X. Presentación de hallazgos
10.1. Existencias de producto acumulado
Given that academic literature in this field attempts to isolate a single trend over a long period of time, the scope of each study with respect to the time period concerned and the metal markets covered may influence results greatly. The time period chosen has an obvious influence on the results. The time around 1975 is an often declared "structural break" in many studies. The notion of falling metal prices in real terms is very much influenced by a perceived persistent decline in metal prices since the end of the 1970s. Furthermore, for lack of accounting for spurious correlation, some analyses based on statistical methodology rather than solid industry knowledge may reach different conclusions than others.78 Finally, the outcome of the analysis depends very much on the deflator used to discount nominal prices.79 Numerous empirical studies on the long-term trend of commodity prices have been conducted and it is beyond the scope of this work to review them all. Representative studies, which for the most part focus on metal prices are listed in Table 2:
76 Humpreys (2009), p. 103 77 Humpreys (2009), p. 104 78 Ahrens/Sharma (1997), p.59 79 Svedberg/Tilton (2006), p.501f.
Author Findings Metal sector (s) Time period Potter/ Christy
(1962)
Falling trend in mineral commodity index comprising 14 metals, 4 energy
commodities, 14 nonmetal prices
Drivers for trend not specified
Index of iron
Price series analysis insufficient for price
Slade (1982) Prices follow u-shaped price curve, first falling, then rising
Copper,
Evidence that real prices of metals are trendless
No particular trend in metal prices can be generalized
Copper,
Author Findings Metal sector (s) Time period Krautkraemer
(1998)
Falling trend in metal real prices
Copper, aluminum, lead, nickel, tin, zinc, silver
1967-1995
Cashin/
McDermott/ Scott (2002)
Declining price trend
No evidence for break in this trend
Industrial commo-dity index
1982-1999
Svedberg/ Tilton (2006)
Trend in prices depends on adjustment of inflation deflator
Copper 1870-2000
Table 2 – Studies on long-term trends in metal price series Source: Own illustration
As summarized in Exhibit 10, Table 2 indicates the range of conclusions reached by authors analyzing the long-term development of metal prices. Rather than reviewing all studies, a summary is provided to distill relevant opinions and patterns.
Several theoretical models have predicted a rise in metal prices due to growing scarcity for centuries.80 Later studies, initiated by a first systematical approach conducted by POTTER/CHRISTY found that a mineral price index comprising fourteen metals, four energy commodities, and fourteen nonmetals, fell by 40% in real terms during the observed period.81 POTTER/CHRISTY's systematic analysis triggered a range of scientific studies, which identify two general opposing market forces influencing metal prices:
Increase of extraction costs due to declining ore quality
Declining cost of production due to technical innovation
The perceived interaction of these two forces is a strong determinant of the predicted price trend. In this context, a study by SLADE received considerable attention. Based on an analysis of the six non-ferrous base metals and silver and a model that uses as variables exogenous technical change and endogenous change in the grade of ores mined, she
80 For a review of the earlier evolution of such concerns, compare Tilton(2003), p. 7ff.
81 Potter/Christy (1962)
concludes that there is "a U-shaped time path for relative prices"82. This trend is driven by
"the historic counterbalancing influences of improvements in technology and deterioration in ore quality in determining production cost"83. For an industry example, she points to the copper industry in the US, where copper ore graded declined from 1900 to 1980 from 5% to 0.7%. Despite such fall in ore quality, real copper prices fell until 1940 due to innovations in mining technology and equipment. As since then prices have been on the rise, she reasons that technological innovation is saturated and that declining ore grade continues to dominate production costs.
Several authors object her findings. For one, it is pointed out that despite warnings of growing scarcity and rising cost of depletion, prices continued to fall after the period examined by SLADE. KRAUTKRAEMER observes that "nonrenewable resource prices did not continue to trend upward after the 1970s"84, thus rejecting SLADE's findings. He concludes that "there isn't a stable linear trend to most resource price time series"85. As a major reason for the lack of an upward trend in prices he points to technological innovation overcompensating the cost increases of declining ore grade. The same argument is repeated by GOMEZ et al. who conclude that the decline in the real price of copper is "a decline largely driven by the highly successful efforts of primary copper producers to reduce their production costs over this period"86. TILTON/LAGOS go even further and suggest that the ability of technical innovation to reduce extraction cost may go on indefinitely: "the challenge for innovation and new technology in keeping the cost-increasing effects of depletion at bay may be no greater in the future than it has been in the past"87.
Other authors generally oppose the idea of identifying any trend on the basis of mere price analysis. Refining POTTER/CHRISTY's analysis, SMITH concludes that their and others findings that metal prices are following a certain trend is premature. He emphasizes that drawing conclusions on price series alone is unwise as a detailed knowledge of the underlying market structure and its changes is inevitable: "evaluations of resource scarcity without
82 Slade (1982), p.126 83 Slade (1986), p.126
84 Krautkraemer (1998), p.2079 85 Ibid.
86 Gómez/Guzmán/Tilton (2007), p.189 87 Tilton/Lagos (2007), p.22
detailed analysis of the character of the markets for the specific commodities within each aggregate, as well as the institutional changes during the period, do not seem possible"88. 3.1.3 Price volatility
Academic literature on commodity price volatility focuses predominantly on two main topics influencing the level of volatility:
the impact of market organization
exchange prices versus administered producer prices
trading of futures and forwards
the impact of supplier concentration
An overview of studies focusing on metal price volatility is presented in Table 3:
Sub-category
Author Findings Metal sector
(s)
Time period Slade (1991) Volatility increased due
to structural change in market organization
Volatility is stationary
Speculation has only short-term influence
Medium-term influence on volatility stems from physical factors
No evidence for increase in volatility
88 Smith (1979), p.426
Sub-category
Author Findings Metal sector
(s)
Decomposition of volatility in long and short-term common factors
Long-term factors found to be influenced by other metal prices
Aluminum,
copper, lead, nickel, tin, zinc
1972-2000
Slade (2006) Volatility and volume forward trading positively correlated
No direct influence, linked by common factor
Aluminum,
Future trading first decreasing, then increasing effect on volatility
Not applicable
Carlton (1986) Higher level of industry concentration correlates
Higher level of industry concentration correlates
Table 3 – Studies on volatility in metal price series Source: Own illustration
Market organization
In a much recognized study on the impact of market organization, SLADE identifies five distinct areas from which price volatility originates:89
Horizontal market structure, i.e. the concentration of the supply side industry90
Marketing method, including the impact of market organization, more specifically price setting mechanisms in metal markets, namely so-called producer price setting, where prices are set by an oligopoly of major producers and exchange-price setting.
This in turn influences the motivation of buyers, i.e. if buyers are solely consumers or consumers, hedgers, and speculators
Demand factors, comprising the stability of output of the consuming sector and the ease of substitution
Supply factors, including the influence of cost stability, by-production, recycling on price stability
Time-period factors, i.e. exchange rate and rate of inflation
Of these five factors she determines two, horizontal market structure and marketing method to be significant, i.e. measurable "with statistical accuracy".91 Of these two, her analysis of the time period from 1970 to 1986 reveals that the structural change in the marketing method appears most influential: "The increase in metal-price instability that has occurred in the last decade is entirely explained by changes in underlying market-structure and organization variables considered here. Foremost among these is increased reliance on commodity exchanges, which accounts for a significant fraction of the systematic variation across markets."92
Subsequent studies question both the finding that volatility has increased and the conclusion that a structural change in market organization is to blame. In their research on the price volatility of aluminum, copper, lead, nickel, tin and zinc between 1972 and 1995, using daily prices, BRUNETTI/GILBERT find no evidence that volatility has increased over time. On the contrary they observe that "except in the case of tin, volatility levels were beneath their
89 Slade (1991), p.1311ff.
90 Slade does not specifically name the supply side when speaking of horizontal concentration. From the data she uses for her analysis, however, it becomes clear which industry she means.
91 Slade (1991), p.1311. It is unclear on what basis she excludes the other three.
92 Slade (1991), p.1337
historic average levels over 1993-95, a period of increased speculative interest in the metals markets"93. However, they find that volatility is itself volatile and distinguish three explanations for changes in metal market volatility:94
Information considerations: price adjustments on the basis of new information
Hedging or speculative pressure
Physical availability
They argue that while informational considerations and speculative pressure do have an impact, theirs appears to be only short-term. They conclude that "much of the medium-term variability in the volatility of non-ferrous metals prices on the LME may be explained by physical (or fundamental) factors"95. Without elaborating what these fundamentals are specifically, they observe that a tight supply market with low stocks and little leeway to respond to demand or supply shocks is more volatile than a market with lower demand and large stocks.
Building on the findings of BRUNETTI/GILBERT, MCMILLAN/SPEIGHT decompose the volatility of six metals in common short- and long-term factors in an attempt to further quantify their impact. Yet they link the price volatility of metals to one another rather than attempting to develop a deeper understanding of the drivers for this proposed connection or generally of the impact of underlying industry factors on volatility. 96
FIGUEROLA-FERRETTI/GILBERT extend the time period used in SLADE's analysis from 1970-1997. They divide this period into four sub-periods based on SLADE's example and visual inspection of data. Using similar statistical methods and monthly data they concur with SLADE that volatility till 1986 increased, but point out that silver was a dominant driver for this and that once silver is excluded, evidence for an increased volatility is much weaker.
Extending the time period to 1997, they refrain from observing a "general tendency for the variability of exchange prices to increase over time"97, thus rejecting SLADE's notion that a structural change in the market organization of metal markets had any influence on price volatility.
93 Brunetti/ Gilbert (1995) , p.237
94 Derived from Brunetti/ Gilbert (1995), p.244 95 Brunetti/ Gilbert (1995), p.245
96 McMillan/Speight(2001), p.206 97 Figuerola-Ferretti/Gilbert (2001), p.175f.
Another topic that has received considerable attention when comparing administered prices with exchange prices is the influence of forward trading on price volatility. Forward trading is made possible through the introduction of exchange trading of commodities. Scientific opinion on whether this form of trading increases of decreases price volatility is divided. An influential work by COX summarizes the empirical findings of his time and concludes that "a significant price effect of future trading reflects an increase in market information"98, and more specifically that the comparison of seven non-metal commodities across a period including times of future trading and no future trading conceded a lesser volatility when future trading was allowed.
SIMPSON/IRELAND investigate the impact of financial futures on the cash market for treasure bills. Summarizing the scientific work of their time, they can find little evidence "on destabilizing speculation in financial futures markets "99. Based on their own analysis they conclude that there was indeed a volatility reducing effect when future trading was introduced but this effect vanished when trading volume increased.100
SLADE/THILLE conclude that while predictions of destabilizing speculation models is mixed, a positive correlation between the two variables price volatility and forward trading volume exist. However, testing for a direct connection between the two, they conceded that
"the link between the two is not direct and that both variables are influenced by a common factor such as the arrival of new information"101, thus refining earlier research, which observed a directly destabilizing effect. Their work is also a notably exception insofar that price data are analyzed together with underlying industry data, namely supplier concentration.
This is unique as the focus on price data alone is the prevalent method of most price analyses.
Supplier concentration
The impact of the level of concentration of the supplier side on price level and volatility has been investigated and debated for some time. Researchers analyzing the impact of industry concentration on price volatility are unusually unanimous in their findings that prices appear to be more stable in oligopolies. Two representative works on this topic are therefore only briefly presented.
98 Cox (1976), p.1232f.
99 Simpson/Ireland (1985), p.372 100 Simpson/Ireland (1985), p.378 101 Slade/Thille (2006), p.251
Analyzing the price rigidity of steel and a price index of non-ferrous metals as well as other commodities and some manufactures, CARLTON concludes that "The level of industry concentration is strongly correlated with rigid prices. The more concentrated the industry, the longer is the average spell of price rigidity."102 SLADE/THILLE observe the prices of the six non-ferrous base metals between 1990 and 1999 and concur with earlier observations that "
commodities that are produced in more concentrated markets tend to have more stable prices."103
Another aspect of supplier concentration that is often analyzed together with its impact on volatility is the influence on price level. Several authors conclude that a higher supplier concentration leads to higher price levels. SLADE/THILLE suggest that "strong evidence that a more concentrated industry is associated with higher prices, as the conventional wisdom predicts"104. MAXWELL predicts a change in the price level of nickel due to a decreasing supplier concentration.105
3.1.4 Summary and evaluation
One can summarize that scientific studies of long-term price trends and price cyclicality based on statistical analysis generally assume an "agnostic view"106 of the subject. The majority of studies are concerned with finding evidence for the existence of a phenomenon in price development, i.e. the "what" but fall short of the "why", i.e. the change in underlying market forces. In the majority of cases, scholars are concerned with analyzing price data and make little effort to relate findings to underlying market forces. The statistical rigour of many studies to identify and define price cyclicality or a long-term trend is thus unmatched by comprehensive empirical research on the causes of such phenomenon. In studies based on visual inspection attempts are made to relate price trends and discontinuities back to underlying market drivers. However, a reference to the stylized fact of technical innovation overcompensating the cost of depletion prevails. A notable exception is RADETKI's study on the anatomy of commodity super cycles. However, his conclusions are based on the analysis of a metals and mineral and other indices, thus remaining on an aggregated level.
Studies on price volatility yield more tangible explanations as to which underlying industry factors influence volatility. Albeit divided in opinion, evidence suggests that a changing
102 Carlton (1986), p.638 103 Slade/Thille (2006), p.249 104 Slade/Thille (2006), p.246 105 Maxwell (1999), p.14 106 Cuddington/Jerret (2008), p.2
market structure towards exchange trading as well as supplier concentration has a measurable impact on metal price volatility.
A further insight from all reviewed studies on metal price development is that refractory or other minor metal prices are rarely in the scope of mineral economists. The exceptions reviewed in this work are listed in Table 4:
Author Refractory metals covered
Other minor metals Data/ analysis
constraint Potter/ Christy
(1962)
Manganese, magnesium, ferro-alloys
Part of a metal index
Labys/
Kouassi/
Terraza (2000)
Tungsten
Jerret/
Cuddington (2008)
Molybdenum
Labys/
Lesourd/
Badillo (1998)
Tungsten
Table 4 - Coverage of refractory and other minor metals in literature on metal price series
Source: Own illustration
Of thirty studies reviewed on metal price development, only four contain price series of refractory metals. POTTER/CHRISTY include manganese and magnesium as well as ferro-alloys in their work but only as part of an index with non-ferrous base metals. LABYS et al.
include a price series of tungsten. Neither authors specify their choice. A notable exception are JERRET/CUDDINGTON, who justify their choice of a price series of molybdenum by assessing that the metal is "critical in the early phases of industrial development and urbanization"107.
107 Jerret/Cuddington (2008), p.188
3.2 Review of literature on metal demand
The purpose of the following chapters is to review economic literature on metal demand with the aim to identify structural factors that are thought to predominantly influence metal prices as well as to gain an overview of methodologies to analyze metal demand. While the importance of these structural factors rooted in demand for metal price development is being acknowledged in studies on metal price development, a proper verification and substantiation in metal price research is largely absent.108 Also, the metal markets covered are evaluated to assess whether the omission of refractory or other minor metals in literature on metal prices prevails.
Myriads of factors may potentially influence metal demand and numerous studies are devoted to understanding underlying structural factors and deduce demand models accordingly. Which factors are considered depends often on the analytical approach chosen. RADETZKI/TILTON identify four methodologies in academic literature109 to analyze metal demand:
Intensity of use technique
Demand function estimation
Production function estimation
Input-output analysis
Of all methodologies, the intensity of use (IU) technique appears to be the most prominent. A series of theories emerged from it, attempting to find recurring, metal independent patterns of the development of metal used by an economy per unit of national income. The IU technique and theories derived from it will be reviewed in chapter 3.2.1. Another albeit less commonly applied methodology is the demand function estimation, which will be reviewed chapter 3.2.2. The production function estimation, which is employed less frequently is covered in brief in chapter 3.2.3. The input-output analysis is rarely used anymore and is therefore not explicitly reviewed in this work.
3.2.1 Intensity of use concept
The intensity of use concept constitutes that an economy's metal demand depends on the economy's macroeconomic development usually measured by GDP as well as by the
108 Compare chapter 3.1
109 Radetzki/Tilton (1990), p.25ff.
economy's mix of product output and the individual metal concentration in each product.110 Specifically, demand Dt of a metal may be expressed as:
where Pi denotes the economy's output of the ith final good in physical units, ai the amount of the metal used for the ith good, and nt the number of goods produced in the economy in the period of time t. Dividing the total output of the ith good Pit by the economy's income Yt during t yields
From equation (4) it becomes clear that the intensity of use (IU) in time period t is a function of the material composition of product (MCP) ait and the production composition of income (PCI) bit. The former expresses the mix of materials used to produce individual goods while
From equation (4) it becomes clear that the intensity of use (IU) in time period t is a function of the material composition of product (MCP) ait and the production composition of income (PCI) bit. The former expresses the mix of materials used to produce individual goods while