equation (1) can only be employed if lnRP is trend stationary. If lnRP is non‐stationary and is ∼I(1), the relevant model to be estimated is:

Δ l n R p = b + u
(2)

Instead of using (1) or (2), Bleaney and Greenaway started with an autoregressive model with a time trend included:

l n R p = a + b t + c 1 n R P t 1 + u
(3)

The main difference between (1) and (3) is the inclusion of a lagged dependent variable as a regressor. Equation (3) can be rearranged to obtain:

Δ l n R p = a + b t + ψ 1 n R P t 1 + u
(4)

where, ψ = c − 1. Equation (4) becomes an ideal error‐correction model if ψ is negative, statistically significant and greater than −1, (i.e. −1 < ψ < 0). In that case, the change in lnRP is negatively related to its current level and this will pull back the short‐run deviations to the steady state long‐run trend path. By contrast, if ψ = 0, lnRP may be considered as a random walk with increasing variance over time. In essence, an error‐correction representation in (2) is only possible if the prices of primary products and manufactured goods are cointegrated.2 In the estimation of (4), if b ≠ 0, and ψ < 0, lnRP has a non‐zero deterministic trend, i.e. it has a long‐run tendency to revert to a non‐zero trend following any short‐term disturbances. The combination of b = 0 and ψ = 0 will imply no long‐term trend of lnRP but the series tends to be pulled back towards its historical mean. Thus both ‘b < 0 and ψ = 0’ and ‘b<0 and ψ < 0’ will provide empirical support for the deteriorating trend hypothesis. For two other combination possibilities, viz. b = 0 with ψ = 0 and b ≠ 0 with ψ = 0, the results should be interpreted as the evidence for the series to be a random walk with zero mean and random walk with a drift respectively.3

(p.38) Estimation of equation (4), therefore, does not require the testing of the variables for unit roots a priori. Let us now briefly consider the Pesaran et al. (2001) methodology that makes the testing for time series properties redundant and provides further justification for the use of (4) in our empirical estimation. The procedure suggested by Pesaran et al. is based on an OLS estimation of an unrestricted error correction model, a general specification of which with respect to two variables, X and Z—whose integrating orders are not determined a priori but are expected to be either zero (i.e. trend stationary) or 1 (i.e. first difference stationary)—and the trend term, T, can be written as:

Δ ln X t = α + θ T + γ ln X t 1 + i = 1 p π i Δ ln X t 1 + i = 0 g δ i + ɛ i
(5)

Estimation of (5) in itself is not interesting because the existence of a long‐run relationship can only be tested by examining the joint null hypothesis that γ = ξ = 0 with the help of either a Wald or an F‐test. The presence of a long‐run relationship requires the rejection of this null. However, as the asymptotic distribution of these statistics is non‐standard, Pesaran et al. provide the necessary critical upper (FU) and lower (FL) bound for the F‐test.4 The FU statistics are derived under the assumption that all variables are I(1) and the FL values consider all of them to be I(0). If the computed F statistic (F), which is obtained by restricting that γ = ξ = 0, is greater than the critical upper value, i.e. F > FU, the null is to be rejected and a valid long‐run relationship among the variables may be ascertained. If F < FL, then no long‐run relationship exists; finally, if FL < F < FU, the test is inconclusive. Pesaran et al. (p. 290) clearly point out that ‘[I]f the computed Wald or F‐statistic falls outside the critical value bounds, a conclusive inference can be drawn without needing to know the integration/cointegration status of the underlying regressors.’5

From (5) it is observed that if there is no other explanatory variable (apart from the trend term), the Pesaran et al. specification becomes the standard Dickey‐Fuller unit root testing equation—just as the one used by Bleaney and Greenay (1993). Under such a circumstance, the statistical significance of the lagged level dependent variable will be regarded as a proof of the long‐run relationship. However, if the dependent is non‐stationary on its level, the (p.39) distribution of T‐statistics is non‐standard and Pesaran et al. suggest that the critical value for testing the statistical significance of the lagged level dependent variable in the absence of any other explanatory variable will correspond to Dickey and Fuller's (1979) unit root T‐statistics.6 Therefore, an error‐correction type trend equation model that encompasses both trend and difference stationary models such as the one in (4) not only avoids the problems of unit root testing procedures but is also justified.

In Dickey‐Fuller type equations, such as the one in (4), special importance is given to the problem of serial correlation. The concern over the presence of serial correlation is usually addressed by the inclusion of one or more lags of the dependent variable as regressor.7 Thus a more general form of equation (4) can be written as:

Δ ln R P = a + b T + i = 1 m η * Δ ln R P t 1 + Φ ln R P t m + u t w h e r e , Φ = ( I η i * )
(6)

And the long‐run trend rate is given by: β = −Φ−1

3.2. Estimation Results

We now turn to the results. Except for one instance, the data used here are for prices of individual commodities. Two different datasets have been used to obtain the information on prices. First, an attempt was made to gather the data on individual commodities in Grilli and Yang (1988). Of the 24 commodities, data was obtained on 13 covering the period 1900–87.8 These are cocoa, coffee, tea, bananas, sugar, rice, wheat, maize, cotton, jute, palm oil, copper and tin.9 (p.40) The series was then updated to 2001 using comparable information.10 All data were gathered in nominal US dollars and then the unit value index of the manufactured goods exports of the industrial countries was used as the deflator to compute commodity‐specific net barter terms of trade.11

Apart from the Grilli‐Yang dataset, the UNCTAD database on commodity prices was used to estimate the trend growth rate for as many as 60 individual commodities.12 The longest span of the data available from UNCTAD is 1960–2002. In most cases these data were available in US dollars and the unit value index of manufactured goods exports of developed market economy countries was used as the deflator.

3.2.1. Trend growth rates of relative prices for commodities in the Grilli‐Yang dataset

Figure 3.1 plots the updated 13 commodity‐specific relative prices in the Grilli‐Yang dataset. All relative prices exhibit wide fluctuations with spectacular peaks and troughs. Nevertheless, even a casual look at the graph clearly reveals a declining trend in the net barter terms of trade of rice, wheat, maize, cotton and palm oil. For bananas a strong declining trend is discernible from around 1930 and for tea and jute from the mid‐1950s. Apart from two skyscrapers, a deteriorating trend in the real price of sugar is also clear. Tin is the only commodity that witnessed a strong positive trend until the early 1970s, largely because of the success of the International Tin Agreement (ITA). Since then, the real tin price began falling before the major crash of the mid‐1980s, which coincided with the collapse of the ITA. The most striking feature of Figure 3.1 is that since the 1970s a strong downward trend in the real prices of all commodities is apparent.

Table 3.1 provides the regression results for the commodities in Figure 3.1. It needs to be mentioned here that except lnRPt − 1, for all variables the standard t‐ratios are valid, which implies that as a rule of thumb if the t‐ratio is greater than 2 the respective coefficient is statistically significantly (p.41)

Long‐Run Trend in the Relative Price: Empirical Estimation for Individual Commodities

Figure 3.1. Relative Prices of 13 Commodities: 1900–2001

Note: The figures correspond to relative commodity prices.

different from zero at the 5 per cent error probability level. For lnRPt − 1, however, the estimated t‐ratios should be compared with those of the critical values computed by Dickey and Fuller (1979) to draw inferences. These critical values are considerably higher than the standard t‐ratios. In fact, in order for the lnRP t − 1 term to be statistically significantly different from zero, the computed t‐ratio should be as high as 3.13 (absolutely) at the 10 per cent significance level. Following the usual practice with Dickey‐Fuller regressions, the first order lagged dependent variable (i.e. ΔlnRPt − 1) is always retained in the equation irrespective of its statistical significance. In only a few cases additional lags were also included to remove the problem of serial correlation.

In a number of equations, regression residuals turn out to be non‐normal, which should be considered as a serious problem preventing the drawing of valid inferences. As sudden and precipitous price fluctuations are common, as reflected in Figure 3.1, it is unsurprising that a simple trend equation will fail to explain such movements, resulting in residuals that are not normally distributed. Bleaney and Greenaway (1993) also encountered the problem of (p.42) non‐normality in estimating the trend growth rate in the aggregate commodity price index for which they re‐estimated their equation after dropping the first 25 years of data from their sample, arguing that those years were associated with exceptionally violent movements of commodity prices. Figure 3.1, however, does not seem to suggest that at the individual commodity levels the movement in commodity prices prior to 1925 was different from that in the latter period and therefore it was decided not to curtail the sample to tackle the problem of non‐normality. Instead, we have used dummy variables to control for the sudden rise(s) and decline(s) in commodity prices. This approach is tantamount to pulling the atypical data points to a normal year, which is defined by the trend equation. All dummies inserted in all equations were found to be highly significant. The estimated equations with the dummies can be considered as the preferred specification and growth rates corresponding to these equations will used for reference.

Results reported in Table 3.1 show that for ten out of thirteen commodities the estimated coefficients on the trend equation are negative; only for cocoa, coffee and tin is the sign on the coefficients positive. For eight commodities—tea, sugar, rice, wheat, maize, cotton, jute and palm oil—the estimated trend is negative and statistically significant at the 10 per cent confidence level. Among the three commodities with a positive sign, only the trend rate for tin is significant over the period 1900–87. In all regressions the lagged level dependent variable (lnRPt − 1) is negative and less than zero, as is expected in the case of an error‐correction model. In as many as eight cases the T‐ratio on lnRPt − 1 is higher than the Dickey‐Fuller critical value (at least at the 10 per cent level), which implies that for these commodities a valid long‐run trend growth rate can be estimated irrespective of the order of integration of the real price series. Although for another five commodities the estimated T‐ratio on the lagged level term is lower than the critical value, it is always significantly different from zero, considering the standard test of significance for stationary variables. Indeed, if any of these relative price series is TSP, estimation of trend growth rate for it from the regression results can be considered valid.13

In the column for ‘trend rate’, the long‐term trend growth rate in relative price (in per cent per annum) has been computed for all commodities for which the coefficient on the trend term appears to be significant at least at the 10 per cent error probability level. For cocoa, coffee, banana, copper and tin the trend term is not significant and the exact interpretation will depend on whether one considers lnRPt − 1 in those equations to be significant or not.14 (p.43) If the coefficient on the lagged level variable is to be considered significant, the real price series of these commodities have no long‐term trend but they tend to be pulled back towards their historical mean.15

Negative trend growth rates have been estimated for tea, sugar, rice, wheat, maize, cotton, jute, palm oil and copper. The trend rates lie between −0.79 and −1.43 per cent per annum and the results show that during the past century most commodity prices have fallen at an annual rate of above 1 per cent. This is considerably higher than the estimates of Grilli and Yang (1988) and Bleaney and Greenaway (1993) which were in the range −0.6 to −0.7 per cent per annum.

3.2.2. Estimation for commodities in the UNCTAD database

3.2.2.1. ESTIMATES FOR BROAD COMMODITY GROUPS

The commodity price bulletin of UNCTAD provides information on prices for a large number of individual commodities since 1960. It also provides an aggregate commodity price index and price indices for another four broad commodity groups, viz. food and beverages, vegetable oils and oilseeds, agricultural raw materials, and minerals and metals. Before analysing the individual commodities, Table 3.2 gives an estimate for the broad commodity groups.

In general, the estimation of the trend equation was affected by the normality problem mainly due to the sudden jump in commodity prices around the mid‐1970s, as Figure 3.2 exhibits one clear peak for all broad commodity groups. Therefore, for most equations dummy variables were included to control for these sharp price movements. As with the previous cases, the equations with the dummies are considered to be the preferred specifications.

The results reported in Table 3.2 show that for every broad commodity group the trend variable appears to be statistically significant. In every preferred specification, apart from the one for the food and beverage group, the computed T‐ratio on the lagged dependent variable (lnRPt − 1) exceeds the Dickey‐Fuller critical values at least at the 90 per cent level.16 Although a firm conclusion about the long‐run relationship cannot be made for food and beverages, separate estimates for the ‘food only’ and ‘beverages only’ sub‐groups strongly rejected the null hypothesis of statistical insignificance of lnRPt − 1, suggesting that irrespective of the order of integration of the dependent variables the estimated trend growth rates are valid. In no regression is lnRPt − 1 insignificant in comparison with the t‐statistics following standard distribution and applicable for drawing inferences in the case of stationary variables.

(p.44)

Table 3.1. Regression Results for 13 Commodities (with Updated Grilli‐Yang Series: 1900–2001)

ΔlnRPt

Constant

T

lnRPt−1

ΔlnRPt−1

ΔlnRPt−2

Dummies

Adjusted R2

Serial Correlation

Functional Form

Normality

Heterosce dasticity

Trend rate (%)

Cocoa

−0.13* (−1.76)

0.00038 (0.044)

−0.114 (−2.15)

−0.15 (1.53)

−0.28** (−2.82)

None

0.125

0.30

0.60

8.96**

0.06

×

0.75** (2.98)

.00093 (0.11)

−0.86 (−1.76)

0.10 (1.13)

−0.29*** (3.09)

D47

0.23

0.03

1.14

2.66

0.66

×

Coffee

−0.16** (−1.90)

0.0014 (0.15)

−0.19 (−2.94)

0.05 (0.55)

None

0.06

0.02

0.48

4.38

0.099

×

Tea

0.04 (1.21)

−0.001* (−1.68)

−0.11 (−2.24)

0.07 (0.1)

None

0.02

4.15**

0.60

11.23**

0.37

−1.04

0.92*** (0.3)

−0.001** (−2.15)

−0.09 (−2.50)

0.04 (0.48)

D85, D77, D84, D54

0.33

0.72

2.38

2.50

1.47

−1.26

Banana

0.05*** (3.11)

−0.004 (−1.16)

−0.14*** (−3.29)

0.10 (1.02)

None

0.05

2.95

0.03

3.91

White

×

Sugar

0.29*** (3.29)

−0.004*** (−3.02)

−0.40*** (−4.84)

0.17* (1.69)

None

0.17

2.50

79.69**

0.64

−1.02

−1.02

2.82*** (5.19)

−0.004*** (−4.97)

−0.38*** (−6.57)

0.16** (2.38)

D20, D21, D63, D65, D74, D80

0.64

3.22

1.86

0.41

3.03

−1.21

Rice

0.19*** (3.48)

−0.003*** (−3.62)

−0.25** (−4.07)

0.29*** (2.98)

None

0.15

3.23

0.91

23.97***

0.05

−1.25

0.47*** (2.16)

−0.003** (−3.81)

−0.24*** (−4.21)

0.29 (3.31)

D73, D82

0.34

0.017

0.57

2.04

0.77

−1.28

Wheat

0.27*** (4.34)

−0.0037*** (−4.19)

−0.32*** (−4.84)

0.28*** (2.89)

None

0.23

0.09

17.41***

1.38

0.18

−0.92

0.82*** (5.23)

−0.003*** (−4.59)

−0.32*** (−5.10)

0.27*** (2.95)

−0.54*** (−3.74)

0.28

0.94

0.10

5.90*

0.69

−1.17

Maize

0.31*** (3.92)

−0.004*** (−3.80)

−0.35*** (−4.41)

0.09 (0.93)

None

0.15

1.33

1.88

21.88***

0.66

−1.18

−1.77*** (−5.41)

−0.0032*** (−3.46)

−0.22* (−3.23)

0.084 (0.33)

D21, D38, D48

0.40

4.40

0.006

4.89

1.08

−1.43

Cotton

0.16*** (3.80)

−0.0025*** (−3.94)

−0.196** (−3.89)

0.09 (0.76)

None

0.08

3.67

0.31

0.29

White

−1.29

Jute

0.11** (2.20)

−0.0017** (−2.11)

−0.19** (−3.26)

0.06 (0.10)

None

0.07

1.07

1.04

14.67***

0.99

−0.9

−0.77*** (−3.65)

−0.0012* (−1.62)

−0.16** (−3.85)

0.05 (0.58)

D86

0.21

0.17

0.39

1.13

0.48

−0.79

Palm Oil

0.18*** (3.11)

−0.0036*** (−3.52)

−0.29*** (−4.16)

0.21** (2.10)

None

0.13

5.19

8.70***

6.33**

3.08

−1.25

−0.60*** (−2.85)

−0.003*** (−3.11)

−0.26** (−3.97)

0.14 (1.47)

D86

0.24

1.58

7.88***

1.00

0.22

−1.17

Copper

0.016 (0.49)

−0.004 (−0.75)

−0.18* (−3.21)

0.11 (1.11)

None

0.07

1.94

0.39

2.51

0.35

×

Tin

−0.17** (−2.47)

0.0007 (1.02)

−0.20** (−3.39)

0.10*** (0.09)

None

0.08

5.65***

3.24

9.17***

0.12

×

−0.76 (−4.64)

0.001 (1.6)

−0.16 (−2.76)

0.153 (1.66)

0.032 (0.09)

D86

0.19

2.89

2.04

2.98

0.48

×

Note: Figures within the parentheses indicate t‐ratios. Statistical significance at the 1, 5, and 10 per cent levels is indicated by

***, **, and * respectively. Critical values for the coefficient of lnRPt−1 at the 10, 5 and 1 per cent significance levels are, respectively, −3.13, −3.45 and −4.10. Variables with the letter ‘D’ followed by two digits indicate a dummy variable. For example, D73 indicates a dummy variable with 0 for 1973 and 1 for all other years. All dummies are inserted separately and are always significant at the 1 per cent level. ‘White’ indicates that due to heteroscedasticity standard errors are derived from the White's (1980) heteroscedasticity consistent variance‐covariance matrix. × implies that the coefficient on the trend term is not significant and hence the trend growth rate is not estimated and can be considered to be zero.

(p.45)

Table 3.2. Regression Results for Broad Commodity Groups as in UNCTAD Commodity Price Bulletin: Annual Data (1960–2002)

(ΔlnRPt)

Constant

T

lnRPt−1

ΔlnRPt−1 }

ΔlnRPt−2

Dummies

Adjusted R 2

Serial Correlation

Functional Form

Normality

Heterosce- dasticity

Trend rate (per cent)

Aggregate Commodity Price Index

0.19** (2.68)

−0.007*** (−2.82)

−0.356 (−3.06)

0.18 (1.15)

None

0.14

1.49

1.15

4.53*

1.77

−1.96

0.09*** (5.96)

−0.007*** (−3.71)

−0.419*** (−4.53)

−0.025 (−0.19)

D73, D74

0.48

0.85

1.33

1.15

0.27

−1.82

Food and Beverages

0.18** (2.34)

−0.007** (−2.45)

−0.30 (−2.78)

0.16 (1.05)

None

0.11

1.72

1.05

3.09

0.53

−2.37

‐ Food only

0.30*** (3.03)

−0.008*** (−2.82)

−0.40*** (−3.51)

0.32** (2.18)

None

0.20

0.94

0.19

8.23***

0.08

−2.19

1.79*** (8.14)

−0.009*** (−4.57)

−0.51*** (−6.53)

0.15 (1.41)

D80, D73, D74

0.65

0.16

0.96

2.05

1.53

−1.85

‐Beverages only

0.05 (0.79)

−0.006* (−1.94)

−0.26** (−2.42)

0.14 (0.89)

None

0.07

0.002

0.21

9.30***

0.009

−2.63

1.27*** (5.01)

−0.006** (−2.29)

−0.28* (−3.33)

−0.002 (0.02)

D76, D77

0.42

0.78

0.30

4.21

0.42

−2.25

Vegetable Oils and Oilseeds

0.16 (1.57)

−0.008* (−1.96)

−0.302 (−2.05)

0.17 (1.16)

None

0.35

0.61

5.79**

1.09

1.09

−2.92

0.70*** (3.48)

−0.01** (−2.84)

−0.49** (−3.79)

0.31** (2.07)

D73

0.29

0.04

4.42

0.02

0.28

−2.38

Agricultural Raw Materials

0.13** (2.17)

−0.005** (−2.56)

−0.43*** (−2.90)

−0.37*** (−2.75)

None

0.37

0.088

1.99

77.86***

0.04

−1.38

0.64*** (8.22)

−0.331** (−2.16)

−0.295*** (−3.14)

−0.32*** (−3.84)

D73, D76

0.76

0.95

0.37

1.00

1.29

−1.08

Minerals and Metals

0.27*** (3.74)

−0.01*** (−3.83)

−0.59*** (−4.19)

0.24* (1.66)

None

0.27

0.05

5.28

0.67

1.21

−1.75

Note: Critical values for the coefficient of lnRPt−1 at the 10, 5 and 1 per cent significance levels are, respectively, −3.13, −3.45 and −4.10. Variables with the letter ‘D’ followed by two digits indicate a dummy variable. For example, D73 indicates a dummy variable with 0 for 1973 and 1 for all other years. Figures within the parentheses indicate t‐ratios. ‘White’ indicates that due to heteroscedasticity standard errors are derived from the White's (1980) heteroscedasticity consistent variance‐covariance matrix. × implies that the coefficient on the trend term is not significant and hence the trend growth rate is not estimated and can be considered to be zero.

(p.46)

Long‐Run Trend in the Relative Price: Empirical Estimation for Individual Commodities

Figure 3.2. Real Prices of Broad Commodity Groups

Note: All prices are relative to the unit value of the index of manufactured goods exports of developed market economy countries.

Source: Authors' calculation based on data from Commodity Price Bulletin (UNCTAD).

Long‐Run Trend in the Relative Price: Empirical Estimation for Individual Commodities

Figure 3.3. Estimated Growth in Relative Prices for Broad Commodity Groups

Source: Based on the results associated with the preferred specifications in Table 3.2.

(p.47) According to the estimates in Table 3.2, trend growth rates for prices of broad commodity groups fell by between −1.08 to −2.92 per cent per annum. The rate of decline during the past 40 years was lowest for agricultural raw materials and highest for vegetable oils and oilseeds and food and beverages (Figure 3.3). On the whole, the aggregate relative price for primary commodities has been subject to an annual trend deterioration of −1.82 per cent. It is important to note that our preferred specification does not exaggerate the rate of decline in commodity prices. In fact, in every case the growth rate associated with the preferred specification in Table 3.2 is lower than the equations that do not include any dummy variable to control for residual non‐normality. Most dummy variables used to control for a sharp rise in prices fall within the relatively early years of the sample, resulting in a negative effect on the magnitude of the trend growth rate.

3.2.2.2. ESTIMATES FOR INDIVIDUAL COMMODITIES

Within each of the broad commodity classifications, it is possible to estimate the growth rate in relative price for several individual commodities. UNCTAD's Commodity Price Bulletin provides a wealth of information on prices of many commodities which are narrowly defined, and data on them are gathered in a consistent manner. Subject to the availability of data for a reasonable time period, 17 individual food and beverages products, 9 vegetable oils and oilseeds, 17 agricultural raw materials and 17 commodities in the minerals, ores and metals sub‐group were used for empirical estimation. The unit value index of manufactured goods exports of developed market economy countries was used to convert the nominal price series in real terms. Despite frequent fluctuations, a close look at the graphical plots of the real prices of the individual commodities, as given in Appendix 3.103.12, reveals a declining trend in real prices for most of the commodities.

Initial experiments with the estimation of the trend equation revealed problems related to the model diagnostic tests in a number of equations. The main source of the problems could be found to be associated with the commodity price boom of the mid‐1970s. Therefore, for some commodities our preferred specification includes dummy variables to control for atypical price rises. One important feature of the specification is that if the dummies were not included, growth rates for real prices of commodities would have been higher (absolutely). Therefore, the estimated models, as presented in Appendices 3.13.4, do not exaggerate the trend decline in commodity prices.

Among the seventeen products in the food and beverages category, the sign on the trend term for all commodities except for white pepper turns out to be negative (see Appendix 3.1).17 Only for cocoa and white pepper is the trend (p.48) term found to be not statistically different from zero. The lagged level dependent variable in every equation is correctly signed and is always significant in comparison with the standard t‐statistics. When compared with the Dickey‐Fuller critical values, the statistical significance of the variable is retained for all commodities except coffee and beef. The estimated trend growth rate varies between −3.27 per cent per annum for tea and −0.92 per cent for bananas. All nine vegetable oil and oilseed commodities have a significant negative trend growth rate along with the statistical significance of lnRPt − 1 (Appendix 3.2).

Turning to agricultural raw materials, Appendix 3.3 only shows a significant positive trend rate for wood items such as non‐coniferous woods, sawn wood, tropical logs and plywood. While the estimated trend in the cases of jute and sisal failed to become statistically significant, for cotton (various types), linseed oil, leaf tobacco, cattle hides and rubber there was evidence of significant declining terms of trade. Among the seventeen products covered in the category of minerals, ores and metals, as many as thirteen have a significant lag dependent variable. However, the coefficient on the trend term in the equations for phosphate rock, nickel (cathodes), refined lead, tin (ex‐smelter), gold and zinc are not significant.

On the whole, the application of the error‐correction model in the estimation of trend rate is therefore found to be satisfactory. The correct sign and the significance of the error‐correction term even after comparing with the Dickey‐Fuller critical values in the overwhelming majority of the equations suggest that the trend growth rates are valid without a priori knowledge about the order of integration of the variables.

Figure 3.4 summarizes the trend growth rates in relative prices over 1960–2002 by individual commodities.18 It is found that two minerals, tungsten ore and silver, have witnessed the steepest decline over the past four decades. The trend declining rates for tea and coffee are found to be much higher than estimates using the very long‐term data of 1900–2001.19 Among the cereals, the real price decline for rice has been much worse than that for wheat and maize. For eight commodities, cocoa, sugar, white pepper, jute, phosphate rock, tin, gold and zinc, the trend growth rates are not statistically significantly different from zero. While the results for cocoa in both the very long‐term sample and the sample beginning from 1960 are qualitatively the same, for tin the positive rate of growth for 1900–2001 has now been turned into one of no significant trend.

The results in Figure 3.4 cannot be readily compared with those of Table 3.1, which uses very long time‐series data. Given the substantial fluctuation in commodity prices, the estimation of the trend growth rate will be affected by (p.49)

Long‐Run Trend in the Relative Price: Empirical Estimation for Individual Commodities

Figure 3.4. Trend (1960–2002) Growth Rates for Individual Commodities with UNCTAD Data

the time span chosen for analysis. The review of the literature in the previous section also highlights this problem. While a very long‐term analysis, such as the one covering 100 years, is useful in understanding the evolution of price movements and in studying the pattern and nature of mean reversion in the data, trends emanating from a relatively recent past are probably more informative in understanding the implications for developing countries. There is not much point in arguing about whether to make the starting point of the sample 1940, 1960, or 1970; nevertheless it might be useful to study the trend in the post‐war period. However, while 1970 should be avoided as the initial point because of the commodity price boom, a starting point in the 1980s reduces the number of observations that can be worked with.20 On the other (p.50)
Long‐Run Trend in the Relative Price: Empirical Estimation for Individual Commodities

Figure 3.5. Trend Growth Rates since the 1960s: Grilli‐Yang versus UNCTAD Data

hand, the data show that in the 1960s most relative commodity prices were quite stable; therefore, the starting point of the estimates presented should not be inappropriate.

One serious concern is whether these results should be considered as evidence for a potential structural break in the very long‐run trend equation of 1900–2001. The issue of structural break has been discussed in a number of studies, including Bleaney and Greenaway (1993), where the authors found statistical support for a once‐for‐all drop in commodity prices after the 1980s. Thus the possibility of a structural break in the very long‐run trend equation cannot be overlooked. However, exactly what time frame should be considered for testing such a break will remain an important issue to be resolved if such a debate is to be informative.21 It is also true that the precise point of structural break might be different for different commodities.

What will be the magnitude of the trend decline rate in real commodity prices if the Grilli‐Yang type long‐run data series is restricted to one comparable with the time frame of the UNCTAD databases? To answer this question trend growth rates for 13 commodities using the 1900–2001 dataset were also estimated for a period from 1960 to the end of the sample. In Figure 3.5 the results are compared with those reported in Table 3.1, together with those (p.51) plotted in Figure 3.4. It now becomes obvious that the estimates from the UNCTAD data and from the Grilli‐Yang data for the comparable period beginning in 1960 do not provide very different results. The biggest discrepancy between the two series is for jute. This is because in the UNCTAD data the trend rate for jute appears to be not significant, while using the updated Grilli‐Yang data, results in the trend term are significant only at a somewhat lower level of statistical significance (i.e. at the 10 per cent level). Figure 3.5 also shows that, with the exception of sugar, jute and cocoa, the growth rate over 1900–2001 is much lower than the sample comprising the data for only the post‐1960 period.

3.3. Conclusion

The empirical evidence presented in this chapter strongly shows the presence of a statistically significant declining trend in the relative price of most individual commodities. When the data spanning the very long period of 1900–2001 are considered, the estimated trend rates lie between −0.79 and −1.43 per cent per annum. Much higher rates of decline are observed for the relatively recent period. Between 1960 and 2002 the aggregate relative price of commodities has fallen at an annual rate of −1.82 per cent, with the corresponding figures for individual commodities ranging from −0.9 to −3.50 per cent. Therefore, the use of very long time‐series data considerably undermines the magnitude of the deterioration of relative commodity prices during the recent past.

(p.52)

Appendix 3.1. Estimated Trends in Relative Prices for 17 Food and Beverage Products (With UNCTAD Data 1960–2002)

ΔlnRPt

Constant

T

lnRPt−1

ΔlnRPt−1

ΔlnRPt−2

Dummies

Adjusted R 2

Serial Correlation

Functional Form

Normality

Heterosce‐ dasticity

Trend rate (per cent)

Sample

Coffee‐1

0.92*** (3.10)

−0.0089** (−2.18)

−0.29 (−2.64)

0.086 (−0.54)

D77

0.23

1.12

4.06

2.41

0.02

−3.07

1960–2002

Coffee‐2

0.87*** (3.86)

−0.0053* (−1.94)

−0.25 (−2.88)

0.30* (1.69)

D76

0.31

Cochrane‐ Orcutt

0.65

1.23

0.06

−2.11

1960–2002

Coffee‐3

0.82*** (3.42)

−0.008** (−2.37)

−0.24 (−2.97)

0.30** (2.06)

D76

0.33

1.66

2.66

1.90

0.16

−3.45

1960–2002

Coffee‐4

0.79*** (3.34)

−0.0084** (−2.10)

−0.325 (−2.97)

0.27* (1.73)

0.21 (1.44)

D76

0.30

2.24

2.16

2.19

0.25

−2.60

1960–2002

Cocoa

0.64*** (2.88)

−0.0043 (−1.31)

−0.26* (−3.11)

0.45*** (2.98)

D76

0.28

0.03

0.81

0.54

2.01

×

1960–2002

Tea

1.49*** (7.46)

−0.022*** (−4.53)

−0.69*** (−5.25)

0.168 (1.34)

D77, D84

0.63

3.58

1.68

2.01

0.68

−3.27

1960–2002

Wheat, Argentina

0.12 (0.58)

−0.0085*** (−2.92)

−0.43*** (−4.30)

0.34** (2.64)

D90, D74

0.44

0.001

0.06

1.33

1.15

−1.98

1960–2002

Wheat, US

0.65*** (5.11)

−0.0077*** (−2.82)

−0.43** (−4.08)

0.36** (2.79)

D73

0.46

0.33

0.41

5.29*

0.16

−1.78

1960–2002

Maize

0.17 (0.92)

−0.008** (−2.48)

−0.41** (−3.60)

0.31 (2.10)

D73, D90

0.35

0.22

2.38

0.44

0.55

−2.00

1960–1997

Rice

1.06*** (5.78)

−0.0138*** (−3.51)

−0.46*** (−4.14)

0.29** (2.27)

D73

0.47

3.14

1.45

1.55

0.73

−2.98

1960–2002

Sugar

2.76*** (7.47)

−0.0039 (−1.05)

−0.47*** (−5.74)

0.27*** (2.88)

D74, D80

0.64

2.46

0.11

0.92

0.38

×

1960–2002

Beef

0.13** (2.11)

−0.0065** (−2.44)

−0.29 (−2.81)

0.14 (0.90)

None

0.11

0.64

5.91**

2.52

2.52

−2.22

1960–2002

Yellow Maize

0.62*** (4.66)

−0.010*** (−3.23)

−0.466** (−3.98)

0.33** (2.48)

D73

0.39

0.70

0.025

0.43

0.13

−2.19

1960–2002

Bananas

0.05 (1.40)

−0.005** (2.45)

−0.54* (−3.31)

0.13 (0.80)

None

0.19

1.23

0.02

0.11

0.55

−0.92

1960–2000

White Pepper

0.32* (1.61)

0.002 (0.85)

−0.33*** (−5.08)

0.69*** (6.92)

D97

0.63

0.007

2.85

1.81

0.011

×

1960–2002

Soybean Meal

1.11*** (7.23)

−0.0105*** (−3.49)

−0.51*** (−4.17)

−0.03 (−0.26)

−0.33*** (−3.40)

D73

0.72

1.07

3.42

1.83

0.29

−2.08

1960–2002

Fish Meal

1.15*** (5.35)

−0.007*** (−2.38)

−0.71*** (−4.97)

0.18 (1.35)

D73

0.46

1.46

1.79

1.55

2.15

−1.02

1960–2002

(p.53)

Appendix 3.2. Trend Growth Rates in Relative Prices for 9 Vegetable Oils and Oilseed Products (With UNCTAD Data 1960–2002)

ΔlnRPt

Constant

T

lnRPt−1

ΔlnRPt−1

Δln RPt−2

Dummies

Adjusted R2

Serial Correlation

Functional Form

Normality

Heterosced‐ asticity

Trend rate (per cent)

Sample

Yellow  Soybean

0.81*** (6.10)

−0.009*** (−3.34)

−0.42*** (−4.15)

0.08 (0.68)

D73

0.49

0.05

0.25

1.46

0.93

−2.18

1960–2002

Crude  Soybean Oil

0.70** (2.26)

−0.0101** (−2.73)

−0.49*** (−4.31)

0.21 (1.63)

D73, D74, D86

0.54

2.61

0.61

1.80

1.58

−2.04

1960–2002

Sunflower  Oil

0.19 (0.67)

−0.006* (−1.89)

−0.37*** (−3.44)

0.12 (0.95)

D74, D86

0.43

0.27

0.26

1.72

2.34

−1.86

1960–2002

Groundnut  Oil

0.65*** (3.19)

−0.008** (−2.54)

−0.56*** (−4.13)

0.18 (1.28)

D74

0.31

0.30

0.09

0.25

1.05

−1.55

1960–2002

Copra  in bulk

1.09*** (3.50)

−0.024*** (−4.07)

−0.86*** (−5.16)

0.24** (1.55)

D74

0.41

3.73

3.58

0.018

0.06

−2.74

1960–2002

Coconut Oil

1.85*** (4.50)

−0.02*** (−4.29)

−0.80*** (−5.33)

0.15* (1.06)

D74, D84

0.47

3.02

4.81

1.29

0.016

−2.93

1960–2002

Palm  Kernel Oil

1.81*** (4.14)

−0.02*** (−3.74)

−0.74*** (−4.88)

0.10 (0.67)

D74, D84

0.43

2.98

3.01

0.44

0.05

−2.75

1960–2002

Palm Oil

1.24*** (3.69)

−0.013*** (−3.03)

−0.54** (−4.09)

0.1 (0.97)

D74, D84

0.33

3.81

4.36

2.43

0.22

−2.55

1960–2002

Cottonseed  Oil

0.70*** (3.73)

−0.10*** (−2.86)

−0.47** (−3.68)

0.002 (0.018)

D74

0.14

3.37

2.11

2.16

0.0003

−2.39

1960–2002

(p.54)

Appendix 3.3. Estimated Trends in Relative Prices of 16 Products in Agricultural Raw Materials (With UNCTAD Data 1960–2002)

ΔlnRPt

Constant

T

lnRPt−1

ΔlnRPt−1

ΔlnRPt−2

Dummies

Adjusted R2

Serial Correlation

Functional Form

Normality

Heterosced‐ asticity

Trend rate (per cent)

Sample

Cotton  (US, Memphis)

0.16** (2.03)

−0.088** (−2.52)

−0.38 (−2.56)

−0.06 (−0.36)

None

0.15

1.25

0.003

4.36

2.31

−2.29

1960– 2002

Cotton  (US, New Orleans)

0.16** (2.08)

−0.0088** (−2.56)

−0.42 (−2.66)

−0.028 (−0.16)

None

0.15

0.79

0.001

1.65

2.06

−2.10

1960– 2002

Cotton  (Outlook Index A)

0.73** (2.77)

−0.026*** (−2.95)

−0.76* (−3.14)

0.07 (0.35)

None

0.29

3.35

0.41

2.13

2.26

−3.50

1960– 2002

Cotton  (Outlook  Index B)

0.71** (2.49)

−0.024** (−2.65)

−0.85* (−3.31)

0.21 (0.99)

None

0.29

0.002

0.58

2.52

0.80

−2.81

1960– 2002

Jute

−0.20 (−0.82)

−0.009 (−1.51)

−0.27 (−2.05)

0.03 (0.29)

D84, D86

0.69

0.23

0.09

1.24

0.85

×

1960– 2002

Sisal  (Tanzania/  Kenya)

0.06 (0.97)

−0.0037 (−1.31)

−0.41* (−3.56)

0.41** (2.72)

None

0.23

0.14

0.008

1.20

0.04

×

1960– 2002

Sisal  (Uganda)

0.15 (1.76)

−0.005** (−1.79)

−0.47** (−3.65)

0.36*** (2.32)

None

0.22

0.08

0.71

1.27

1.36

−1.23

1960– 2002

Non‐coniferous  wood

−0.18** (−2.42)

0.007** (2.77)

−0.41 (−2.28)

0.15 (0.74)

None

0.33

0.26

7.37**

0.47

White

+1.88

1972– 2002

Sawn Wood

−0.26*** (−2.12)

0.109** (2.46)

−0.58** (−3.34)

0.29 (1.60)

None

0.21

0.11

0.002

0.02

0.10

+1.87

1970– 2002

Tropical Logs

0.24*** (3.13)

−0.005** (−2.31)

−0.80*** (−4.55)

0.24 (1.50)

None

0.39

0.05

1.05

5.47

2.83

−0.69

1970– 2002

Tropical  Logs (Gabon)

0.006 (0.10)

0.0024 (0.97)

−0.48 (−2.84)

0.166 (0.87)

None

0.15

0.36

0.88

0.002

2.13

×

1970– 2002

Plywood/sheet

0.85 (0.16)

0.10*** (4.15)

0.76*** (−5.59)

0.17 (1.62)

None

0.70

0.001

0.05

0.51

0.77

+1.35

1963– 2002

Plywood/cubic  metre

0.85*** (5.17)

0.010*** (4.15)

−0.76*** (−5.59)

0.17 (1.62)

D73, D93

0.70

0.002

0.05

0.51

0.77

+1.35

1963– 2002

Linseed Oil

0.79*** (3.90)

−0.103** (−2.73)

−0.57*** (−4.81)

0.47*** (3.72)

D74

0.51

0.002

3.23

0.72

0.01

−1.79

1960– 2003

Leaf Tobacco

0.10** (2.15)

−0.006** (−2.85)

−0.51** (−3.71)

0.36** (2.26)

None

0.22

3.70

1.81

3.38

0.55

−1.23

1963– 2003

Cattle Hides

0.27** (2.70)

−0.015*** (−3.38)

−0.64*** (−4.39)

0.40** (2.62)

None

0.29

0.44

0.05

0.28

0.03

−2.48

1962– 2002

Rubber in bales

0.45*** (3.42)

−0.016*** (−3.48)

−0.66*** (−4.25)

0.29** (1.87)

None

0.28

0.76

2.07

2.14

0.07

−2.45

1960– 2002

Phosphate  rock

1.04*** (9.77)

0.0003 (0.02)

−0.27*** (−4.61)

−0.27*** (−4.61)

D74

0.79

0.05

0.008

2.98

0.42

×

1960– 2002

Manganese  ore

0.11*** (2.51)

−0.002* (−1.66)

−0.41*** (−4.91)

0.66*** (5.52)

None

0.49

1.28

2.40

1.72

2.09

−0.63

1960– 2002

Iron ore

0.64 (5.84)

−0.003* (−1.76)

−0.33** (−3.52)

0.26** (2.26)

D75, D82

0.60

0.73

1.44

2.23

1.27

−0.83

1960– 2002

Tungsten ore

0.19 (1.80)

−0.009** (−2.14)

−0.18 (−2.38)

0.23 (1.46)

None

0.10

0.35

0.24

1.15

0.16

−4.96

1960– 2002

Copper,  Grade A

0.36*** (3.17)

−0.011** (−3.07)

−0.43** (−3.47)

0.16 (1.04)

None

0.19

0.17

5.34

0.57

2.02

−2.60

1960– 2002

Copper,  Wire Brass

0.23** (2.83)

−0.007** (2.76)

−0.40* (−3.15)

0.12 (0.77)

None

0.15

0.29

0.40

1.03

0.16

−1.77

1960– 2002

Nickel, LME

0.99*** (4.96)

−0.007* (−1.71)

−0.58* (−3.40)

0.22 (1.51)

D88

0.58

1.11

0.001

2.37

0.45

−1.30

1970– 2002

Nickel,  Cathodes

0.84*** (5.04)

−0.003 (−1.56)

−0.53*** (−4.07)

0.22* (1.78)

D88

0.56

0.76

0.002

1.76

0.50

×

1960– 2002

Lead, LME

0.33*** (2.93)

−0.009** (−2.74)

−0.42* (−3.27)

0.13 (0.85)

None

0.17

3.69

0.54

4.11

0.74

−2.29

1960– 2002

Refined Lead

0.1977 (1.90)

−0.002 (−0.99)

−0.31 (−2.32)

0.04 (0.28)

None

0.07

2.42

0.01

0.64

0.18

×

1960– 2002

Aluminium,  high grade

0.29*** (3.42)

−0.0094*** (−3.22)

−0.80*** (−4.47)

0.28** (1.74)

None

0.31

0.04

0.82

6.01**

0.34

−1.18

1960– 2002

Tin, LME

0.05 (1.32)

−0.005** (−2.08)

−0.10 (−1.12)

−0.14 (−0.91)

None

0.08

0.06

0.013

14.24***

0.77

ק

1960– 2002

Tin, Malaysia

0.04 (0.55)

−0.006 (1.85)

−0.14 (1.28)

−0.01 (−0.07)

None

0.05

0.97

1.26

70.02***

3.18

ק

1960– 2002

Gold

0.58*** (4.04)

−0.002 (−0.98)

−0.29*** (−4.49)

0.10 (0.78)

D80

0.58

1.49

0.34

2.17

0.43

×

1970– 2002

Silver

0.26 (1.65)

−0.011** (−2.0)

−0.29** (−2.59)

0.20 (1.13)

None

0.13

0.03

0.89

6.07**

0.04

−3.96

1970– 2002

Zinc, Special

0.06 (0.93)

−0.0022 (−0.87)

−0.54*** (−4.10)

0.35*** (2.32)

None

0.26

0.15

0.09

11.06***

0.51

×

1960– 2002

Zinc,  Prime Western

0.0077 (0.16)

−0.002 (−1.06)

−0.60** (−3.94)

0.28* (1.81)

None

0.25

0.88

0.24

4.30

0.56

×

1960– 2002

Note: Zinc, Special: The use of dummy for 1973 to control non‐normality of errors did not make the coefficient on the trend variable significant.

(p.55)

(p.56)

Appendix 3.4: Estimated Trends in Relative Prices of 17 Products in Minerals, Ores, and Metals (With UNCTAD Data 1960–2002)

ΔlnRPt

Constant

T

lnRPt−1

ΔlnRPt−1

Δln RPt−2

Dummies

Adjusted R2

Serial Correlation

Functional Form

Normality

Heterosced‐ asticity

Trend rate (per cent)

Sample

Phosphate rock

1.04***(9.77)

0.0003(0.02)

−0.27***(−4.61)

−0.27***(−4.61)

D74

0.79

0.05

0.008

2.98

0.42

×

1960–2002

Manganese ore

0.11***(2.51)

−0.002*(−1.66)

−0.41***(−4.91)

0.66***(5.52)

None

0.49

1.28

2.40

1.72

2.09

−0.63

1960–2002

Iron ore

0.64(5.84)

−0.003*(−1.76)

−0.33**(−3.52)

0.26**(2.26)

D75, D82

0.60

0.73

1.44

2.23

1.27

−0.83

1960–2002

Tungsten ore

0.19(1.80)

−0.009**(−2.14)

−0.18(−2.38)

0.23(1.46)

None

0.10

0.35

0.24

1.15

0.16

−4.96

1960–2002

Copper, Grade A

0.36***(3.17)

−0.011**(−3.07)

−0.43**(−3.47)

0.16(1.04)

None

0.19

0.17

5.34

0.57

2.02

−2.60

1960–2002

Copper, Wire Brass

0.23**(2.83)

−0.007**(2.76)

−0.40*(−3.15)

0.12(0.77)

None

0.15

0.29

0.40

1.03

0.16

−1.77

1960–2002

Nickel, LME

0.99***(4.96)

−0.007*(−1.71)

−0.58*(−3.40)

0.22(1.51)

D88

0.58

1.11

0.001

2.37

0.45

−1.30

1970–2002

Nickel, Cathodes

0.84***(5.04)

−0.003(−1.56)

−0.53***(−4.07)

0.22*(1.78)

D88

0.56

0.76

0.002

1.76

0.50

×

1960–2002

Lead, LME

0.33***(2.93)

−0.009**(−2.74)

−0.42*(−3.27)

0.13(0.85)

None

0.17

3.69

0.54

4.11

0.74

−2.29

1960–2002

Refined Lead

0.1977(1.90)

−0.002(−0.99)

−0.31(−2.32)

0.04(0.28)

None

0.07

2.42

0.01

0.64

0.18

×

1960–2002

Aluminium , high grade

0.29***(3.42)

−0.0094***(−3.22)

−0.80***(−4.47)

0.28**(1.74)

None

0.31

0.04

0.82

6.01**

0.34

−1.18

1960–2002

Tin, LME

0.05(1.32)

−0.005**(−2.08)

−0.10(−1.12)

−0.14(−0.91)

None

0.08

0.06

0.013

14.24***

0.77

ק

1960–2002

Tin, Malaysia

0.04(0.55)

−0.006(1.85)

−0.14(1.28)

−0.01(−0.07)

None

0.05

0.97

1.26

70.02***

3.18

ק

1960–2002

Gold

0.58***(4.04)

−0.002(−0.98)

−0.29***(−4.49)

0.10(0.78)

D80

0.58

1.49

0.34

2.17

0.43

×

1970–2002

Silver

0.26(1.65)

−0.011**(−2.0)

−0.29**(−2.59)

0.20(1.13)

None

0.13

0.03

0.89

6.07**

0.04

−3.96

1970–2002

Zinc, Special

0.06(0.93)

−0.0022(−0.87)

−0.54***(−4.10)

0.35***(2.32)

None

0.26

0.15

0.09

11.06***

0.51

×

1960–2002

Zinc, Prime Western

0.0077(0.16)

−0.002(−1.06)

−0.60**(−3.94)

0.28*(1.81)

None

0.25

0.88

0.24

4.30

0.56

×

1960–2002

Note: Zinc, Special: The use of Dummy for 1973 to control of non‐normality of errors did not make the coefficient on the trend variable significant.

(p.57)

Appendix 3.5. Description for Food‐Commodities used from UNCTAD Commodity Price Bulletin

Name

Product Description

Coffee‐1

Coffee, Brazilian and other natural Arabicas, ex‐dock NY (ȼ/lb.)

Coffee‐2

Coffee, other mild Arabicas, ex‐dock NY (ȼ;/lb.)

Coffee‐3

Coffee, Robustas, ex‐dock NY (ȼ/lb.)

Coffee‐4

Coffee, composite indicator price 1976 (ȼ/lb.)

Cocoa

Cocoa, average daily prices NY/London (ȼ/lb.)

Tea

All teas, London auction prices

Wheat, Argentina

Wheat, Argentina, Trigo Pan Upriver, f.o.b.

Wheat, US

Wheat, US, n° 2, Hard Red Winter, f.o.b. Gulf ports

Maize

Maize, Argentina, c.i.f. Rotterdam

Yellow Maize

Yellow maize, n° 3, US, c.i.f. Rotterdam

Rice

White milled rice, 5% broken, Thailand, f.o.b. Bangkok

Sugar

Sugar in bulk, Caribbean ports, f.o.b. (I.S.A.) (ȼ/lb.)

Beef

Frozen and boneless beef (mainly Australia), US ports (ȼ/lb.)

Bananas

Fresh bananas, Central America and Ecuador, f.o.b. US ports (ȼ/lb.)

Pepper

White pepper, 100% Sarawak, Singapore, closing quotations

Appendix 3.6. Description of Vegetable Oils and Oilseeds used from UNCTAD Commodity Price Bulletin

Name

Product Description

Soybean Meal

Soybean meal 44/45%, Hamburg f.o.b. ex‐mill

Fish Meal

Fish meal 64/65%, any origin, candf, Hamburg

Yellow Soybean

Yellow soybeans, n° 2, US, c.i.f. Rotterdam

Crude Soybean Oil

Crude soybean oil, Dutch, f.o.b. ex‐mill

Sunflower Oil

Sunflower oil, E.U., f.o.b. N.W. European ports

Ground Nut Oil

Groundnut oil, any origin, c.i.f. Rotterdam

Cora in bulk

Copra in bulk, Philippines/Indonesia, c.i.f. European ports

Coconut Oil

Coconut oil, Philippines/Indonesia c.i.f. N.W. European ports

Palm Kernel Oil

Palm kernel oil, Malaysia, c.i.f. Rotterdam

Palm Oil

Palm oil, 5% ffa, Indonesia/Malaysia, c.i.f., N.W European ports

Cottonseed Oil

Cottonseed oil, PBSY, US, f.o.b. Gulf ports

Appendix 3.7. Description of Agricultural Raw Materials used from UNCTAD Commodity Price Bulletin

Commodity Name

Product Description

Cotton, US, Memphis

Cotton, US Memphis/Eastern, Midd.1–3/32″, c.i.f. (ȼ/lb.)

Cotton, US, Orleans

Cotton, US Orleans/Texas, Midd.1‐1/32″, c.i.f. (ȼ/lb.)

Cotton Outlook Index A

Cotton Outlook Index A (M 1–3/32″) (ȼ/lb.)

Cotton Outlook Index B

Cotton Outlook Index B (coarse count) (ȼ/lb.)

Jute

Jute BWD, Bangladesh, f.o.b. Mongla

Sisal, Tanzania/Kenya

Sisal, n° 3, long, Tanzania/Kenya, c.i.f. London

Sisal, Uganda

Sisal UG, East Africa, c.i.f. London

Non‐coniferous Wood

Non‐coniferous woods, UK Import price index ($ equivalent) [1995=100]

Tropical Logs

Tropical logs, Sapelli LM, UK import price, f.o.b. ($/m3)

Tropical Logs, Gabon

Tropical logs, Okoume, LM, f.o.b. Gabon ($/m3)

Sawn Wood

Sawn wood, Dark Red Meranti, Malaysia, select and better, c.i.f. French ports ($/m3)

Plywood, sheet

Plywood, S.E. Asian Lauan, 4 mm, wholesale price, Tokyo (ȼ/sheet)

Plywood, metre

Plywood, S.E. Asian Lauan, 4 mm, wholesale price, Tokyo ($/m3)

Linseed Oil

Linseed oil, any origin, ex‐tank, Rotterdam

Leaf Tobacco

Leaf tobacco, US import unit value

Hides

Cattle hides, suspension dried, 8/12 lb. Tanzania ($/100 kg)

Rubber in Bale

Rubber in bales, Singapore n°1 RSS, f.o.b.

(p.58)

Appendix 3.8. Description of Minerals, Ores, and Metals used from UNCTAD Commodity Price Bulletin

Name

Product Description

Phosphate rock

Phosphate rock, 70% BPL, Khouribga, f.a.s. Casablanca

Manganese ore

Manganese ore, 48/50% Mn, c.i.f. Europe

Iron ore

Iron ore, Brazilian to Europe, 64.5% Fe, f.o.b. (ȼ/Fe unit)

Tungsten ore

Tungsten ore, Wo3 > 65%, c.i.f. UK ($/t.Wo3)

Aluminium

Aluminium high grade, LME, cash

Copper

Copper, grade A, LME, cash

Copper wire bars

Copper, wire bars, US producer, f.o.b. refinery (ȼ/lb.)

Nickel, LME

Nickel, LME, cash

Nickel, cathodes

Nickel cathodes, New York dealer (ȼ/lb.)

Lead, LME

Lead, LME, cash settlement ($/t)

Refined Lead

Refined lead, North America producer price (ȼ/lb.)

Zinc, Special high grade

Zinc, special high grade, LME, cash settlement

Zinc, Prime Western

Zinc, Prime Western, delivered, North America (ȼ/lb.)

Tin, LME

Tin, LME, cash

Tin, Malaysia

Tin, ex‐smelter price, Kuala Lumpur

Gold

Gold, 99.5% fine, afternoon fixing London ($/troy ounce)

Silver

Silver, 99.9%, Handy and Harman, New York (ȼ/troy ounce)

Long‐Run Trend in the Relative Price: Empirical Estimation for Individual Commodities

Appendix 3.9. MUV of Grilli-Yang Dataset and UNCTAD Unit Value of Exports of Manufactured Goods from Developed Market Economy Countries

(p.59)

Long‐Run Trend in the Relative Price: Empirical Estimation for Individual Commodities

Appendix 3.10. Real Prices of 23 Food and Beverages Items

(p.60)

Long‐Run Trend in the Relative Price: Empirical Estimation for Individual Commodities

Appendix 3.11. Real Prices of 17 Agricultural Raw Materials

(p.61)

Long‐Run Trend in the Relative Price: Empirical Estimation for Individual Commodities

Appendix 3.12. Real Prices of 17 Agricultural Raw Materials

(p.62)

Long‐Run Trend in the Relative Price: Empirical Estimation for Individual Commodities

Appendix 3.13. Real Price of Four Different Types of Coffee

Long‐Run Trend in the Relative Price: Empirical Estimation for Individual Commodities

Appendix 3.14. Real Price of Two Types of Wheat

Long‐Run Trend in the Relative Price: Empirical Estimation for Individual Commodities

Appendix 3.15. Real Price of Crude Soybean, Sunflower, and Groundnut Oil

(p.63)

Appendix 3.16. Data Set for 13 Commodity Prices (the Updated Grilli‐Yang Series: 1900–2001)

Year

Cocoa

Coffee

Tea

Bananas

Sugar

Rice

Wheat

Maize

Cotton

Jute

Palm oil

Copper

Tin

MUV

1900

8.990

4.594

18.814

13.010

33.190

22.771

20.383

16.770

16.990

13.900

16.580

21.724

4.780

14.607

1901

8.460

3.598

17.994

13.470

26.857

20.467

19.533

21.465

17.090

12.460

15.700

21.616

2.676

13.858

1902

8.510

3.044

19.067

13.940

21.462

19.789

22.364

27.628

17.650

11.970

16.570

14.800

4.283

13.483

1903

8.560

3.100

14.521

14.430

22.986

25.035

22.364

19.705

21.520

13.110

16.860

17.765

4.491

13.483

1904

8.780

4.317

17.425

14.940

30.609

20.877

26.044

21.130

20.670

13.610

16.570

17.202

4.475

13.858

1905

8.620

4.594

16.920

15.470

32.720

23.488

25.478

22.346

19.260

17.960

16.280

20.919

5.014

13.858

1906

8.830

4.428

14.837

16.010

24.863

27.608

21.515

19.663

21.240

22.780

18.310

25.870

6.367

14.607

1907

10.710

3.598

14.521

16.580

26.153

29.637

24.912

22.262

21.520

20.390

19.770

26.836

6.103

15.356

1908

7.440

4.594

13.448

17.164

30.023

30.746

29.441

29.054

20.580

15.060

16.570

17.725

4.710

14.232

1909

6.370

4.871

14.710

16.606

29.554

24.195

30.857

28.341

23.510

12.650

17.440

17.417

4.752

14.232

1910

5.990

5.757

15.152

17.013

31.900

24.153

27.177

23.813

27.750

14.380

21.220

17.095

5.455

14.232

1911

6.320

7.804

15.279

18.100

35.418

31.609

26.894

25.490

24.170

19.710

20.930

16.611

6.759

14.232

1912

6.690

8.856

15.468

18.889

30.609

36.468

27.743

28.665

21.900

20.410

20.060

21.925

7.371

14.607

1913

7.440

7.196

15.658

18.635

22.869

28.552

22.647

25.990

24.070

25.750

21.510

20.489

7.075

14.607

1914

6.640

6.365

15.658

18.965

30.961

26.907

28.309

29.472

20.860

26.660

23.550

18.248

5.484

13.858

1915

8.940

5.314

15.152

18.504

38.819

27.253

35.953

30.916

19.730

20.030

24.710

23.186

6.170

14.232

1916

7.600

5.867

15.152

19.207

51.250

25.154

39.067

34.823

29.550

29.480

35.470

36.497

6.952

17.603

1917

5.990

5.646

19.319

22.065

54.182

21.707

62.280

70.495

46.160

37.460

52.030

36.470

9.881

20.974

1918

6.900

7.030

22.602

26.692

49.726

24.719

61.714

68.457

57.210

37.370

97.380

33.049

14.189

25.468

1919

9.960

13.727

21.693

24.092

59.343

26.553

61.714

67.692

63.530

44.240

52.330

25.078

10.125

25.966

1920

7.280

10.572

17.381

27.067

140.150

31.241

65.960

60.345

50.030

32.760

36.630

23.428

7.717

28.839

1921

4.170

5.757

15.092

24.924

36.356

35.767

41.331

23.994

32.850

21.200

20.350

16.773

4.784

24.439

1922

4.920

7.915

21.149

23.673

32.838

39.560

34.537

26.287

41.440

27.220

21.510

17.953

5.204

21.723

1923

4.070

8.192

27.052

25.037

58.991

37.559

30.008

34.738

52.770

23.730

22.090

19.349

6.821

21.723

1924

4.070

11.790

27.656

25.974

44.800

40.847

35.669

40.726

51.260

28.030

23.840

17.470

8.023

21.723

1925

5.080

13.616

27.486

29.839

26.270

41.170

46.427

43.783

41.440

47.890

27.030

18.839

9.256

22.097

1926

6.160

12.343

29.650

31.406

26.036

44.269

42.464

31.765

32.190

42.560

25.000

18.517

10.437

20.974

1927

8.460

10.351

29.206

31.319

30.961

41.005

41.897

36.649

32.190

31.820

23.260

17.336

10.288

19.850

1928

6.850

12.841

25.698

30.676

25.567

37.098

38.217

41.448

36.250

32.820

23.550

19.550

8.063

19.850

1929

5.570

12.232

25.034

30.858

20.172

37.027

37.934

39.622

32.570

31.020

23.840

24.300

7.220

19.101

1930

4.390

7.140

23.370

30.926

14.425

29.171

26.894

34.823

23.980

19.400

18.900

17.417

5.067

18.727

1931

2.780

4.871

17.496

29.376

13.018

16.366

16.136

21.955

14.630

14.390

13.950

10.895

3.912

15.356

1932

2.360

5.867

10.422

27.847

8.327

14.392

13.588

12.952

12.370

11.260

11.050

7.460

3.521

12.734

1933

2.360

5.037

15.675

28.439

11.367

12.232

15.853

16.902

16.990

12.600

11.050

9.419

6.253

14.232

1934

2.780

6.144

21.194

28.188

13.956

15.912

21.515

27.561

21.900

14.180

15.700

11.311

8.344

16.854

1935

2.680

4.926

19.996

28.410

18.530

20.785

24.063

34.526

22.660

16.530

22.380

11.607

8.061

16.479

1936

3.640

5.258

20.592

27.423

20.289

20.405

26.611

35.502

22.940

17.460

22.670

12.707

7.425

16.479

1937

4.500

6.089

23.764

26.327

20.641

21.633

37.934

43.741

20.200

19.430

25.000

17.672

8.688

16.854

1938

2.780

4.262

22.264

27.103

17.005

19.850

28.309

23.144

16.340

17.270

19.700

13.418

6.763

17.603

1939

2.570

4.096

19.355

28.385

17.709

19.208

17.552

21.233

17.560

23.540

20.350

14.706

8.045

16.105

1940

2.730

3.930

19.770

30.986

15.950

23.088

18.684

24.461

19.540

22.360

20.930

15.162

7.967

17.603

1941

4.070

6.255

25.300

32.080

19.820

25.814

18.684

29.939

27.090

21.230

28.200

15.833

8.317

18.727

1942

4.760

7.417

32.973

33.169

29.671

28.353

20.949

35.375

35.020

19.940

34.590

15.806

8.314

21.723

1943

4.760

7.417

29.655

34.676

28.616

29.628

34.254

43.868

36.630

26.620

25.000

15.806

8.314

24.345

1944

4.760

7.417

28.964

37.051

28.968

29.628

36.519

48.115

37.950

32.410

25.000

15.806

8.314

27.715

1945

4.760

7.528

28.964

38.652

34.480

29.628

39.350

49.559

42.480

31.600

25.000

15.806

8.314

28.464

1946

6.160

10.240

27.443

43.921

41.634

34.177

64.545

69.306

54.470

39.560

25.000

18.544

8.720

28.839

1947

18.680

14.779

38.020

45.998

56.293

54.719

75.019

87.524

62.020

64.330

51.740

28.124

12.463

34.831

1948

21.250

15.000

40.439

46.981

49.608

57.992

67.376

40.738

58.910

77.920

63.370

29.573

15.868

35.581

1949

11.560

18.103

43.072

52.050

48.788

50.082

56.028

60.771

56.920

56.710

46.800

25.763

15.883

33.333

1950

17.180

25.862

38.370

54.383

58.404

42.235

49.645

65.586

67.680

62.800

42.440

28.500

15.275

30.337

1951

19.000

30.517

48.151

54.383

66.496

44.614

57.447

69.342

75.330

93.500

67.150

32.472

20.318

35.955

1952

18.950

29.483

40.157

55.126

48.905

48.290

60.283

60.289

66.930

56.020

40.120

32.472

19.261

36.704

1953

19.860

29.483

48.057

55.126

39.992

53.975

54.610

57.978

60.890

44.480

35.470

38.644

15.323

35.206

1954

30.940

39.828

69.405

56.647

38.233

48.785

47.518

56.148

60.420

52.240

36.340

39.838

14.683

34.457

1955

20.070

31.034

65.831

55.903

37.998

43.718

45.886

46.998

60.510

47.330

37.790

50.304

15.147

34.831

1956

14.610

35.690

63.668

56.647

40.813

42.297

45.886

49.695

58.620

49.750

43.600

56.114

16.214

36.330

1957

16.380

35.172

58.401

59.621

60.515

42.389

44.539

45.843

57.110

59.510

44.190

39.690

15.390

36.704

1958

23.710

25.862

60.470

55.126

41.047

43.965

43.759

45.843

57.870

52.010

41.860

34.565

15.209

36.330

1959

19.590

22.241

60.000

49.177

34.831

40.844

45.319

44.398

56.070

51.290

42.440

41.837

16.316

36.330

1960

15.200

21.207

60.659

48.434

36.825

38.527

44.539

41.702

53.620

74.580

41.280

43.005

16.218

37.079

1961

12.100

19.655

58.025

46.947

34.128

42.173

45.106

44.206

56.260

91.500

42.440

40.147

18.116

37.453

1962

11.240

18.621

58.684

44.716

34.949

47.209

46.879

49.503

59.190

62.070

40.410

41.059

18.330

37.453

1963

13.540

18.103

55.674

56.674

99.686

44.274

47.447

52.681

58.720

61.050

40.700

41.059

18.650

37.453

1964

12.530

24.310

56.521

57.391

68.847

42.544

49.787

53.740

55.980

66.860

41.280

42.884

25.197

38.202

1965

9.260

23.793

55.110

53.639

24.863

42.111

46.666

52.970

52.490

71.200

45.930

46.990

28.491

38.951

1966

13.060

21.724

53.605

52.152

21.814

50.422

49.787

57.207

43.800

79.230

43.900

48.533

26.232

39.700

1967

15.580

20.172

54.075

53.639

23.338

63.587

49.220

48.058

36.250

56.560

43.020

51.297

24.531

39.700

1968

18.410

20.172

44.577

51.408

23.221

62.286

47.163

47.287

35.680

61.850

42.150

56.154

23.686

39.326

1969

24.460

20.690

41.473

53.639

39.523

57.744

45.603

51.910

35.400

69.160

36.340

63.776

26.300

40.449

1970

18.310

26.897

46.740

55.903

43.979

44.490

44.822

56.244

37.850

66.340

46.220

77.422

27.853

42.697

1971

14.340

23.276

44.953

47.690

53.010

39.856

45.319

56.244

48.050

72.750

43.900

69.009

26.756

45.318

1972

17.290

25.862

44.859

54.383

87.137

45.448

50.567

53.933

49.273

69.430

34.880

67.922

28.374

48.689

1973

34.470

32.069

45.141

55.734

112.940

108.140

104.400

94.382

84.825

67.170

48.260

78.992

36.376

58.801

1974

52.510

34.138

59.812

62.224

351.360

167.460

148.010

127.130

88.567

83.720

77.910

102.850

63.356

71.161

1975

40.040

33.621

59.060

83.382

240.420

112.180

128.580

115.180

72.351

83.140

75.580

85.258

54.331

79.026

1976

58.500

73.965

65.549

86.965

135.810

78.630

105.740

108.250

105.408

70.640

59.300

92.343

60.726

78.652

1977

108.390

124.660

114.640

92.846

95.229

84.099

82.127

91.782

96.676

89.550

88.370

88.304

85.472

86.517

1978

97.410

85.862

93.386

97.104

91.477

113.540

95.602

96.983

97.923

104.740

101.450

87.901

100.660

98.876

1979

94.200

89.483

91.975

110.050

113.290

102.360

122.270

111.240

105.408

105.710

110.170

123.790

113.870

114.610

1980

74.400

80.690

95.173

128.100

336.240

134.060

135.320

120.670

127.861

98.370

99.130

137.370

135.260

125.470

1981

59.410

66.207

86.050

135.640

198.900

149.230

139.290

125.970

115.387

88.520

95.640

112.380

115.510

119.100

1982

49.600

72.419

82.447

126.510

99.053

90.532

118.040

105.280

99.794

82.120

73.550

97.791

104.530

115.730

1983

60.570

68.048

99.339

144.930

99.534

85.596

120.220

130.970

115.387

86.140

87.210

104.440

103.070

110.490

1984

70.110

74.550

147.520

125.000

61.310

77.930

117.310

130.870

111.645

157.370

126.900

89.580

99.730

108.610

1985

65.660

75.530

84.420

128.378

48.090

66.740

104.350

108.050

82.330

173.060

87.210

87.980

94.660

109.590

1986

60.980

100.520

82.290

129.050

73.030

65.540

88.238

85.710

66.114

79.080

43.520

86.670

58.560

130.300

1987

58.760

58.660

72.879

123.310

81.563

71.705

86.703

74.099

102.913

94.500

58.012

112.460

63.583

147.337

1988

44.657

72.258

67.634

161.486

118.280

91.238

111.256

102.885

87.320

100.749

73.487

167.439

79.55

156.486

1989

35.497

56.996

76.324

184.797

150.538

98.485

129.671

107.692

104.160

101.656

58.857

183.269

83.85

155.224

1990

36.355

46.980

76.702

182.770

150.538

89.262

104.350

104.808

113.516

111.186

48.767

171.300

100.67

170.999

1991

34.351

44.595

69.523

189.189

107.527

96.509

98.980

102.885

104.784

99.502

57.007

150.515

72.94

170.999

1992

31.489

33.625

75.569

159.797

107.527

88.274

115.860

100.000

79.835

76.016

66.256

146.782

65.57

176.047

1993

32.061

37.202

70.279

149.662

118.280

77.404

107.420

98.077

79.835

73.860

63.566

123.102

71.60

166.898

1994

40.077

78.935

69.145

148.311

145.161

88.274

115.092

103.846

109.774

80.509

88.790

148.456

60.62

171.315

1995

40.936

79.412

61.966

150.338

155.914

105.731

135.809

118.269

132.851

99.570

105.607

188.932

64.05

188.351

1996

41.794

64.150

62.722

158.784

139.785

111.660

159.595

159.615

110.397

123.690

89.295

147.683

72.70

182.357

1997

46.374

99.444

77.836

169.932

134.409

99.802

121.998

112.500

109.150

82.233

91.817

146.525

72.25

169.422

1998

48.092

71.066

77.458

166.216

107.527

100.132

96.678

98.077

89.815

70.546

112.838

106.435

66.17

162.796

1999

33.779

52.942

68.767

144.932

75.269

83.004

86.703

87.500

75.469

75.063

75.842

98.970

66.94

156.486

2000

26.050

45.787

71.035

143.243

96.774

66.535

87.470

85.577

81.083

75.924

52.131

116.667

63.24

148.283

2001

30.630

32.671

60.455

196.959

102.151

56.983

97.445

86.538

66.114

89.743

48.095

101.544

65.20

144.497

Note: All price data are in nominal US dollars with 1977–79 prices. MUV is the prices index of manufactured goods exported by developed countries.

(p.64) (p.65)

(p.66)

Appendix 3.17. UNCTAD Data on Commodity Prices

Year

UVXM

Composite/ Aggregate

All Food

Food and Beverages

Food Only

Vegetable oils and oilseeds

Agri raw materials

Minerals and Metals

1960

34.68208

45.20833

43.90833

30.475

49.95

43.975

54.26667

41.85

1961

35.4528

43.44167

43.10833

28.05833

49.2

46.64167

48.3

40.70833

1962

35.83815

43.25833

43.83333

27.34167

51.74167

41.75

45.50833

40.15833

1963

35.83815

52.04167

58.55833

28.36667

75.25833

43.99167

45.46667

40.19167

1964

36.41618

51.525

54.28333

33.1

65.53333

46.34167

45.725

48.675

1965

37.28324

48.36667

45.79167

30.45833

51.30833

52.61667

46.08333

56.65

1966

38.15029

49.90833

46.61667

31.05

53.23333

48.53333

46.1

61.11667

1967

38.15029

47.66667

47.475

30.43333

55.71667

44.90833

42.73333

51.75833

1968

38.15029

47.23333

45.83333

30.49167

53.84167

40.70833

42.79167

54.10833

1969

38.15029

51.8

49.875

32.36667

59.61667

41.2

45.84167

61.075

1970

39.49904

53.575

53.55833

36.70833

61.11667

53.81667

42.41667

61.75833

1971

42.00382

51.63333

53.075

32.26667

62.03333

55.20833

42.2

54.9

1972

45.37572

58.98333

64.2

36.05

80.39167

47.75

46.78333

54.60833

1973

53.46821

95.375

107.25

46.35833

138.8667

87.825

81.54167

75.21667

1974

64.45087

138.9417

169.0417

54.48333

226.5167

141.9667

84.90833

101.6167

1975

73.12139

109.175

127.0417

53.21667

168.15

90.36667

76.275

87.60833

1976

72.83237

106.3833

115.1167

96.125

130.15

84.375

97.25

90.78333

1977

80.0578

117.3583

131.4333

169.0667

119.3333

107.9417

100.875

93.50833

1978

91.6185

115.0833

125.0917

121.8917

128.1417

117.5083

107.4583

95.13333

1979

104.0462

132.7917

137.825

125.6583

144.0083

134.625

124.8583

125.725

1980

115.6069

168.5167

188.4083

117.55

235.4833

116.825

137.225

140.6083

1981

109.5376

140.4083

154.3083

96.55

189.6833

110.2583

117.95

121.35

1982

105.7803

110.6417

114.5083

91.675

130.1583

89.11667

101.4667

107.5083

1983

102.3121

117.95

122.525

96.225

137.7083

106.8333

108.3667

113.2667

1984

98.84393

113.1667

117.6333

109.7167

115.6083

144.0917

110.1417

104.0167

1985

100

100.0083

100.0083

100

100

100.0083

100

100.0083

1986

119.3642

104.0167

107.4333

124.2

109.475

61.59167

103.5167

95.65833

1987

134.9711

106.625

100.8167

80.70833

115.8167

72.6

121.0917

110.8833

1988

143.3526

134.8333

125.7333

81.725

151.8667

96.075

129.825

161.7417

1989

142.1965

135.9583

126.8

69.99167

161.2667

84.925

132.4417

161.9583

1990

156.6474

128.8083

117.3667

61.55833

151.6917

74.04167

142.3083

148.1667

1991

156.6474

119.7917

110.2667

56.85833

140.875

79.625

133.6667

133.95

1992

161.2717

115.85

106.575

48.59167

137.125

85.975

130.0167

129.175

1993

152.8902

110.3833

107.9

52.31667

137.6833

85.61667

120.8917

109.0333

1994

156.9364

130.1667

129.85

91.25

151.9917

107.4417

140.0917

123.7333

1995

172.5434

143.3583

135.9417

91.925

159.5333

118.425

161.3167

149.1667

1996

167.052

137.0667

137.425

77.525

169.625

113.05

144.525

130.7417

1997

155.2023

136.3667

140.075

103.7

162.475

111.7083

130.1083

131.4667

1998

149.1329

118.45

122.45

85.525

139.5833

120.3833

116.3167

109.8083

1999

143.3526

101.8583

98.80833

68.075

114.225

91.60833

104.2

107.9583

2000

135.8382

104.4

97.31667

58.7

120.2833

71.06667

106.3833

120.9917

2001

132.3699

101.0333

96.825

46.10833

126.3333

65.45

104.1833

109.4667

2002

129.4798

99.225

96.84167

50.125

121.0917

82.03333

96.925

107.0667

(p.67)

Notes:

(1) For example, how to choose a break point in unit root testing procedure or how many breaks are to be considered.

(2) This result is due to Engle and Granger (1987). Therefore, Bleaney and Greenaway (1993) suggest that equation (4) should be considered as an alternative to test for cointegration between primary product and industrial goods prices, as is done in Powell (1991) and noted in the previous section.

(3) For a series that is a random walk with zero mean, its history gives no indication of its future path. In future its value can be greater or less than its current value. For a random walk with drift, if the estimated b is positive, it is more probable that it will be greater than its current value in the future and the opposite is true if b turns out to be negative.

(4) Pesaran et al. give both the critical values for Wald and F‐statistics. In this paper we will only consider the F‐statistics.

(5) In equation (5) πi and δi give the short‐run estimates of the parameters. The long‐run parameter values can be obtained by noting that there is no change in the steady state such that: ΔXt = ΔZt = 0. This would imply the long‐run coefficient on Z as: [ ξ γ ] .

(6) This, in effect, implies that in the absence of any other explanatory variables (apart from the constant and trend term) the statistical significance of the lagged level dependent variable is to be considered as evidence for a valid long‐run relationship irrespective of the unit root property of the data.

(7) A general practice in the case of annual data is to include at least one lag of the dependent variable and then to check for the residual autocorrelation problem. For quarterly data at least four lags are used.

(8) We thank Angus Deaton for providing us with the Grilli‐Yang data on commodity prices for these 13 commodities. From an e‐mail communication, it was learnt that the World Bank no longer has access to the information on the individual commodities price series used in the Grilli‐Yang study.

(9) The commodities for which information could not be obtained were aluminium, beef, hides, lamb, lead, rubber, silver, timber, tobacco, wool and zinc.

(10) Apart from jute, price series for commodities were updated using the information in various issues of Global Economic Prospects, published by the World Bank. Price data in the International Financial Statistics Yearbook of the IMF were used to build the series on jute for 1987–2001.

(11) Note that Grilli and Yang (1988) used MUV as the deflator. For later periods we use what UNCTAD now publishes as the unit value index of manufactured goods exports from the developed market economy countries. Appendix 1 gives the graphical plots of these two series, which show that the series are almost the same. A linear trend line fitted through the scatter of the two series resulted in a R2 value of 0.999 with the coefficient on the explanatory variable very close to one (the restriction that the coefficient was exactly one could not be rejected at the 1 per cent error probability level).

(12) These data were accessed from the Commodity Price Bulletin of UNCTAD. For this study the online version of the dataset was used from the website: www.unctad.org

(13) Note that since there is no pre‐testing for unit root, it is not known a priori whether any of these five series are TSP. In some studies, when pre‐testing for unit roots is not done, the standard T‐ratios are used to make inferences about the statistical significance of the lagged dependent variable (e.g. Athukorala, 2000).

(14) As mentioned above, if the statistical significance of ln RPt − 1 is to be determined on the basis of the Dickey‐Fuller critical values, then the variable is significant only in the case of copper.

(15) Otherwise, the relative price series form a random walk with zero mean.

(16) The equation for food and beverages did not show any residual non‐normality problem and therefore no dummy variable was inserted to control for the sharp rise in 1973, as shown in Figure 3.2.

(17) Among the 17 commodities, four types of coffee and two types of wheat are included. Appendices 3.13 and 3.14 show that prices of different varieties of the two products move quite closely.

(18) In the case of different varieties, a simple average of the estimated growth rates has been used.

(19) For tea, the long‐term trend growth over the period of 1900–2001 was found to be −1.25 per cent per annum, while for coffee no significant rate could be found.

(20) Moreover, in the 1980s commodities prices were already very low. Maizels (1992) shows that relative prices in the 1980s were lower than those during the great depression of the 1930s. Bleaney and Greenaway (1993) report a 37 per cent downward jump in commodity prices after 1980 compared to the average for 1925–1991.

(21) In the previous section it was found that using the aggregate relative price of primary commodities, Cuddington and Urzua (1989), Sapsford (1985) and Powell (1991) found structural breaks in different years.