Copper Concentrates Benchmark Prices Estimation with an

Assessment of the Trading Business

Mención Doctorado Industrial

Programa de Doctorado

“Gestión Estratégica y Negocios Internacionales”

Directores de Tesis

Prof. Dr. Bernabé Escobar Pérez

Profª. Dra. M.ª del Mar Miras Rodríguez

Doctorando

Francisco José Díaz Borrego

Sevilla, 2023
INDEX

Chapter 1: Introduction ............................................................................................... 10

1.1. Copper and its trading .......................................................................................... 12

1.1.1. Copper production ......................................................................................... 13

1.1.2. Copper concentrates ...................................................................................... 14

1.1.3. Copper trading ............................................................................................... 16

1.2. Objectives ............................................................................................................ 18

1.3. Structure ............................................................................................................... 20

Chapter 2: Theoretical Foundations ........................................................................... 24

2.1. Theoretical framework ......................................................................................... 24

2.2. Copper concentrates prices .................................................................................. 28

2.2.1. Payable metal terms ...................................................................................... 28

2.2.2. Discounts ....................................................................................................... 29

2.2.3. Penalty terms ................................................................................................. 32

2.3. Concentrates layout .............................................................................................. 33

2.4. Copper concentrates price modelling................................................................... 36

Chapter 3: Looking for an accurate forecasting of Copper TC/RC Benchmark

Levels ............................................................................................................................. 39

3.1. Introduction .......................................................................................................... 39

3.2. The need for accurate forecasts of Copper TC/RC .............................................. 40

3.3. Related Work........................................................................................................ 42

3.4. Methodology ........................................................................................................ 45

3.4.1. Models in Methodology ................................................................................ 46

3.4.2. Models Comparison Method ......................................................................... 50

2
 3.4.3. Models Calibration ........................................................................................ 53

3.5. Analysis of Results............................................................................................... 56

3.5.1. GBM Forecasts .............................................................................................. 56

3.5.2. OUP Forecasts ............................................................................................... 59

3.5.3. LES Forecasts ................................................................................................ 62

3.6. Discussion and Model Comparison ..................................................................... 64

Chapter 4: Estimating Copper Concentrates benchmark prices under dynamic

market conditions ......................................................................................................... 68

4.1. Introduction .......................................................................................................... 68

4.2. Copper concentrates benchmark price model ...................................................... 69

4.3. Copper concentrates benchmark price forecasting .............................................. 74

4.3.1. Parameters estimation ................................................................................... 75

4.4. Metals and Copper TC/RC Forecasts .................................................................. 76

4.5. Copper concentrates benchmark price forecasts and errors ................................. 78

4.2.1. Effects of discounts and punishable elements over the price of concentrates82

Chapter 5: Managing a High Uncertainty Scenario through a Real Option

Assessment: Evidence from a Copper Concentrate Trader...................................... 85

5.1. Introduction .......................................................................................................... 85

5.2. Related Work........................................................................................................ 86

5.3. Materials and Method .......................................................................................... 89

5.3.1. Concentrate Price .......................................................................................... 89

5.3.2. Copper concentrate trading companies valuation model .............................. 94

5.3.3. Equity free cash flows ................................................................................... 95

5.3.4. Terminal value ............................................................................................... 97

5.3.5. Discount factor .............................................................................................. 97

3
 5.3.5. Real options ................................................................................................... 99

5.4. Results and Discussion ...................................................................................... 103

5.4.1. Concentrate price forecasts ......................................................................... 103

5.4.2. Equity free cash flows ................................................................................. 105

5.4.3. Discounted Cash Flows Valuation................................................................110

5.4.4. Real options valuation .................................................................................. 111

5.4.5. Abandonment option ....................................................................................112

5.4.6. Growth Option..............................................................................................113

5.4.7. Company Value ............................................................................................114

Chapter 6: Conclusions ............................................................................................... 117

6.1. Implications of developing forecasts for Copper TC/RC ...................................117

6.2. Accuracy of TC/RC benchmark levels forecasts ................................................119

6.3. The importance of having a benchmark on copper concentrates prices ............ 120

6.4. Applying the models and methodologies to anticipate future business

performance .............................................................................................................. 121

6.5. Limitations and future lines ............................................................................... 122

References.................................................................................................................... 126

Appendix: Published Papers ..................................................................................... 141

4
INDEX OF TABLES

Table 1. Recovery factors (r) and minimum deductions (MD) for copper and silver. ... 29

Table 2. Recovery factors (r) and minimum deductions (MD) for gold......................... 29

Table 3. Copper TC/RC benchmark levels ..................................................................... 31

Table 4. Gold and silver refining charges values applied ............................................... 31

Table 5. Penalty parameters for punishable elements..................................................... 33

Table 6. Concentrates analyzed layout. Payable elements. ............................................ 34

Table 7. Concentrates analyzed layout. Punishable elements. ....................................... 35

Table 8. Summary of previous literature research .......................................................... 44

Table 9. Iterations on LES smoothing constants optimization ....................................... 56

Table 10. Average of main measures of error for GBM simulations 2013 to 2017........ 57

Table 11. Number of paths for which measures of error are higher for monthly steps than

for annual steps in GBM ................................................................................................. 58

Table 12. Average for main measures of error for GBM simulations 2013 ................... 58

Table 13. Average for main measures of error for GBM simulations 2013-2014 .......... 58

Table 14. OUP parameters obtained in the calibration process ...................................... 59

Table 15. Number of paths for which measures of error are higher for monthly steps than

for annual steps in MR-OUP 5-Steps MC simulation. ................................................... 61

Table 16. Average for main measures of error for MR-OUP simulations 2013-2017 .... 61

Table 17. Average for main measures of error for MR-OUP simulations 2013 ............. 62

Table 18. Average for main measures of error for MR-OUP simulations 2013-2014 .... 62

Table 19. LES Model Optimised Parameters ................................................................. 63

Table 20. LES error measures for different steps-ahead forecasts ................................. 64

Table 21. Improvement for averaged monthly steps before annual steps ...................... 65

Table 22. TC/RC best error measures for 2013 forecasts after simulations ................... 65

5
Table 23. TC/RC best error measures for 2013 – 2014 forecasts after simulations ....... 66

Table 24. TC/RC Best error measures for 2013 – 2017 forecasts after simulations ...... 67

Table 25. Contracts considered for parameters estimation. ............................................ 76

Table 26. Parameters values for Schwartz and Smith’s model ....................................... 77

Table 27. Fitting errors for concentrates price estimation (2004 – 2018) ...................... 79

Table 28. Errors for concentrates price forecasting ........................................................ 80

Table 29. Copper Concentrate Specifications................................................................. 90

Table 30. Statistical Distribution of Historical daily returns of metal prices ................. 91

Table 31. Constant parameters for Concentrate Pricing (2021 – 2025) ......................... 93

Table 32. MASCOFLAPEC weighing ........................................................................... 99

Table 33. Trader’s supply capacity increasement steps ................................................ 101

Table 34. Expected Annual Average Price for Copper Concentrate 2021 – 2025 ........ 105

Table 35. Expected Annual Average Price for Copper Concentrate 2021 – 2025 ........ 106

Table 36. Market Strategy 2021 – 2025 ....................................................................... 109

Table 37. Expected Income Statements 2021 – 2025 ................................................... 109

Table 38. Values for the calculation of the Discount Factor ..........................................110

Table 39. Discounted Equity Free Cash Flows 2021 – 2025......................................... 111

Table 40. Concentrate exercise price for growth option increase steps .........................112

Table 41. Concentrate exercise price for abandonment option in each year .................112

Table 42. Discounted Equity Free Cash Flows 2021 – 2025.........................................112

Table 43. Growth Option Probabilities and Value .........................................................113

6
INDEX OF FIGURES

Figure 1. Cu ores flotation flowsheet. ............................................................................ 15

Figure 2. Industrial processing of copper sulphides ore to obtain copper cathodes. ...... 16

Figure 3. Monte Carlo paths for either TC or RC levels using monthly steps ............... 57

Figure 4. The first 20 Monte Carlo paths following the OUP model ............................. 60

Figure 5. Linear Exponential Smoothing model (LES) forecasts .................................. 63

Figure 6. Observed and Schwartz and Smith’s two-factor model estimations ............... 76

Figure 7. Estimations of concentrates 1 and 2 prices compared to objective price........ 78

Figure 8. 2019 Model's forecasts for concentrates ......................................................... 79

Figure 9. Discounts’ relative weight in payable metal content in concentrate ............... 83

Figure 10. Concentrate price paths calculation process ................................................. 91

Figure 11. Rate of decrease of sigma for different number of simulations .................... 92

Figure 12. Real Options valuation scheme ..................................................................... 95

Figure 13. Global average concentrate prices. Expected Value calculation workflow 100

Figure 14. ICDF for annual Copper Concentrates prices used for ROV...................... 103

Figure 15. Five sample price paths for copper concentrate 2021 – 2025 ..................... 104

Figure 16. Annual average price paths for copper concentrates 2021 – 2025 .............. 104

Figure 17. Average price path forecasts. Epected annual average price 2021 – 2025.. 105

Figure 18. Revenue, EBITDA, Self-financing variations ............................................ 107

Figure 19. Revenue, EBITDA, Self-financing relative variations ............................... 108

Figure 20. Real Options valuation alternative for Copper Concentrate Trader .............115

7
8
9
 Chapter 1: Introduction

Commodity trading is one of the oldest forms of economic activity, albeit its functioning is yet

scarcely understood nowadays. On the other hand, the large-scale expansion of fuels, minerals and

grains trading in recent decades has been one of the vital elements which has facilitated globalization

(Buchan and Errington, 2016).

The activity of sourcing the natural resources needed by the industry and society in general

constitutes one of the pillars of the global economy, obtaining commodities from where there is

abundance of and moving them to where they are most needed in time and form, equilibrating thus

mismatches between offer and demand (Meersman et al., 2012). Commodity trading stretches from

grain like wheat or corn, which comprises the base of human and animal feeding, to copper, iron or

aluminium, metals that have laid the foundations the developing of information technologies,

transportation and infrastructures, as well as oil and natural gas, key elements for the world’s energy

production.

The modern form of commodity trading dates back to the beginning of the 17th century, with the

stablishing of the East India Company (EIC) in 1600, and the Vereeningde Oostindische

Compagnie (VOC), or the Netherlands East India Company in 1602 (Robins, 2012). Both these

two companies, which were in hands of private investors, possessed the concessions of their

monarchies that granted them the exclusive right to trade all kind of commodities, mainly cotton,

opium, spices or tea, between the middle and far east and their European metropolis.

Later on, the huge demand for metals that arose in the beginning of the 19th century stemming from

the advances of the industrial revolution fostered the creation of the London Metal Exchange (LME)

10
in 1877 (Geman and Smith, 2013), formalizing a trading which had been taking place informally in

the outskirts of the Royal Exchange of London since decades earlier. At the LME, copper suppliers

from Chile or tin suppliers from Malaisia would find their buyers, enabling them to set prices for

future deliveries. Some of the first standards withing commodity trading were then first set,

remaining in force still today, such as the 3-month future contract for copper. This originally was

equivalent to the time that copper shipments from Chile took to arrive to England (The London

Metal Exchange, 2019). In a similar fashion, the first future exchanges were stablished across the

world, like the Chicago Board of Trade (CBOT), which in1864 created the first standardized futures

contract for grain, being still nowadays a reference market for setting of agricultural commodities

prices.

Additionally, the generalization of the use of oil in the 20th century, mainly since in 1911 gasoline

became the main product that Standard Oil’s refineries made, overtaking kerosene (Murty, 2020),

to meet the unstoppable increase of its demand due to the growing use of cars, transformed the

commodities trading business. The Seven Sisters appeared in the 40s. This seven large vertically-

integrated petroleum conglomerate, were capable of controlling every stages of the oil production,

from extraction to refining, going through transportation and retailing, hence becoming the first oil

cartel (Buchan and Errington, 2016).

The 1973 Oil Crisis, caused by the embargo by the Arab countries to western nations due to the

Yom Kippur war, terminated with this model of vertical integration. This allowed a new model of

free competence between private companies that traded with commodities to appear. These new

independent companies, like March Rich & Co. transform commodities in space through logistical

systems, in time through warehousing and in form through processing by the making of blends

11
(Pirrong, 2013). Many companies involved in commodity trading ceased therefore to be present in

all the parts of the process and focused solely in arbitrating in commodity markets.

Since 1981, with the financialization of futures markets after the creation of the first futures contract

for oil by the New York Mercantile Exchange (NYMEX), like the ones that already existed for

copper or corn, and large investment banks opening dedicated commodities trading desks (Blas and

Farchy, 2021), commodity trading reshaped to look like today. Commodity trading is now

decentralised, dominated by companies that reach both physical and financial markets, for either

hedging their exposure in the physical world as well as for speculating. These companies have

developed sophisticated financial structures for their operations, reaching in some cases to acquire

a meaningful importance in the economy, even to be considered as systemic -too big to fail-

(Jacques and Simondet, 2016).

1.1. Copper and its trading

Copper is a key element for the current transformations in energy and technology. It is the best non-

precious metal conductor, which makes it a key component of electrical wires, generators, motors,

transformers and renewable energy production systems. Additionally, copper and copper brass are

key components for construction materials and industrial machinery and equipment, thanks to its

resistance to corrosion, its durability, as well as its machinability and ability to be cast.

Also, copper plays a major role in transportation uses, being critical in the electrification of fossil-

fuel-based transportation means, being intensively used an in Electrical Vehicles (EV). Indeed, a

conventional car contains 23 Kg of copper on average, whereas a Hybrid Electric Vehicle (HEV)

has almost twice as much on average, while a Plug-in Hybrid Electric Vehicle (PHEV) contains

about 60 Kg (International Copper Study Group, 2022).

12
In 2021, the use of refined copper in the world reached its historical peak with 25.3 million tons

consumed globally. Having achieved an annualized consumption increase rate of 3.3% since 1990.

Meanwhile, the total used of copper in the world, also accounted for recycling, approached 32.5

million tons (International Copper Study Group, 2022).

1.1.1. Copper production

The world’s copper production is essentially achieved through alternative processes which depend

on the chemical and physical characteristics of the copper ores extracted. According to the USGS’

2017 Mineral Commodity Summary on Copper (U.S. Geological Survey, 2017), global identified

copper resources contained 2.1 billion tons of copper as of 2014, of which about 80% are mainly

copper sulphides, whose copper content has to be extracted through pyrometallurgical processes

(Schlesinger and Davenport, 2011).

The main source of copper, despite the growing importance of scrap recycling, is still primary

copper, which is obtained by the extraction and processing of copper ores. According to the U.S.

Geological Survey (2023) the world’s primary copper production reached 21.2 million tons in 2021,

of which 3.9 million were only obtained through the Solvent Extraction/Electro-Winning (SX/EW)

process, whereas the rest were achieved by smelting and refining of concentrates of different grades

(International Copper Study Group, 2022).

The majority presence in the earth’s crust of sulphide deposits, having low copper grades between

0.5% - 2.0%, (Schlesinger and Davenport, 2011) makes concentrates the only technically and

economically feasible mining product from most copper mines. Copper concentrates are thus the

main products offered by copper mines, representing 80% of the origin of all cathodic copper

obtained to be used with industrial applications (Norgate and Jahanshahi, 2010). Concentrates are

produced from copper sulphide ores, occurring naturally in different kinds of deposits, being

13
porphyry deposits the most relevant ones. According to the USGS’ 2015 Assessment of

Undiscovered Copper Resources of the World (Hammarstrom et al., 2019) porphyry deposits

represented 1.8 billion tons of all identified copper resources, whereas undiscovered deposits of

these kind alone are estimated to contain 3.1 billion tons of copper.

Ore grade differs depending on the deposit class, which for the case of porphyry deposits oscillates

between 0.3% and 2.0% of copper, with the top 25 mines by Cu content having a mean grade of

0.49% (Mudd et al., 2013). These copper deposits may also contain some other valuable elements,

such as gold or silver with variable gradings, as is the case of the Indonesian mine Grasberg with an

average gold grade in the ore mined during 2018 of 1.58 g/t, as well as having reached an average

copper grade at the mill of 0.98% (Freeport-McMoRan, 2018).

Other relevant, though fewer, common classes of deposits, such as sediment-hosted copper

deposits, which account for 0.31 billion tons of global total identified resources and 0.42 billion tons

of undiscovered world resources, are likely to contain significant amounts of additional by-products,

for example cobalt or silver, as well as higher average copper grades (Hammarstrom et al., 2019),

such as the Katanga mine in D.R.C. with an average Cu grade of 3.49% and an average cobalt grade

of 0.46% in 2018 (Katanga Mining Limited, 2018), and KGHM Polska Miedź’s Polish mines

whose ores graded on average 1.49% of copper and 48.6 g/t of silver throughout 2018 (KGHM

Polska Miedz, 2018).

1.1.2. Copper concentrates

Except for oxidized ores, which are mainly treated through a hydrometallurgical process, ores

undergo an enrichment process known as froth flotation to achieve concentrates. Through flotation,

grades are significantly increased while the tonnage of the concentrate outflow is much less than the

tonnage of the ore inflow current. Most of the copper content in the ore, as well as the by-products,

14
pass into the concentrate with minimum losses of valuable elements being removed to the tail

outflow current (Figure 1).

Figure 1. Cu ores flotation flowsheet.

Source: Schlesinger and Davenport (2011)

Copper sulphide ores must undergo froth-flotation to obtain concentrates containing around 30%

of copper, as the average grades of ores mined globally range from 0.5% - 2% Cu which makes

direct smelting unfeasible for economic and technical reasons (Glöser et al., 2013; International

Copper Study Group, 2017).

Concentrates are latter further treated at smelting plants through pyrometallurgical processes

(Figure 2) to achieve copper anodes, containing 99.5% of Cu. Copper is eventually refined

electrochemically to achieve copper cathodes with 99.9% Cu (The London Metal Exchange, 2019).

15
In addition, gold, silver and other impurities in the concentrates are recovered from the anode slimes,

produced at the electrorefining stage, using multiple techniques. The grade of these slimes is a direct

reflection of the levels of these elements in the concentrate at the beginning of the smelting process

(Schlesinger and Davenport, 2011).

Figure 2. Industrial processing of copper sulphides ore to obtain copper cathodes.

Source: Schlesinger and Davenport (2011)

1.1.3. Copper trading

The relevance of copper trading is undeniable. In 2020 copper ores and concentrates trade globally

reached 61.8 b $USD, being China the main importer and Chile the leading exporter followed by

Peru (OEC, 2023). In 2016 exports of copper ores, concentrates, copper matte and cement copper

increased by 1.5%, reaching 47.3 b $USD, while imports attained 43.9 b $USD (United Nations,

2016). In addition, the global mining capacity is expected to rise by 10% from the 23.5 million

16
tonnes recorded in 2016 to 25.9 million tonnes in 2020, with smelter production having reached the

record figure of 19.0 million tonnes in 2016 (International Copper Study Group, 2017).

Copper cathodes are regarded as the pure form of copper and are traded as the underlying asset of

futures contracts at major physical future exchanges such as the London Metal Exchange (LME) or

the New York Mercantile Exchange (COMEX). The price of copper on these exchanges is a

reflection of the actual supply and demand of physical copper (Sánchez Lasheras et al., 2015), and

is taken as the core reference for pricing all copper products.

Copper concentrates are usually sold directly by the mining companies that produce them to

smelters and refiners that process them to obtain cathodic copper, frequently under the framework

of long-term supply agreements. The absence of a reference market for concentrates generates that

both producers and processors, as well as independent traders, agree on complex price fixing

mechanisms, which are generally of exclusive application to each specific agreement. These

mechanisms tend to be kept confidentially, as they are regarded as highly sensitive for the parties

involved in concentrates trading. Nonetheless, they have broad repercussions in the financial aspects

of such a relevant trade at a global scale.

Products that are less refined than copper cathodes, such as mattes or concentrates, are priced at a

discount on the exchange price, while products requiring further processing than cathodes, such as

copper wires or rods, are sold at the exchange reference price plus a premium. Discounts and

premiums on the exchange reference prices are usually decomposed into a series of factors

corresponding to the specifications of the products being traded. Discounts associated with

concentrates widely vary in function of the content of the payable elements - copper, gold and silver

- the excess of punishable elements, such as arsenic or bismuth, as well as the current availability of

17
smelters demanding concentrates with those specifications, and the global supply and demand

levels of copper, among other causes.

1.2. Objectives

As previously mentioned, one of the main problems identified is the lack of a reference market for

copper concentrates, unlike there is for cathodic copper or for other industrial or precious metals, as

well as for other commodities. The absence of a specific reference market adds an additional layer

of complexity to copper concentrate trading, as the market price for copper is not the only factor

that affects the price of concentrates despite being the main one. Reference markets for commodities

are of key importance. Apart from the ability of price discovery, market participants have also the

possibility of using financial instruments to mitigate, transfer and, essentially, hedge, their risks

related to commodities’ high volatility.

Much previous research has primarily focused on possible modelling alternatives for reference

commodity prices, such as Brennan and Schwartz (1985), Liu et al. (2017) and Dehghani and

Bogdanovic (2018), who explore different alternatives to forecast copper prices at major

commodity exchanges. Others, have modelled the exchanges’ reference prices of some

commodities to conduct applied studies, for instance Zhang et al. (2015), who employ the LBMA

price of gold for the valuation of a mining project, and Guj and Chandra (2019), who model the

LME copper price to apply to the real option valuation of a copper mine.

In addition, previous research have instead dodged the problem of copper concentrates pricing,

focusing on cathodic copper price modelling. This has been applied to multiple projects that may

be affected to a certain extent by copper price variations at some major metal futures exchange, such

as the London Metal Exchange (LME). This simplification of reality, either referred to the mining

18
business or the smelter side of the process, along with trading itself, falls short of what market

participant come across on a daily basis when dealing with copper concentrates.

Therefore, we have set the main objective of this research to develop a model for copper concentrate

benchmark prices, so we can set an useful reference for market participant to value their

concentrates, as well as to be able to forecast potential price trends on copper concentrates. Through

the analysis of available data, as well as the existent literature, this research aims to achieve a better

understanding on the mechanisms for copper concentrates price setting, as well as on the

agreements reached between traders, miners and smelters, along with the corporate finances of

commodity traders.

On the other hand, there are also multiple examples in the literature on the finances of mining

companies and on the valuation of mining projects, as well as the assessment of the economic

aspects of mining investments considering the uncertainty involved. However, regarding to mineral

commodities trading, specifically when treated as a standalone element, separated to the mining or

smelting activities, little has been researched so far. Hence, aiming to improve our understanding of

mineral commodities trading, we have set the second objective of this research the financial

assessment of a copper concentrates trader through Real Options, as a case of analysis of mineral

commodities trading. Treating the activity of trading in an isolated and independent way from the

extractive and processing stages.

Additionally, this research pursues to provide practical, usable tools, that may be of application by

market participants in their regular operations. Emphasizing on improving the current process of

copper concentrate price setting, as well as developing price forecasting models, which may be

19
especially beneficial for a better financial planning of any company that trades, process or produce

copper concentrates.

1.3. Structure

Following the introduction there is a chapter dedicated to the theoretical foundations in which we

review the most recent advances on commodities pricing and trading, setting the theoretical

framework for the development of our research. In this chapter, the methodology we have employed

to develop the copper concentrates benchmark prices model is briefly explained, as well as the Real

Options methodology and variables which we have resorted to for the valuation of the copper

concentrates trader.

Additionally, the different concentrates’ layouts that have been used to test the benchmark price

model forecasting ability are also described. These concentrates have been chosen as they depict

both regular trading operations, as well as most complex ones. Also, in this chapter we have set the

parameters, such as the main error measures we have used, needed to develop our research and

perform further testing of the models’ forecasting capacity, as well as to carry out the Real Options

analysis.

Chapter 3 is committed to find an accurate model that allows us to forecast copper TC/RC

benchmark levels. The model should deliver reliable forecasts, while at the same time may be of

practical application by industry participants and of easy integration in a latter, larger-scale copper

concentrates benchmark prices model.

To find the best solution, we have compared five measures of error of different forecasts for copper

TC/RC annual benchmark levels obtained using well-known models in literature: Geometric

20
Brownian Motion (GBM), Mean-Reverting Orstein Uhlenbeck Process (OUP) and Linear

Exponential Smoothing (LES).

Three forecasting horizons were set, one, two and three years ahead, in order to compare the

different models short- and longer-term forecasting capabilities. To carry out the different

simulations, as well as the different models’ calibration process, we have employed an historical

dataset of copper TC/RC benchmark levels from 2004 to 2017.

The copper concentrates benchmark prices model is mathematically described in Chapter 4, where

we have also performed a series of simulations to check its forecasting capacity. The model delivers

short-term (up to one year ahead) price estimations for copper concentrates basing on independent

forecast for metal prices (copper, gold and silver), as well as resorting to forecasts of main discounts

(copper TC/RC and gold and silver RC), along with other relevant factors.

Both historical future contracts and spot price data have been used for copper, gold and silver from

2004 to 2018 to perform the Monte Carlo simulations to obtain concentrate prices forecasts. The

results achieved in Chapter 3 on TC/RC benchmark levels have been integrated in this model to

give full coherence to our research. The five concentrates described in Chapter 2 have been used to

test the model’s accuracy, while also employing the Mean Average Percentage Error (MAPE) as

main indicator of precision.

Chapter 5, entitled ‘Managing a High Uncertainty Scenario Through a Real Option Assessment:

Evidence from a Copper Concentrate Trader’ is dedicated to the analysis of the copper concentrates

trading business through the use of the Real Options methodology. Real Options allow to assess the

uncertainties associated to the high volatility that concentrate prices experience, while also allowing

21
to take advantage of potentially beneficial opportunities on the trading business development that

may arise over time.

The trading company valuation is carried out by projecting ahead for a total of five years its market

strategy, as well as its balance sheet and profits and losses statements. The concentrate benchmark

price model developed in Chapter 4 is used here to obtain numerical results of the expected value

of the trading company, which clearly depends of future concentrate price. Simulations have been

carried out in a simplified version of the methodology proposed in Chapter 4, opting form an

statistical approach for metal pricing, using the distributions of daily metal prices returns, instead of

the Kalman-Filter technique which requires also futures contracts data and which increases the

simulation complexity meaningfully.

To validate the proposed methodology, a single layout of concentrate has been assumed while

expansion and abandonment criteria for the trader operations have been set as a function of potential

profitability increases or declines as copper concentrate prices vary over time. The Real Options

methodology, modelled with Monte Carlo simulations, have proven an useful and applicable tool

for trading operations as well, like they had already been tested for mining operations in multiple

previous research.

Finally, conclusions, limitations and future research are explained in Chapter 6. The results of our

research indicate that the copper concentrate benchmark price model is sufficiently accurate for

short-term price forecasts, delivering reliable estimations of copper concentrates prices for up to one

year ahead. Additionally, the Linear Exponential Smoothing model is capable of providing a single

forecast for TC/RC benchmark levels that may give market participants a proper indication of the

22
potential market evolution based on preceding trends, while also being suitable to be included in the

copper concentrates benchmark price model.

When applied to a real-life case, both the models and the Real Options methodology has

applications of interest for the market participant to help them optimize their trading operations as

well as to advert or prevent potentially loss-generating situation as a consequence of future market

downturns.

On the other hand, the lack of publicly available data on commodity traders’ financials, as well as

on actual copper concentrates transactions, with precise details on prices and terms, presents the

main limitation of our research, as well as the main future line of research of interest should further

data be available.

23
 Chapter 2: Theoretical Foundations

2.1. Theoretical framework

There is also an abundance of references in literature regarding flows of primary and secondary

copper (Glöser et al., 2013; Jaunky, 2013; Mudd et al., 2013; Chen et al., 2016), which mainly

focuses on analysing and quantifying where the copper used in human activities comes from, what

it used for and what its transformations take place from an aggregated perspective. Many others

focus their efforts in modelling and forecasting the price of copper and of copper future contracts

using different methods (Kriechbaumer et al., 2014; Buncic and Moretto, 2015; Wets and Rios,

2015; Sverdrup, 2016; Liu et al., 2017; Dehghani and Bogdanovic, 2018; Guzmán and Silva, 2018).

It is also frequent to find examples in the literature where cathodic copper is taken as what mines

produce, leaving out copper concentrates themselves due to their complexity to analyze them, or

simply taking into consideration the copper content within the concentrates to simplify. Thus,

works like Guj and Chandra (2019) use a forecast of cathodic copper prices at the London Metal

Exchange (LME) and get rid of the complicated process of pricing concentrates through the

assumption of a Net Smelter Value (NSV), as a percentage to apply to the value of copper in the

concentrates sold by a certain mine.

Although these kinds of approaches might be valid, even desirable in certain context to specifically

aimed towards a practical application, we must highlight that the reality, either regarding to mining

operations, or just trading agreements, is notably different. Most mines produce concentrates

containing form 20% to 40% of copper (Delbeke and Rodriguez, 2014), being its price linked to

that of copper in international markets, but also to other important circumstances, as well as

experiencing significant differences between mines or concentrates layouts, among other factors.

24
Thus, if precision is desired, one cannot assume homogenic prices for concentrates, whereas

simplifications make no sense. The complex task of pricing copper concentrates implies considering

both their content and the prices at reference metal markets of those metals that give value to the

concentrates, such as copper, gold and silver. Also, it becomes necessary to reflect the remaining

factors that, at least in an indirect way, have an impact on the price that concentrates are sold, such

as the levels of concentrates supply/demand globally, or the processing capacity of the world’s

copper smelters at a given time.

The most common price setting method for copper concentrates prices using copper Treatment

Charges/Refining Charges (TC/RC), along with other discounts and penalties, has scarcely been

analysed in the literature, despite its importance for copper concentrate trading. Furthermore,

TC/RC setting has not been analysed in depth, even though they represent the largest part of copper

smelters’ revenue. Being their levels a factor of major concern for not only smelters, but also copper

mines and traders around the world.

To find an adequate forecasting model for copper TC/RC benchmark levels, in this research we

have treated the historical available TC/RC benchmark data as a price timeseries, resorting to the

available models for commodity price modelling of which there are abundance in the literature.

In this sense Xiong et al. (2015), compares various forecasting models after applying them to

agricultural commodities, or Zhang et al. (2015) that employs a mean-reverting model for the price

of gold, specifically applied to the valuation of a mining project, just like Brennan and Schwartz

(1985) do for an unidentified commodity in their well-known work.

Finally, despite the great importance of trading activities within commodity markets, there is also a

remarkable lack of systematic research in literature regarding this crucial aspect for commodities.

25
By the contrary, when the financial aspects of commodities are related to the productive or

transforming stages, as well as with the extractive phases, literature becomes prolific (Saramak,

2011; Jovanović and Stanimirović, 2012; Jovanović et al., 2013; Boulamanti and Moya, 2016).

Additionally, the copper mining is also the research topic of many works, where not only the

technical aspects are analysed, but also those closer to the financial optimization of copper mining

operations, putting special attention to the uncertainties related to the variability of metals prices

over time. Works like Dehglani et al. (2014) approach the problem from the Net Present Value

(NPV) point of view of a mining project applying different techniques to forecast the price of copper

along the proposed time horizon.

On the other hand, since Brennan & Schwartz (1985) first employed Real Options to value mining

assets, this has become one of the main approaches, if not the main one, to perform the valuation of

mining projects. Real Options share a similar view with financial options when valuing tangible

assets under. They allow the conversion of uncertainty related to a mining project to management

alternative based on probable situations.

As summarized by Savolainen (2016) in its extensive historical review, many recent works used

Real Options to value metallic mining projects with different approaches when taking into

consideration those factors considered to be generators of uncertainty, such as price fluctuations or

exchange rates. Hence, works like Botín et al. (2012) pursues to find the optimal design for a copper

mine while trying to maximize the NPV of the project using Real Options to account for the effect

that horizontal dilution has on the deposit.

26
Another series of works have instead used Real Options to examine the complex process of mine

design, considering the significant variation that the profitability of the project is prone to experience

as the price of metals move. In Thompson and Bar (2014) Real Options has been used as a tool to

determine the optimal cut-off grade to set the mine planning of an open-pit mine, while assuming

the stochastic behaviour of the prices of any mineral.

Equally, Asad (2007) resorts to Real Options as a tool to develop a decision policy for the active

design of a copper open-pit mine, looking to maximize its NPV, while simultaneously considering

the dynamic evolution of prices as well as an increase of costs throughout all its lifetime.

It becomes evident that in high uncertainty environments, particularly in those related to extractive

industries like mining, the valuation of investment projects is a complex challenge. In the first place,

widely applied traditional methos like Discounted Cash Flows (DCF) have important deficiencies

as they require a high level of concretion of future scenarios. This, due to the nature of certain types

of investments, either requires making too many assumptions or simply unfeasible. As can be seen

in Samis et al. (2005), Real Options offer clear advantages when the uncertainty of an investment

is high, provided their ability to treat with flexibility natural resource assets, being capable of

reflecting the uncertainty of future cash flows in an efficient way.

The advantages that the Real Options method show when valuing mining projects, and especially

when tackling the variability of metal prices and deposit uncertainties are also explained in Abdel

Sabour et al. (2008), which compares the Real Option method to traditional discounted cash flow

methods by making a comparative ranking of twelve possible designs of an open-pit copper mine

and a gold mine. The results of this analysis show that mine designs chosen following the Real

Option methodology outperforms those obtained using the other methodology.

27
2.2. Copper concentrates prices

Market participants lack a market-wide accepted reference for copper concentrates pricing. The

only possible reference available is the exchanges’ price for cathodic copper for both current and

future expected prices, as well as certain information on part of the discounts agreed by the largest

miners and smelters at the LME Week. This is usually provided by some accredited specialized

consultant, such as Platts or Wood Mackenzie.

Copper concentrates supply agreements set specific mechanisms to price the copper concentrates

to be traded according to their chemical specifications. Usually, most pricing systems establish the

price of copper concentrates as the sum of the payable elements present in the concentrates minus

the deductions and the penalties (Soderstrom, 2008; Seabridge Gold, 2010; Teck, 2012, 2015).

In this research we have followed the same logic to model copper concentrate prices, treating

payable elements separately from discounts and punishable elements. Thus, determining the price

of concentrates as the sum of the amount paid for all payable elements, which is the amount paid

for the copper, gold and silver content in the concentrate, M, minus the sum of deductions applied

by smelters and refiners, D, along with the sum of penalties, P, due to the presence of impurities and

punishable elements that worsen the final properties of cathodic copper or make the concentrates

harder to process.

𝐶𝐶 = 𝑀𝑀 − 𝐷𝐷 − 𝑃𝑃

2.2.1. Payable metal terms

Contract terms for payable metals may vary depending on the market conditions as well as on the

layout of the concentrates. Table 1 shows typical recovery factors and minimum deductions for

copper and silver, which have also been used in this research.

28
 Table 1
Recovery factors (r) and minimum deductions (MD) for copper and silver.
r MD
Cu 96.5% 1.0%
Ag 90.0% 30 g/dmt
Source: Own elaboration. Based on Teck (2015), Ghaffari et al. (2016) and OECD et
al. (2017).

In addition, recovery factors and minimum deductions for gold are shown in Table 2. Unlike copper

and silver, payable gold usually increases as its grade rises in concentrates.

Table 2
Recovery factors (r) and minimum deductions (MD) for gold.
𝑔𝑔 (g/dmt) r MD (g/dmt)
<1 00.00% 1
1–3 90.00% 1
3–5 93.00% 1
5–7 95.00% 1
7 – 10 96.50% 1
10 – 20 97.00% 1
20 – 30 97.50% 1
> 30 97.75% 1
Source: Own elaboration. Based on Teck (2015), Ghaffari et al. (2016) and OECD et
al. (2017).

2.2.2. Discounts

The main discounts involved in copper concentrates price determination are copper Treatment

Charges and Refining Charges, most commonly known as TC/RC, along with gold and silver

Refining Charges. These discounts comprise the bulk of deductions in most copper concentrate sale

agreement.

TC/RC levels for copper concentrates continuously vary throughout the year, relying on private and

individual agreements between miners, traders and smelters worldwide. Nonetheless, the TC/RC

29
benchmark fixed during the LME week each October is used by market participants as the main

reference to set actual TC/RC levels for each supply agreed upon for the following year. Hence, the

year’s benchmark TC/RC is taken here as a good indicator of a year’s TC/RC average levels.

Analysed time series of benchmark TC/RC span from 2004 through to 2018, as shown in Table 3,

as well as the source each value was obtained from. We have not intended to reflect the continuous

variation of TC/RC for the course of any given year, though we have however considered

benchmark prices alone as we intuitively assume that annual variations of actual TC/RC in contracts

will eventually be reflected in the benchmark level that is set at the end of the year for the year to

come.

Historical TC/RC benchmark levels in Table 3 have been used as the in-sample dataset for one-

step ahead forecasts in Díaz-Borrego et al. (2019), which seeks the most suitable forecasting model

for copper TC/RC benchmark levels among frequently-used options for commodity prices.

Copper TC and RC normally maintain a 10:1 relation with different units, with TC being expressed

in US dollars per metric tonne and RC in US cents per pound of payable copper content in

concentrates. In fact, historical benchmark data shows that only in 2010 those values of TC and RC

did not conform to this relation, though it did remain close to it (46.5/4.7).

30
 Table 3
Copper TC/RC benchmark levels.
TC (USD/dmt) RC (USc/lb)
2004 45 4.5
2005 85 8.5
2006 95 9.5
2007 60 6.0
2008 45 4.5
2009 75 7.5
2010 46.5 4.7
2011 56 5.6
2012 63.5 6.35
2013 70 7.0
2014 92 9.2
2015 107 10.7
2016 97.35 9.735
2017 92.50 9.25
2018 82.25 8.225
Note: Data available free of charge. Source: Johansson (2007), Svante
(2011), Teck (2012), Willbrandt and Faust (2013, 2014), Aurubis AG
(2015), Drouven and Faust (2015), Shaw (2015), Aurubis AG (2016),
EY (2017), Schachler (2017), Nakazato (2017).

On the other hand, gold and silver RC have been assumed to remain constant, since smelters tend

to apply around 6.00 to 8.00 USD per troy ounce of payable gold content and 0.40 to 0.50 USD per

troy ounce of payable silver content (Seabridge Gold, 2010; Teck, 2012, 2015; Ghaffari et al., 2016).

Gold and silver RC may be seen in Table 4.

Table 4
Gold and silver refining charges values applied.
RC (USD/troz)
Au 7.50
Ag 0.45
Source: Own elaboration. Based on Seabridge
Gold (2010), Teck (2012), Teck (2015) and
Ghaffari et al. (2016).

31
2.2.3. Penalty terms

Punishable elements are generally undesired in concentrates since some of them may hinder the

concentrate smelting and refining process, such as Chlorine, which may condense as hydrochloric

acid and cause corrosion problems. Others may seriously deteriorate the cathode’s chemical,

mechanical, thermal or electrical properties, such as Arsenic or Antimony which reduce copper

conductivity, or could even represent an important hazard for human health and the environment if

they are not properly managed or mitigated, such as Cadmium, that is a well-known carcinogen, or

Lead, which is highly contaminating as well as toxic to humans (Organisation for Economic Co-

operation and Development et al., 2017).

Normally, each country custom’s authority sets the maximum allowance of the most toxic elements

in imported concentrates See Announcement No. 106 of 2017 (AQSIQ, 2017). Nonetheless,

smelters tend to have stricter requirements than authorities, depending on their ability to process

these elements, considerably reducing the market for most so-called ‘complex’ concentrates, which

sometimes are only accepted by traders for blending purposes with cleaner concentrates. Penalties

applied to concentrates have also assumed to remain at constant levels for each element based on

limits and unitary penalties outlined on reports by Ghaffari et al. (2016) and the OECD (2017).

These penalties have hardly changed in recent years, as may be seen if current levels are compared

with older reports, such as Ghaffari el al.’s (2016) and Seabridge Gold’s (2010), which both assess

the feasibility of the same mining project. Yet a certain downward pressure on the most

environmentally harmful elements’ limits is observable as well as a slight increase in some of the

unitary penalties imposed. The parameters employed to determine penalties due to the excess of

punishable elements may be seen in Table 5.

32
Table 5
Penalty parameters for punishable elements.
F h 𝛿𝛿 (USD)
Arsenic (As) 0.10% 0.10% 3.00
Antimony (Sb) 0.10% 0.10% 3.50
Bismuth (Bi) 0.03% 0.01% 4.00
Cadmium (Cd) 10 ppm 10 ppm 3.00
Chlorine (Cl) 300 ppm 100 ppm 2.00
Fluorine (F) 300 ppm 100 ppm 1.50
Lead (Pb) 0.50% 1.00% 2.50
Mercury (Hg) 10 ppm 10 ppm 2.00
Nickel (Ni) + Cobalt (Co) 0.50% 0.10% 1.00
Selenium (Se) 0.05% 0.01% 4.00
Tellurium (Te) 0.03% 0.01% 4.50
Zinc (Zn) 2.00% 1.00% 2.50
Source: Own elaboration. Based on Seabridge (2010), Ghaffari (2016), OECD (2017).

2.3. Concentrates layout

According to the statistical assessment made by Delbeke and Rodriguez (2014), 50% of world-

wide copper concentrates grade less than 26.67% of copper, while 30% of the world’s production

grades more than 28.452% Cu. Also, according to the statistical distribution of punishable elements,

70% of world’s copper concentrates can be regarded as clean.

On the other hand, less than 10% of all concentrates produced worldwide grade over 34% Cu,

generally with significant levels of some impurities such as lead or arsenic, which qualifies them as

complex concentrates. By the contrary, high values of precious metals are rare in concentrates,

though may be frequent to some extent in those produced by some mines, such as CODELCO’s

Chuquicamata (CODELCO, 2018).

For the purpose of this research, we have selected five sample concentrates that may be

representative of the full scope of possibilities according to the statistical distribution exposed by

Delbeke and Rodriguez (2014). Hence being able to analyze both regular, good quality

concentrates, as well as dirty concentrates, with too high levels of impurities, or precious

33
concentrates with meaningful levels of either gold or silver. The layouts of the five concentrates

considered in this study are shown in Tables 6 and 7.

Table 6
Concentrates analysed layout. Payable elements.
Cu Au Ag
C1 26.62% 0.7 g/dmt 18.4 g/dmt
C2 28.10% 38.1 g/dmt 51.6 g/dmt
C3 27.60% 25.7 g/dmt 187.0 g/dmt
C4 28.33% 6.7 g/dmt 81.4 g/dmt
C5 36.71% 2.4 g/dmt 48.3 g/dmt
Source: Own elaboration. Based on Delbeke and
Rodriguez (2014) for penalty levels.

34
Table 7
Concentrates analysed layout. Punishable elements.

As Sb Bi Cd Cl F Pb Hg Ni Co Se Te Zn
C1 0.002% 0.002% 0.003% 50ppm 49ppm 210ppm 0.04% 0ppm 0.004% 0.009% 0.002% 0.001% 0.22%

C2 0.110% 0.010% 0.004% 40ppm 105ppm 128ppm 0.14% 6ppm 0.002% 0.005% 0.004% 0.001% 0.62%

C3 0.272% 0.042% 0.033% 28ppm 332ppm 258ppm 1.47% 1ppm 0.010% 0.024% 0.035% 0.021% 1.87%

C4 0.62% 0.21% 0.051% 73ppm 44ppm 105ppm 3.05% 4ppm 0.000% 0.003% 0.021% 0.005% 3.12%

C5 0.48% 0.32% 0.066% 41ppm 512ppm 405ppm 2.91% 0ppm 0.000% 0.008% 0.012% 0.040% 4.75%
Source: Own elaboration. Based on Delbeke and Rodriguez (2014) for penalty levels.

35
2.4. Copper concentrates price modelling

The fundamental problem underlaying beneath the surface of our research is the lack of reference

for copper concentrates prices which market participants can rely to. Thus, the cornerstone of our

research is indeed the development of a copper concentrates price model which can be easily used

as a benchmark tool, that can be implemented to value any copper concentrate, while also capable

of providing reliable forecasts for prices.

Throughout this research we have employed different approaches to obtain numerical results out of

the copper concentrates pricing model, depending on whether the forecasting accuracy of the model

was to be tested or if, on the other hand, we were to perform a practical valuation of a trader’s copper

concentrates business. The simulating workload was thus adjusted in order to focus in each specific

objective’s need at each time.

Thus, at time t, the price of the ton of concentrate can be expressed as:

𝐶𝐶𝑡𝑡 = 𝑀𝑀𝑡𝑡 − 𝐷𝐷𝑡𝑡 − 𝑃𝑃𝑡𝑡

Though copper concentrates may contain multiple payable elements apart from copper, should

there be some gold and silver are the most common. Metals in concentrates are paid at market price,

though the full content is not usually paid, as smelters are normally not capable of recovering all the

content. Hence, the price of payable metals in the concentrates is calculated as the payable content

(pp) multiplied by its market price at time t (𝑆𝑆𝑡𝑡 ).

𝑀𝑀𝑡𝑡 = 𝐶𝐶𝐶𝐶𝑡𝑡 + 𝐴𝐴𝐴𝐴𝑡𝑡 + 𝐴𝐴𝐴𝐴𝑡𝑡

𝐶𝐶𝐶𝐶𝑡𝑡 = 𝑝𝑝𝑝𝑝𝐶𝐶𝐶𝐶 × 𝑆𝑆𝑡𝑡𝐶𝐶𝐶𝐶

𝐴𝐴𝐴𝐴𝑡𝑡 = 𝑝𝑝𝑝𝑝𝐴𝐴𝐴𝐴 × 𝑆𝑆𝑡𝑡𝐴𝐴𝐴𝐴

𝐴𝐴𝐴𝐴
𝐴𝐴𝐴𝐴𝑡𝑡 = 𝑝𝑝𝑝𝑝𝐴𝐴𝐴𝐴 × 𝑆𝑆𝑡𝑡

36
Additionally, the payable copper content in concentrates is also dependent on the metal grade (𝑔𝑔),

multiplied by a recovery factor (𝑟𝑟), subject to a minimum deduction (MD).

𝑔𝑔 − 𝑀𝑀𝑀𝑀𝐶𝐶𝐶𝐶 , 𝑔𝑔𝐶𝐶𝐶𝐶 − 𝑀𝑀𝑀𝑀𝐶𝐶𝐶𝐶 < 𝑔𝑔𝐶𝐶𝐶𝐶 × 𝑟𝑟𝐶𝐶𝐶𝐶

𝑝𝑝𝑝𝑝𝐶𝐶𝐶𝐶 = � 𝐶𝐶𝐶𝐶 ∀ 𝑔𝑔𝐶𝐶𝐶𝐶 ≥ 𝑀𝑀𝑀𝑀𝐶𝐶𝐶𝐶
𝑔𝑔𝐶𝐶𝐶𝐶 × 𝑟𝑟𝐶𝐶𝐶𝐶 , 𝑔𝑔𝑖𝑖 − 𝑀𝑀𝑀𝑀𝑖𝑖 ≥ 𝑔𝑔𝐶𝐶𝐶𝐶 × 𝑟𝑟𝐶𝐶𝑢𝑢

𝑝𝑝𝑝𝑝𝐶𝐶𝐶𝐶 = 0 ∀ 𝑔𝑔𝐶𝐶𝐶𝐶 < 𝑀𝑀𝑀𝑀𝐶𝐶𝐶𝐶

On the other hand, precious metals in concentrates are paid only if their content is above the

minimum deduction levels, otherwise they are priced at zero. The final payable content of precious

metals is determined by multiplying the metal grade by a recovery factor.

0 , 𝑔𝑔𝑖𝑖 ≤ 𝑀𝑀𝑀𝑀𝑖𝑖
𝑝𝑝𝑝𝑝𝑖𝑖 = �
𝑔𝑔𝑖𝑖 × 𝑟𝑟𝑖𝑖 , 𝑔𝑔𝑖𝑖 > 𝑀𝑀𝑀𝑀𝑖𝑖

𝑖𝑖 = 𝐴𝐴𝐴𝐴, 𝐴𝐴𝐴𝐴

In addition, Dt is the sum of deductions applied, including copper Treatment Charges (TC) and

copper Refining Charges (RCu), as well as gold Refining Charges (RAu) and silver Refining Charges

(RAg).
𝐴𝐴𝐴𝐴
𝐷𝐷𝑡𝑡 = 𝑇𝑇𝑇𝑇𝑡𝑡 + 𝑅𝑅𝑡𝑡𝐶𝐶𝐶𝐶 + 𝑅𝑅𝑡𝑡𝐴𝐴𝐴𝐴 + 𝑅𝑅𝑡𝑡

Discounts linked to copper concentrates pricing, of which copper TC/RC represent the largest share,

are normally set in each concentrate supply contract agreement between the two parties of the

contract. Copper TC are quoted in US Dollars per tonne of concentrate, usually expressed on a dry

basis, while copper RC is quoted in US cents of dollars per pound of payable copper content.

In addition, gold and silver RC are applicable as long as there is gold or silver payable content in the

concentrate, gold RC being quoted in USD per troy ounce of payable gold content and silver RC in

USD per gram of payable silver content.

𝑅𝑅𝑡𝑡𝑖𝑖 = 𝑅𝑅𝑅𝑅𝑡𝑡𝑖𝑖 × 𝑝𝑝𝑝𝑝𝑖𝑖

Finally, Pt is the overall penalty for an n number of punishable elements in the concentrate which

negatively affect its quality as well as the chemical, mechanical and electrical properties of the
37
copper products made with that concentrate. These penalties are calculated as a function of the

excessive content of the punishable element beyond the penalty-free limit, below which the

concentrate would be exempt of penalty for that element.

𝑛𝑛

𝑃𝑃𝑡𝑡 = � 𝑝𝑝𝑖𝑖 ∀ 𝑡𝑡
𝑖𝑖

0, 𝑔𝑔𝑖𝑖 < 𝐹𝐹𝑖𝑖

𝑝𝑝𝑖𝑖 = �
(𝑔𝑔𝑖𝑖 − 𝐹𝐹𝑖𝑖 ) × 𝜌𝜌𝑖𝑖 , 𝑔𝑔𝑖𝑖 ≥ 𝐹𝐹𝑖𝑖

Where 𝑔𝑔𝑖𝑖 is the content of the punishable element i, Fi is the penalty-free limit content for that

element and 𝜌𝜌𝑖𝑖 is the penalty per unit of excessive content of the punishable element i beyond the

penalty-free limit. The unitary penalty is set as the nominal penalty 𝛿𝛿𝑖𝑖 , divided by the nominal

punishable increment of element i, hi.

𝛿𝛿𝑖𝑖
𝜌𝜌𝑖𝑖 = ∀ 𝑖𝑖
ℎ𝑖𝑖

38
 Chapter 3: Looking for an accurate forecasting of Copper TC/RC

Benchmark Levels

3.1. Introduction

There is a lack of research about the price at which mines sell copper concentrate to smelters, while

numerous investigations have instead focused on forecasting copper prices. Smelters however

obtain the copper they sell from the concentrate that most mines produce by processing the ore

which they have extracted. It therefore becomes necessary to thoroughly analyse the price at which

smelters buy the concentrates from the mines, besides the price at which they sell the copper. In

practice, this cost is set by applying discounts to the price of cathodic copper, the most relevant being

those corresponding to the smelters’ benefit margin (Treatment Charges-TC and Refining Charges-

RC).

The valuation of copper concentrates is a recurrent task undertaken by miners or traders following

processes in which market prices for copper and other valuable metals such as gold and silver are

involved, as well as relevant discounts or coefficients that usually represent a major part of the

revenue obtained for concentrate trading, smelting or refining. The main deductions are applied to

the market value of the metal contained by concentrates such as Copper Treatment Charges (TC),

Copper Refining Charges (RC), the Price Participation Clause (PP) and Penalties for Punishable

Elements (Soderstrom, 2008).

These are fixed by the different parties involved in a copper concentrate long-term or spot supply

contract, where TC/RC are fixed when the concentrates are sold to a copper smelter/refinery. The

sum of TC/RC is often viewed as the main source of revenue for copper smelters along with copper

premiums linked to the selling of copper cathodes. Furthermore, TC/RC deductions pose a concern

39
for copper mines as well as a potential arbitrage opportunity for traders, whose strong financial

capacity and in-depth knowledge of the market make them a major player (Crundwell, 2008, p.

491).

Due to their nature, TC/RC are discounts normally agreed upon taking a reference which is

traditionally set on an annual basis at the negotiations conducted by the major market participants

during LME Week every October and, more recently, during the Asia Copper Week each

November, an event that is focused more on Chinese smelters. The TC/RC levels set at these events

are usually taken as benchmarks for the negotiations of copper concentrate supply contracts

throughout the following year.

Thus, as the year goes on, TC/RC average levels move up and down depending on supply and

demand, as well as on concentrate availability and smelters’ available capacity. Consultants, such

as Platts, Wood Mackenzie and Metal Bulletin, regularly carry out their own market surveys to

estimate the current TC/RC levels. Furthermore, Metal Bulletin has created the first TC/RC index

for copper concentrates (“Metal Bulletin TC/RC Index,” 2017).

3.2. The need for accurate forecasts of Copper TC/RC

The information available for market participants may be regarded as sufficient to draw an accurate

assumption of market sentiment about current TC/RC levels, but not enough to foresee potential

market trends regarding these crucial discounts, far less as a reliable tool which may be ultimately

applicable by market participants to their decision-making framework or their risk-management

strategies. Hence, from an organisational standpoint, providing accurate forecasts of copper TC/RC

benchmark levels, as well as an accurate mathematical model to render these forecasts, is a research

40
topic that we have explored in depth. This is a question with undeniable economic implications for

traders, miners and smelters alike, due to the role of TC/RC in the copper trading revenue stream.

As a first step, our research seeks to determine an appropriate forecasting technique for TC/RC

benchmark levels for copper concentrates that meets the need of reliability and accuracy. To achieve

these three different and frequently applied techniques have been preselected from among the

options available in the literature. Then, their forecasting accuracy at different time horizons have

been tested and compared. These techniques (Geometric Brownian Motion -GBM-; the Mean

Reversion -MR- ;Linear Exponential Smoothing -LES-), have been chosen primarily because they

are common in modelling commodities prices and their future expected behaviour, as well as in

stock indices’ predictive works, among other practical applications (Pinheiro and Senna, 2016;

Savolainen, 2016; Zhang et al., 2017; Hloušková et al., 2018).

The selection of these models is further justified by the similarities shared by TC/RC with indices,

interest rates or some economic variables that these models have already been applied to. Also in

our view, the predictive ability of these models in commodity prices such as copper is a major asset

to take them into consideration. The models have been simulated using historical data of TC/RC

annual benchmark levels from 2004 to 2017 agreed upon during the LME Copper Week. The

dataset employed has been split into two parts, with two thirds as the in-sample dataset and one third

as the out-of-sample one.

With our study we aim to provide a useful and applicable tool to all parties involved in the copper

trading business to forecast potential levels of critical discounts to be applied to the copper

concentrates valuation process. To carry out our research, we have based ourselves on the following

premises: 1) GBM would deliver good forecasts if copper TC/RC benchmark levels vary randomly

41
over the years, 2) a mean-reverting model, such as the Ornstein-Uhlenbeck Process (OUP), would

deliver the best forecasts if TC/RC levels were affected by market factors and consequently they

move around a long-term trend, 3) a moving average model would give a better forecast than the

other two models if there were a predominant factor related to precedent values affecting the futures

ones of benchmark TC/RC.

In addition, we have also studied the possibility that a combination of the models could deliver the

most accurate forecast as the time horizon considered is increased, since there might thus be a

limited effect of past values, or of sudden shocks, on future levels of benchmark TC/RC. So, after

some time, TC/RC levels could be ‘normalised’ towards a long-term average.

The remainder of this chapter is structured as follows: Section 3 revises the related work on

commodity discounts forecasting and commodity prices forecasting techniques, as well as different

forecasting methods; Section 4 presents the reasoning behind the choice of each of the models

employed, as well as the historic datasets used to conduct the research and the methodology

followed; Section 5 shows the results of the simulations of the different models, comparing the

different error measures used to evaluate the best forecasting alternative for TC/RC benchmark

levels amongst all those presented; Section 6 contains the conclusions of the chapter and discusses

the results.

3.3. Related Work

The absence of any specific method in the specialised literature in relation to copper TC/RC leads

us to revisit previous literature in order to determine the most appropriate model to employ for our

purpose, considering those that have already been used in commodity price forecasting as the logical

starting point due to the application similarities that they share with ours.

42
Commodity prices and their forecasting have been a topic intensively analysed in much research.

Hence, there are multiple examples in literature with an application to one or several commodities,

such as Xiong et al. (2015), where the accuracy of different models was tested to forecast the interval

of agricultural commodity future prices; Shao and Dai (2018), whose work employs the

Autoregressive Integrated Moving Average (ARIMA) methodology to forecast food crop prices

such as wheat, rice and corn; and Heaney (2002), who tests the capacity of commodities future

prices to forecast their cash price were the cost of carry to be included in considerations, using the

LME Lead contract as an example research.

Table 8 summarizes similar research in the field of commodity price behaviours.

43
Table 8
Summary of previous literature research
AUTHOR RESEARCH
Shafiee & Topal (2010) Validates a modified version of the long-term trend reverting jump and dip diffusion model for
forecasting commodity prices and estimates the gold price for the next 10 years using historical monthly
data.
Li et al. (2012) Proposes an ARIMA-Markov Chain method to accurately forecast mineral commodity prices, testing the
method using mineral molybdenum prices.
Issler et al.(2014) Investigates several commodities’ co-movements, such as Aluminium, Copper, Lead, Nickel, Tin and
Zinc, at different time frequencies, and uses a bias-corrected average forecast method proposed by Issler
and Lima (Issler and Lima, 2009) to give combined forecasts of these metal commodities employing
RMSE as a measure of forecasting accuracy.
Hamid & Shabri (2017) Models palm oil prices using the Autoregressive Distributed Lag (ARDL) model and compares its
forecasting accuracy with the benchmark model ARIMA. It uses an ARDL bound-testing approach to
co-integration in order to analyse the relationship between the price of palm oil and its determinant
factors.
Duan et al. (2018) Predicts China’s crude oil consumption for 2015-2020 using the fractional-order FSIGM model.

Brennan & Schwartz (1985) Employs the Geometric Brownian Motion (GBM) to analyse a mining project’s expected returns
assuming it produces a single commodity.
McDonald & Siegel (1986) Uses GBM to model the random evolution of the present value of an undefined asset in an investment
decision model.
Zhang et al. (2015) Models gold prices using the Ornstein-Uhlenbeck Process (OUP) to account for a potentially existent
long-term trend in a Real Option Valuation of a mining project.

Sharma (2016) Forecasts gold prices in India with the Box Jenkins ARIMA method.

44
From the summary above, it is seen that moving average methods have become increasingly

popular in commodity price forecasting. The most broadly implemented moving average

techniques are ARIMA and Exponential Smoothing. Both consider the entire time series data as

well as do not assign the same weight to past values than those closer to the present as they are seen

as affecting greater to future values.

The Exponential Smoothing models’ predictive accuracy has been tested by Makridakis and Hibon

(1979), concluding that there are small differences between them (Exponential Smoothing) and

ARIMA models. Also, GBM and MR models have been intensively applied to commodity price

forecasting. Nonetheless, MR models present a significant advantage over GBM models which

allows them to consider the underlying price trend.

This advantage is of particular interest for commodities that, according to Dixit and Pindyck (1994)

– pp. 74, regarding the price of oil: “in the short run, it might fluctuate randomly up and down (in

responses to wars or revolutions in oil producing countries, or in response to the strengthening or

weakening of the OPEC cartel), in the longer run, it ought to be drawn back towards the marginal

cost of producing oil. Thus, one might argue that the price of oil should be modelled as a mean-

reverting process.”

3.4. Methodology

This section presents both the justification of the models chosen to forecast copper TC/RC

benchmark levels, as well as the core reference dataset used as inputs for each model compared in

the methodology. The dataset employed comprises publicly available annual benchmark levels for

copper TC/RC from 2004 to 2017, which are taken as market-wide reference to agree each years’

45
contracts specific TC/RC. The steps followed to carry out an effective comparison in terms of the

forecasting ability of each model are also laid out herein.

3.4.1. Models in Methodology

Geometric Brownian Motion (GBM)

GBM has been used in much earlier research as a way of modelling prices that are believed not to

follow any specific rule or pattern, hence seen as random. Black and Scholes (Black and Scholes,

1973) first used GBM to model stock prices and since then others have used it to model asset prices

as well as commodities, these being perhaps the most common of all, in which prices are expected

to increase over time, as does their variance (Zhang et al., 2017). Hence, following our first premise,

concerning whether TC/RC might vary randomly, there should not exist a main driving factor that

would determine TC/RC future benchmark levels and therefore GBM could to a certain extent be

a feasible model for them.

GBM can be written as a generalisation of a Wiener, a continuous time-stochastic Markov process,

with independent increments and whose changes over any infinite interval of time are normally

distributed, with a variance that increases linearly with the time interval (Dixit and Pindyck,

1994).process:

𝑑𝑑𝑑𝑑 = 𝛼𝛼𝛼𝛼𝛼𝛼𝛼𝛼 + 𝜎𝜎𝜎𝜎𝜎𝜎𝜎𝜎

Where, according to Marathe and Ryan (Marathe and Ryan, 2005), the first term is known as the

Expected Value, whereas the second is the Stochastic component, with α being the drift parameter

and σ the volatility of the process. Also, dz is the Wiener process which induces the abovementioned

stochastic behaviour in the model:

𝑑𝑑𝑑𝑑 = ∈𝑡𝑡 √𝑑𝑑𝑑𝑑 → ∈𝑡𝑡 ~𝑁𝑁(0,1)

46
The GBM model can be expressed in discrete terms according to Equation 3:

∆𝑥𝑥 = 𝑥𝑥𝑡𝑡 − 𝑥𝑥𝑡𝑡−1 = 𝛼𝛼𝑥𝑥𝑡𝑡−1 ∆𝑡𝑡 + 𝜎𝜎𝑥𝑥𝑡𝑡−1 𝜖𝜖√∆𝑡𝑡

In GBM percentage changes in x, ∆𝑥𝑥/𝑥𝑥 are normally distributed, thus absolute changes

in x, ∆𝑥𝑥 are lognormally distributed. Also, the expected value and variance for 𝑥𝑥(𝑡𝑡) are:

𝔼𝔼[𝑥𝑥(𝑡𝑡)] = 𝑥𝑥0 𝑒𝑒 𝛼𝛼𝛼𝛼

2
𝑣𝑣𝑣𝑣𝑣𝑣[𝑥𝑥(𝑡𝑡)] = 𝑥𝑥02 𝑒𝑒 2𝛼𝛼𝛼𝛼 (𝑒𝑒 𝜎𝜎 𝑡𝑡 − 1)

Orstein-Uhlenbeck Process (OUP)

Although GBM or ‘random walk’ may be well suited to modelling immediate or short-term price

paths for commodities, or for TC/RC in our case, it lacks the ability to include the underlying long-

term price trend should we assume that there is one. Thus, in accordance with our second premise

on benchmark TC/RC behaviour, levels would move in line with copper concentrate supply and

demand as well as the smelters’ and refineries’ available capacity to transform concentrates into

metal copper. Hence, a relationship between TC/RC levels and copper supply and demand is known

to exist and, therefore, is linked to its market price, so to some extent they move together coherently.

Therefore, in that case, as related works on commodity price behaviour such as Foo, Bloch and

Salim (Foo et al., 2018) do, we have opted for the MR model, particularly the OUP model.

The OUP process was first defined by Uhlenbeck and Orstein (Uhlenbeck and Ornstein, 1930) as

an alternative to the regular Brownian Motion to model the velocity of the diffusion movement of

a particle that accounts for its losses due to friction with other particles. The OUP process can be

regarded as a modification of Brownian Motion in continuous time where its properties have been

changed1 (Tresierra Tanaka and Carrasco Montero, 2016).

1
Stationary, Gaussian, Markov and stochastic process.

47
These modifications cause the process to move towards a central position, with a stronger attraction

the further it is from this position. As mentioned above, the OUP is usually employed to model

commodity prices and is the simplest version of a mean-reverting process (Dixit and Pindyck,

1994):

𝑑𝑑𝑑𝑑 = 𝜆𝜆(𝜇𝜇 − 𝑆𝑆)𝑑𝑑𝑑𝑑 + 𝜎𝜎𝑑𝑑𝑑𝑑𝑡𝑡

Where S is the level of prices, µ the long-term average to which prices tend to revert and 𝜆𝜆 the

speed of reversion. Additionally, in a similar fashion to that of the GBM, σ is the volatility of the

process and 𝑑𝑑𝑑𝑑𝑡𝑡 is a Wiener process with an identical definition. However, in contrast to GBM,

time intervals in OUP are not independent since differences between current levels of prices, S, and

long-term average prices, µ, make the expected change in prices, dS, more likely either positive or

negative.

The discrete version of the model can be expressed as follows:

1−𝑒𝑒 −2𝜆𝜆Δ𝑡𝑡
𝑆𝑆𝑡𝑡 = 𝜇𝜇(1 − 𝑒𝑒 −𝜆𝜆Δ𝑡𝑡 ) + 𝑒𝑒 −𝜆𝜆Δ𝑡𝑡 𝑆𝑆𝑡𝑡−1 + 𝜎𝜎� 𝑑𝑑𝑑𝑑𝑡𝑡
2𝜆𝜆

Where the expected value for 𝑥𝑥(𝑡𝑡) and the variance for (𝑥𝑥(𝑡𝑡) − 𝜇𝜇) are:

𝔼𝔼[𝑥𝑥(𝑡𝑡)] = 𝜇𝜇 + (𝑥𝑥0 − 𝜇𝜇)𝑒𝑒 −𝜆𝜆𝜆𝜆

𝜎𝜎2
𝑣𝑣𝑣𝑣𝑣𝑣[𝑥𝑥(𝑡𝑡) − 𝜇𝜇] = 2𝜆𝜆 �1 − 𝑒𝑒 −2𝜆𝜆𝜆𝜆 �

It can be derived from previous equations that as time increases prices will tend to long-term average

levels, µ. In addition, with large time spans, if the speed of reversion, λ becomes high, variance tends

to 0.

On the other hand, if the speed of reversion is 0 then 𝑣𝑣𝑣𝑣𝑣𝑣[𝑥𝑥(𝑡𝑡)] → 𝜎𝜎 2 𝑡𝑡 , making the process a

simple Brownian Motion.

48
Holt’s Linear Exponential Smoothing (LES)

Both GBM and MR are Markov processes which means that future values depend exclusively on

the current value, while the remaining previous time series data are not considered. On the other

hand, moving average methods employ the average of a pre-established number of past values in

different ways, evolving over time, so future values do not rely exclusively on the present, hence

behaving as though they had only a limited memory of the past.

This trait of the moving average model is particularly interesting when past prices are believed to

have a certain, though limited, effect on present values, which is another of the premise for this

research. Existing alternatives of moving average techniques pursue considering this ‘memory’

with different approaches. As explained by Kalekar (2004), Exponential Smoothing is suitable only

for the behaviours of a specific time series, thus Single Exponential Smoothing (SES) is reasonable

for short-term forecasting with no specific trend in the observed data, whereas Double Exponential

Smoothing or Linear Exponential Smoothing (LES) is appropriate when data shows a cyclical

pattern or a trend. In addition, seasonality in observed data can be computed and forecasted through

the usage of Exponential Smoothing by the Holt-Winters method, which adds an extra parameter

to the model to handle this characteristic.

Linear Exponential Smoothing models are capable of considering both levels and trends at every

instant, assigning higher weights in the overall calculation to values closer to the present than to

older ones. LES models carry that out by constantly updating local estimations of levels and trends

with the intervention of one or two smoothing constants which enable the models to dampen older

value effects. Although it is possible to employ a single smoothing constant for both the level and

the trend, known as Brown’s LES, to use two, one for each, known as Holt’s LES, is usually

preferred since Brown’s LES tends to render estimations of the trend ‘unstable’ as suggested by

49
authors such as Nau (2014). Holt’s LES model comes defined by the level, trend and forecast

updating equations, each of these expressed as follows, respectively:

𝐿𝐿𝑡𝑡 = 𝛼𝛼𝑌𝑌𝑡𝑡 + (1 − 𝛼𝛼)(𝐿𝐿𝑡𝑡−1 + 𝑇𝑇𝑡𝑡−1 )

𝑇𝑇𝑡𝑡 = 𝛽𝛽(𝐿𝐿𝑡𝑡 − 𝐿𝐿𝑡𝑡−1 ) + (1 − 𝛽𝛽)𝑇𝑇𝑡𝑡−1

𝑌𝑌�𝑡𝑡+𝑘𝑘 = 𝐿𝐿𝑡𝑡 + 𝑘𝑘𝑇𝑇𝑡𝑡

With α being the first smoothing constant for the levels and 𝛽𝛽 the second smoothing constant for

the trend. Higher values for the smoothing constants imply that either levels or trends are changing

rapidly over time, whereas lower values imply the contrary. Hence, the higher the constant, the more

uncertain the future is believed to be.

3.4.2. Models Comparison Method

The works referred to in the literature usually resort to different measures of errors to conduct the

testing of a model’s forecasting capacity regardless of the specific nature of the forecasted value, be

it cotton prices or macroeconomic parameters. Thus, the most widely errors used include Mean

Squared Error (MSE), Mean Absolute Deviation (MAD), Mean Absolute Percentage Error

(MAPE) and Root Mean Square Error (RMSE) (Makridakis et al., 1982; Heaney, 2002; Kalekar,

2004; Khashei and Bijari, 2010, 2011; Shafiee and Topal, 2010; Gómez-Valle and Martínez-

Rodríguez, 2013; Choudhury and Jones, 2014; Xiong et al., 2015). These error measures come

defined as follows:

Mean Square Error (MSE)

𝑛𝑛
1
𝑀𝑀𝑀𝑀𝑀𝑀 = �(𝑌𝑌�𝑖𝑖 − 𝑌𝑌𝑖𝑖 )2
𝑛𝑛
𝑖𝑖=1

Where 𝑌𝑌�𝑖𝑖 are the forecasted values and 𝑌𝑌𝑖𝑖 those observed.

50
Mean Absolute Deviation (MAD)

𝑛𝑛
1
𝑀𝑀𝑀𝑀𝑀𝑀 = ��𝑌𝑌�𝑖𝑖 − 𝑌𝑌��
𝑛𝑛
𝑖𝑖=1

Where 𝑌𝑌�𝑖𝑖 are the forecasted values and 𝑌𝑌� the average value of all the observations.

Mean Absolute Percentage Error (MAPE)

𝑛𝑛
1 �𝑌𝑌�𝑖𝑖 − 𝑌𝑌𝑖𝑖 �
𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 = �
𝑛𝑛 𝑌𝑌𝑖𝑖
𝑖𝑖=1

The above formula is expressed in parts-per-one and is the one used in the calculations conducted

here. Hence, multiplying the result by 100 would deliver percentage outcomes.

Root Mean Square Error (RMSE)

𝑛𝑛
1
𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 = √𝑀𝑀𝑀𝑀𝑀𝑀 = � �(𝑌𝑌�𝑖𝑖 − 𝑌𝑌𝑖𝑖 )2
𝑛𝑛
𝑖𝑖=1

In this research Geometric Brownian Motion (GBM), Ornstein-Uhlenbeck Process (OUP) and

Linear Exponential Smoothing (LES) models have been used to forecast annual TC/RC benchmark

levels, using all the four main measures of error mentioned above to test the predictive accuracy of

these three models. The GBM and OUP models have been simulated and tested with different step

sizes, while the LES model has been analysed solely using annual time steps.

The GBM and OUP models were treated separately to the LES model, thus GBM and OUP were

simulated using Monte Carlo (MC) simulations, whereas LES forecasts were simple calculations.

Monte Carlo simulations (MC) were carried out using Matlab software to render the pursued

51
forecasts of GBM and OUP models, obtaining 1000 simulated paths for each step size. MC

simulations using monthly steps for the 2013 – 2017 timespan were averaged every 12 steps to

deliver year forecasts of TC/RC benchmark levels.

On the other hand, forecasts obtained by MC simulations taking annual steps for the same period

were considered as year forecasts for TC/RC annual benchmark levels without the need for extra

transformation. Besides, LES model forecasts were calculated at different time horizons to be able

to compare the short-term and long-term predictive accuracy. LES forecasts were obtained for one-

year-ahead, hence using known values of TC/RC from 2004 to 2016; for two-year-ahead, stretching

the dataset from 2004 to 2015, and for five-year-ahead, thus using the same input dataset as for the

GBM and OUP models, from 2004 to 2012.

Finally, for every forecast path obtained, we have calculated the average of the squares of the errors

with respect to the observed values, MSE, the average distance of a forecast to the observed mean,

MAD, the average deviation of a forecast from observed values, MAPE, and the square root of

MSE, RMSE. The results of MSE, MAD, MAPE and RMSE calculated for each forecast path were

averaged by the total number of simulations carried out for each case. Averaged values of error

������, MAD
measures of all simulated paths, MSE �������, MAPE
��������, RMSE
��������, for both annual-step forecasts and

monthly-step forecasts have been used for cross-comparison between models to test predictive

ability at every step size possible.

Also, to test each model’s short-term forecasting capacity against its long-term forecasting capacity,

one-year-ahead forecast errors of the LES model were compared with the errors from the last year

of the GBM and OUP simulated paths. Two-year-ahead forecast errors of LES models were

compared with the average errors of the last two years of GBM and OUP, and five-year-ahead

52
forecast errors of LES models were compared with the overall average of errors of the GBM and

OUP forecasts.

3.4.3. Models Calibration

In a preliminary stage, the models were first calibrated to forecast values from 2013 to 2017 using

available data from 2004 to 2012 for TC and RC separately, hence disregarding the well-known

inherent 10:1 relation. The GBM and OUP models were calibrated for two different step sizes, -

monthly steps and annual steps- in order to compare forecasting accuracy with each step size in

each model. The available dataset is thus comprised of data from 2004 to 2017, which make the

calibration data approximately 2/3 of the total.

GBM Calibration

Increments in the logarithm of variable x are distributed as follows:

𝜎𝜎 2
∆(𝑙𝑙𝑙𝑙𝑙𝑙)~𝑁𝑁((𝛼𝛼 − )𝑡𝑡, 𝜎𝜎𝜎𝜎)
2

Hence, if m is defined as the sample mean of the difference of the natural logarithm of the time

series for TC/RC levels considered for the calibration and n as the number of increments of the

series considered, with n=9:

𝑛𝑛
1
𝑚𝑚 = �(𝑙𝑙𝑙𝑙𝑥𝑥𝑡𝑡 − 𝑙𝑙𝑙𝑙𝑙𝑙𝑡𝑡−1 )
𝑛𝑛
𝑡𝑡=1

𝑛𝑛
1
𝑠𝑠 = � �(𝑙𝑙𝑙𝑙𝑥𝑥𝑡𝑡 − 𝑙𝑙𝑙𝑙𝑙𝑙𝑡𝑡−1 − 𝑚𝑚)2
𝑛𝑛 − 1
𝑖𝑖=1

𝜎𝜎 2
𝑚𝑚 = 𝛼𝛼 −
2

𝑠𝑠 = 𝜎𝜎

53
OUP Calibration

The OUP process is an AR1 process (Dixit and Pindyck, 1994) whose resolution is well presented

by Woolridge (2011) using OLS techniques, fitting the following equation:

𝑦𝑦𝑡𝑡 = 𝑎𝑎 + 𝑏𝑏𝑦𝑦𝑡𝑡−1 + 𝜀𝜀𝑡𝑡

Hence, the estimators for the parameters of the OUP model are obtained by OLS for both TC and

RC levels independently:

𝑙𝑙𝑙𝑙𝑙𝑙
𝜆𝜆̂ = −
Δ𝑡𝑡
𝑎𝑎
𝜇𝜇̂ =
1 − 𝑏𝑏

2ln (1 + 𝑏𝑏)
𝜎𝜎� = 𝜀𝜀𝑡𝑡 �
(1 + 𝑏𝑏)2 − 1

LES Calibration

A linear regression is conducted on the input dataset available to find the starting parameters for the

LES model, the initial Level, 𝐿𝐿0 , and the initial value of the Trend, 𝑇𝑇0 , irrespective of TC values

and RC values. Here, as recommended by Gardner (2006), the use of OLS is highly advisable due

to the erratic behaviour shown by trends in the historic data, so the obtaining of negative values of

𝑆𝑆0 is prevented. Linear regression fulfils the following equations:

𝑌𝑌𝑡𝑡 = 𝑎𝑎𝑎𝑎 + 𝑏𝑏

𝐿𝐿0 = 𝑏𝑏

𝑇𝑇0 = 𝑎𝑎

By fixing the two smoothing constants, the values for the forecasts, 𝑌𝑌�𝑡𝑡+𝑘𝑘 , can be calculated at each

step using the model equations. There are multiple references in the literature on what is the

optimum range for each smoothing constant; Gardner (2006) speaks of setting moderate values for

54
both parameter less than 0.3 to obtain the best results. Examples pointing out the same may be found

in Brown (1963), Coutie (1964), Harrison (1967) and Montgomery and Johnson (1976). Also, for

many applications, Makridakis and Hibon (1979) and Chatfield (1978) found that parameter values

should fall within the range of 0.3-1. On the other hand, McClain and Thomas (1973) provided a

condition of stability for the non-seasonal LES model given by:

0 < 𝛼𝛼 < 2
� 4 − 2𝛼𝛼
0 < 𝛽𝛽 <
𝛼𝛼

Also, the largest possible value of α that allows the avoidance of areas of oscillation is proposed by

McClain and Thomas (1973) and McClain (1974):

4𝛽𝛽
𝛼𝛼 <
(1 + 𝛽𝛽)2

However, according to Gardner, there is no tangible proof that this value improves accuracy in any

form. Nonetheless, we have opted to follow what Nau (2014) refers to as ‘the usual way’, namely,

minimising the Mean Squared Error (MSE) of the one-step-ahead forecast of TC/RC for each input

data series previously mentioned. To do so, Matlab’s fminsearch function has been used with

function and variable tolerance levels of 1x10-4 as well as a set maximum number of function

iterations and function evaluations of 1x106 to limit computing resources. In Table 9 the actual

number of necessary iterations to obtain optimum values for smoothing constants is shown. As can

be seen, the criteria are well beyond the final results, which ensured that an optimum solution was

reached with assumable computing usage (the simulation required less than one minute) and with

a high degree of certainty.

55
 Table 9
Iterations on LES smoothing constants optimization
Function
Iteration
Evaluations
TC 1-Step 36 40
TC 2-Steps 249945 454211
TC 5-Steps 40 77
RC 1-Step 32 62
RC 2-Steps 254388 462226
RC 5-Steps 34 66

3.5. Analysis of Results

Monte Carlo simulations of GBM and OUP models render 1000 different possible paths for TC

and RC, respectively, at each time step size considered. Accuracy errors for both annual time steps

and averaged monthly time steps, for both GBM and OUP forecasts, were first compared to

determine the most accurate time step size for each model. In addition, the LES model outcome for

both TC and RC at different timeframes was also calculated and measures of errors for all the three

alternative models proposed at an optimum time step were finally compared.

3.5.1. GBM Forecasts

The first 20 of 1000 Monte Carlo paths for the GBM model with a monthly step size using Matlab

software may be seen in Figure 3 for both the TC and RC levels compared to their averaged paths

for every 12 values obtained. The tendency for GBM forecasts to steeply increase over time is easily

observable in the non-averaged monthly-step paths shown.

56
 Figure 3. The upper illustrations show the first 20 Monte Carlo paths for either TC or

RC levels using monthly steps. The illustrations below show the annual averages of

monthly step forecasts for TC and RC levels.

The average values of error of all 1000 MC paths obtained through simulation for averaged

monthly-step and annual-step forecasts are shown in Table 10 for both TC and RC discounts over

the period from 2013 to 2017, which may lead to preliminary conclusions in terms of an accuracy

comparison between averaged monthly steps and annual steps.

Table 10
Average of main measures of error for GBM after 1000 MC simulations
from 2013 to 2017.
������
𝑀𝑀𝑀𝑀𝑀𝑀 �������
𝑀𝑀𝑀𝑀𝑀𝑀 ��������
𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 ��������
𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅
3
TC Averaged Monthly Steps 5.60x10 46.98 0.50 56.38
TC Annual Steps 5.20x103 49.02 0.52 58.57
RC Averaged Monthly Steps 58.74 4.55 0.48 5.46
RC Annual Steps 50.85 4.91 0.52 5.84

57
However, a more exhaustive analysis is shown in Table 11, where the number of times the values

of each error measure is higher for monthly steps is expressed over the total number of MC

simulations carried out. The results indicate that better values of error are reached the majority of

times for averaged monthly-step simulations rather than for straight annual ones.

Table 11
Number of paths for which measures of error are higher
for monthly steps than for annual steps in GBM.
𝑀𝑀𝑀𝑀𝑀𝑀 𝑀𝑀𝑀𝑀𝑀𝑀 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅
TC 457/1000 455/1000 453/1000 453/1000
RC 439/1000 430/1000 429/1000 429/1000

In contrast, short term accuracy was also evaluated by analysing the error measures of one-year-

ahead forecasts (2013) in Table 12.

Table 12
Average for main measures of error for GBM after 1000 MC
simulations for 2013.
������
𝑀𝑀𝑀𝑀𝑀𝑀 �������
𝑀𝑀𝑀𝑀𝑀𝑀 ��������
𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 ��������
𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅
TC Averaged Monthly Steps 336.58 14.55 0.21 14.55
TC Annual Steps 693.81 20.81 0.30 20.81
RC Averaged Monthly Steps 2.97 1.38 0.20 1.38
RC Annual Steps 6.57 2.05 0.29 2.05

Two-year-ahead forecasts (2013–2014) are shown in Table 13. The results indicate, as one may

expect, that accuracy decreases as the forecasted horizon is widened, with the accuracy of averaged

monthly-step forecasts remaining higher than annual ones as found above for the five-year

forecasting horizon.

Table 13
Average for main measures of error for GBM after 1000 MC simulations
for 2013-2014.
������
𝑀𝑀𝑀𝑀𝑀𝑀 �������
𝑀𝑀𝑀𝑀𝑀𝑀 �������� 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅
𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 ��������
TC Averaged Monthly Steps 994.92 21.87 0.31 24.35
TC Annual Steps 1.33x103 28.33 0.40 31.04
RC Averaged Monthly Steps 10.35 2.40 0.34 2.71
RC Annual Steps 13.13 2.84 0.34 3.11

58
3.5.2. OUP Forecasts

Long-term levels for TC and RC, µ, the speed of reversion, λ, and the volatility of the process, σ,

are the parameters determined at the model calibration stage, which define the behaviour of the

OUP, shown in Table 14. The calibration was done prior to the Monte Carlo simulation for both

TC and RC, with each step size using available historical data from 2004 to 2012. The OUP model

was fitted with the corresponding parameters for each case upon MC simulation.

Table 14
OUP parameters obtained in the calibration process.
µ λ σ
5
TC Monthly 63.45 4.792x10 2.534
TC Annual 63.45 2.974 1.308
5
RC Monthly 6.35 4.763x10 2.519
RC Annual 6.35 2.972 1.305

Figure 4 shows the Mean Reversion MC estimations of the TC/RC benchmark values from 2013

to 2017 using monthly steps. The monthly forecasts were rendered from January 2012 through to

December 2016 and averaged every twelve values to deliver a benchmark forecast for each year.

The averaged results can be comparable to actual data as well as to the annual Monte Carlo

simulations following Mean Reversion. The lower-side figures show these yearly-averaged

monthly-step simulation outcomes which clearly move around a dash-dotted red line, indicating the

long-term run levels for TC/RC to which they tend to revert.

59
Figure 4. The first 20 Monte Carlo paths following the OUP model using monthly steps

are shown on the upper illustrations for either TC and RC. The annual averaged

simulated paths every 12 values are shown below.

The accuracy of monthly steps against annual steps for the TC/RC benchmark levels forecast was

also tested by determining the number of simulations for which average error measures became

higher. Table 15 shows the number of times monthly simulations have been less accurate than

annual simulations for five-year-ahead OUP forecasting by comparing the four measures of errors

proposed. The results indicate that only 25-32% of the simulations drew a higher average error,

which clearly results in a better predictive accuracy for monthly-step forecasting of TC/RC annual

benchmark levels.

60
 Table 15
Number of paths for which measures of error are
higher for monthly steps than for annual steps in
MR-OUP 5-Steps MC simulation.
𝑀𝑀𝑀𝑀𝑀𝑀 𝑀𝑀𝑀𝑀𝑀𝑀 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅
TC 311/1000 283/1000 266/1000 266/1000
RC 316/1000 281/1000 250/1000 250/1000

The averaged measures of errors obtained after the MC simulations of the OUP model for both

averaged-monthly-steps and annual steps giving TC/RC benchmark forecasts from 2013 to 2017

are shown in Table 16.

Table 16
Average for main measures of error for MR-OUP after 1000 MC
simulations 2013-2017.
������
𝑀𝑀𝑀𝑀𝑀𝑀 �������
𝑀𝑀𝑀𝑀𝑀𝑀 �������� 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅
𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 ��������
TC Averaged Monthly Steps 842.54 26.14 0.27 28.95
TC Annual Steps 1.14x103 29.31 0.31 33.25
RC Averaged Monthly Steps 8.40 2.61 0.27 2.89
RC Annual Steps 11.48 2.94 0.31 3.33

The error levels of the MC simulations shown above point towards a higher prediction accuracy of

averaged-monthly-step forecasts of the OUP Model, yielding an averaged MAPE value that is

12.9% lower for TC and RC 5-step-ahead forecasts. In regard to MAPE values, for monthly steps,

only 26.6% of the simulations rise above annual MC simulations for TC, and 25% for RC 5-step-

ahead forecasts, which further underpins the greater accuracy of this OUP set-up for TC/RC level

forecasts. A significant lower probability of higher error levels for TC/RC forecasts with monthly

MC OUP simulations is reached for the other measures provided. In addition, short-term and long-

term prediction accuracy was tested by comparing errors of forecasts for one-year-ahead in Table

17, two-year-ahead in Table 18, as well as for five-year-ahead error measures above in Table 16.

61
 Table 17
Average for main measures of error for MR-OUP after 1000 MC
simulations 2013.
������
𝑀𝑀𝑀𝑀𝑀𝑀 �������
𝑀𝑀𝑀𝑀𝑀𝑀 �������� 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅
𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 ��������
TC Averaged Monthly Steps 39.14 5.22 0.07 5.22
TC Annual Steps 366.43 15.55 0.22 15.55
RC Averaged Monthly Steps 0.39 0.51 0.07 0.51
RC Annual Steps 3.63 1.54 0.22 1.54

Table 18
Average for main measures of error for MR-OUP after 1000 MC
simulations 2013-2014.
������
𝑀𝑀𝑀𝑀𝑀𝑀 �������
𝑀𝑀𝑀𝑀𝑀𝑀 ��������
𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 ��������
𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅
TC Averaged Monthly Steps 371.65 15.67 0.18 19.00
TC Annual Steps 682.44 21.85 0.26 23.24
RC Averaged Monthly Steps 3.76 1.57 0.18 1.91
RC Annual Steps 6.89 2.19 0.26 2.43

With a closer forecasting horizon error, measures show an improvement of forecasting accuracy

when average monthly steps are used rather than annual ones. For instance, the MAPE values for

2013 forecast for TC is 68% lower for averaged monthly steps than for annual steps, also MAPE

for 2013 – 2014 were 30% lower for both TC and RC forecasts. Similarly, better values of error are

achieved for the other measures for averaged monthly short-term forecasts than in other scenarios.

In addition, as expected, accuracy is increased for closer forecasting horizons where the level of

errors shown above become lower as the deviation of forecasts is trimmed with short-term

predictions.

3.5.3. LES Forecasts

In contrast to GBM and OUP, the LES model lacks any stochastic component, so non-deterministic

methods such as the Monte Carlo are not required to obtain a forecast. Nonetheless, the LES model

relies on two smoothing constants which must be properly set in order to deliver accurate

predictions, hence the values of the smoothing constants were first optimised. The optimisation was

carried out for one-year-ahead forecasts, two-year-ahead forecasts, and five-year-ahead forecasts by

minimising the values of MSE for both TC and RC. The different values used for smoothing

62
constants, as well as the initial values for level and trend obtained by the linear regression of the

available dataset from 2004 through to 2012 are shown in Table 19.

Table 19
LES Model Optimised Parameters.
L0 T0 α β
TC 1 year 71.3611 -1.5833 -0.2372 0.1598
RC 1 year 7.1333 -0.1567 -0.2368 0.1591
TC 2 years 71.3611 -1.5833 -1.477x10-4 777.4226
RC 2 years 7.1333 -0.1567 -1.448x10-4 789.8336
TC 5 years 71.3611 -1.5833 -0.2813 0.1880
RC 5 years 7.1333 -0.1567 -0.2808 0.1880

Compared one-year-ahead, two-year-ahead and five-year-ahead LES forecasts for TC and RC are

shown in Figure 5, clearly indicating a stronger accuracy for shorter term forecasts as the observed

and forecasted plotted lines overlap.

Figure 5. One-step-ahead forecasts (2013), two-step-ahead forecasts (2013-2014) and

five-step-ahead forecasts (2013-2017) for TC and RC using the Linear Exponential

Smoothing model (LES).

63
Minimum values for error measures achieved through the LES model parameter optimisation are

shown in Table 20. The values obtained confirm the strong accuracy for shorter-term forecasts of

TC/RC seen in the figures.

Table 20
LES error measures for different steps-ahead forecasts.
������
𝑀𝑀𝑀𝑀𝑀𝑀 �������
𝑀𝑀𝑀𝑀𝑀𝑀 ��������
𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 ��������
𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅
-5
TC 1 year 1.0746x10 0.0033 4.6831x10-5 0.0033
-8 -4
RC 1 year 5.1462x10 2.2685x10 3.2408x10-5 2.2685x10-4
TC 2 years 189.247 9.7275 0.1057 13.7567
RC 2 years 1.8977 0.9741 0.1059 1.3776
TC 5 years 177.5531 7.8881 0.0951 10.8422
RC 5 years 1.1759 0.7889 0.0952 1.0844

3.6. Discussion and Model Comparison

The forecasted values of TC/RC benchmark levels could eventually be applied to broader valuation

models for copper concentrates and their trading activities, as well as to the copper smelters’ revenue

stream, thus the importance of delivering as accurate a prediction as possible in relation to these

discounts to make any future application possible. Each of the models presented may be a feasible

method on its own with eventual later adaptations to forecasting future values of benchmark TC/RC.

Nonetheless, the accuracy of these models as they have been used in this research requires, firstly,

a comparison to determine whether any of them could be a good standalone technique and,

secondly, to test whether a combination of two or more of them would deliver more precise results.

When comparing the different error measures obtained for all the three models, it is clearly

established that results for a randomly-chosen simulation of GBM or OUP would be more likely to

be more precise had a monthly-step been used to deliver annual forecasts instead of an annual-step

size. In contrast, average error measures for the entire population of simulations with each step size

employed showing that monthly step simulations for GBM and OUP models are always more

accurate than straight annual step forecasts when a shorter time horizon, one or two-year-ahead, is

64
 taken into consideration. However, GBM presents a higher level of forecasting accuracy when the

average for error measures of all simulations is analysed, employing annual steps for long-term

horizons, whereas OUP averaged-monthly-step forecasts remain more accurate when predicting

long-term horizons. Table 21 shows the error improvement for the averaged-monthly-step forecasts

of each model. Negative values indicate that better levels of error averages have been found in

straight annual forecasts than for monthly-step simulations.

Table 21
Error Average Improvement for averaged monthly steps before annual steps.

������
𝑀𝑀𝑀𝑀𝑀𝑀 �������
𝑀𝑀𝑀𝑀𝑀𝑀 ��������
𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 ��������
𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅
GBM TC/RC 1 Year 53.26%/56.53% 32.72%/35.87% 33.33%/35.48% 32.72%/35.87%
GBM TC/RC 2 Years 26.95%/26.35% 25.75%/15.97% 26.19%/2.86% 24.34%/13.69%
GBM TC/RC 5 Years -11.73%/-11.04% 8.73%/7.09% 9.26%/7.69% 7.47%/5.65%
OUP TC/RC 1 Year 88.08%/87.08% 63.71%/62.50% 62.91%/63.19% 63.68%/62.24%
OUP TC/RC 2 Years 46.28%/44.21% 27.71%/26.17% 29.99%/30.75% 22.16%/20.75%
OUP TC/RC 5 Years 26.02%/25.53% 10.93%/9.97% 13.12%/12.18% 13.06%/12.69%

Considering the best results for each model, and comparing their corresponding error measures, we

can opt for the best technique to employ among the three proposed in this paper. Hence, GBM

delivers the most accurate one-year forecast when averaging the next twelve-month predictions for

TC/RC values, as does the MR-OUP model. Table 22 shows best error measures for one-year-

ahead forecasts for GBM, OUP and LES models.

Table 22
TC/RC best error measures for 2013 forecasts after 1000 MC simulations.
������
𝑀𝑀𝑀𝑀𝑀𝑀 �������
𝑀𝑀𝑀𝑀𝑀𝑀 ��������
𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 ��������
𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅
GBM TC/RC 1 Year* 331.11/3.36 14.25/1.43 0.20/0.20 14.25/1.43
OUP TC/RC 1 Year* 41.05/0.42 5.36/0.54 0.0766/0.0777 5.36/0.54
1.07x10-5/ 0.003/ 4.68x10-5/ 0.003/
LES TC/RC 1 Year -8 -4
5.14x10 2.26x10 3.24x10-5 2.26x10-4
*Averaged monthly steps.

Unarguably, the LES model generates minimal error measures for one-year-ahead forecasts,

significantly less than the other models employed. A similar situation is found for two-year-ahead

forecasts where minimum error measures are also delivered by the LES model, shown in Table 23.

65
Table 23
TC/RC best error measures for 2013 – 2014 forecasts after 1000 MC simulations.
������
𝑀𝑀𝑀𝑀𝑀𝑀 �������
𝑀𝑀𝑀𝑀𝑀𝑀 ��������
𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 ��������
𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅
3
GBM TC/RC 2 Steps* 1.03x10 /10.06 22.00/2.42 0.31/0.34 24.47/2.71
OUP TC/RC 2 Steps* 373.31/3.76 15.73/1.58 0.1802/0.1813 19.00/1.91
LES TC/RC 2 Steps 189.24/1.89 9.72/0.97 0.1057/0.1059 13.75/1.37
*Averaged monthly steps.

Finally, accuracy measures for five-year-ahead forecasts of the GBM model might result in

somewhat contradictory terms for MSE reaching better values for annual steps than for averaged

monthly steps, while the other figures do better on averaged monthly steps. Addressing the

definition of MSE, this includes the variance of the estimator as well as its bias, being equal to its

variance in the case of unbiased estimators. Therefore, MSE measures the quality of the estimator

but also magnifies estimator deviations from actual values since both positive and negative values

are squared and averaged.

In contrast, RMSE is calculated as the square root of MSE and, following the previous analogy,

stands for the standard deviation of the estimator if MSE were considered to be the variance. Though

RMSE overreacts when high values of MSE are reached, it is less prone to this than MSE since it

is calculated as its squared root, thus not accounting for large errors as disproportionately as MSE

does.

Furthermore, as we have compared an average of 1000 measures of errors corresponding to each

MC simulation performed, the values obtained for average RMSE stay below the square root of

average MSE, which indicates that some of these disproportionate error measures are, to some

extent, distorting the latter. Hence, RMSE average values point towards a higher accuracy for GBM

five-year forecasts with averaged monthly steps, which is further endorsed by the average values of

MAD and MAPE, thus being the one used for comparison with the other two models as shown in

Table 24.

66
Table 24
TC/RC Best error measures for 2013 – 2017 forecasts after 1000 MC simulations.
������
𝑀𝑀𝑀𝑀𝑀𝑀 �������
𝑀𝑀𝑀𝑀𝑀𝑀 ��������
𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 ��������
𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅
3
GBM TC/RC 5 Steps* 6.00x10 /61.65 46.51/4.59 0.49/0.48 55.71/5.51
OUP TC/RC 5 Steps* 843.34/8.43 26.16/2.62 0.2715/0.2719 28.96/2.89
LES TC/RC 5 Steps 177.55/1.17 7.88/0.78 0.0951/0.0952 10.84/1.08
*Averaged monthly steps.

The final comparison clearly shows how the LES model outperforms the other two at all average

measures provided, followed by the OUP model in accuracy, although the latter more than doubles

the average MAPE value for LES.

The results of simulations indicate that measures of errors tend to either differ slightly or not at all

for either forecast of any timeframe. A coherent value with the 10:1 relation can then be given with

close to the same level of accuracy by multiplying RC forecasts or dividing TC ones by 10.

67
 Chapter 4: Estimating Copper Concentrates benchmark prices under

dynamic market conditions

4.1. Introduction

Copper concentrates are the primary product sold by copper mines to traders or smelters.

Their price is set in private agreements following unofficial market practices which, along

with the fact of the wide variety of concentrates layouts traded worldwide, make

problematic for any market participant to have a clear price reference for a specific

concentrate. Despite the large number of transactions of copper concentrates taking place,

there is no market-wide accepted concentrates price reference from an official, publicly

accessible and reliable source that can be taken as a benchmark. By the contrary, miners,

smelters and traders alike are forced to price concentrates following indirect references

and, sometimes, complex and poorly transparent calculation procedures.

In this research a copper concentrates benchmark price model was developed, which provides

suitable short-term price estimations based on metals and discounts forecasts, as well as on copper,

gold and silver spot and future prices data from the LME and COMEX. The model, which redeems

considerably low forecast error values for the short-term concentrate prices, constitutes a useful and

applicable tool for miners, traders and smelters to set a benchmark price level for their copper

concentrate transactions, also helping them optimize their operations, as well as estimate their

immediate liquidity needs or their actual necessity to hedge the price risks associated to their

concentrate trading.

68
In addition, five different concentrates layouts have been analysed to test the model’s behaviour

with the most common specification and blends of copper concentrates demanded by the market,

as well as to portray the model’s forecasting capacity and its ability to convey information on the

future relevance of the different components of pricing in the most frequent timeframe in which

copper concentrate trading takes place.

To address this our concentrates benchmark price model provides suitable short-term price

estimations as a function of the concentrate’s chemical specifications and the market conditions at

each moment. Our model puts together separate forecasts for the LME copper and COMEX gold

and silver spot prices, along with the main deductions and penalties applied by smelters, to achieve

copper concentrates price forecasts based on their layout.

The model resorts to historical data on spot and futures prices at different maturities for copper, gold

and silver as well as to historical values on copper TC/RC benchmark levels to render copper

concentrate price forecasts. These concentrate benchmark price model constitute a powerful source

of information for market participants, helping them to determine the price optimality of the copper

concentrates being traded as well as their future conditions. It also serves as an applicable tool for

miners, smelters and traders alike, allowing them to improve their production planning, to increase

the benefits from their trading operations, as well as helping them optimize their risk management

and their short-term liquidity needs.

4.2. Copper concentrates benchmark price model

Copper concentrates supply agreements set specific mechanisms to price the copper concentrates

to be traded according to their chemical specifications, as already explained in Chapter 2. Usually,

most pricing systems establish the price of copper concentrates as the sum of the payable elements

69
present in the concentrates minus the deductions and the penalties (Soderstrom, 2008; Seabridge

Gold, 2010; Teck, 2012, 2015). The model presented in this research is based on the same logic,

treating payable elements separately from discounts and punishable elements, though differing from

usual market practices in its forecasting capabilities for concentrates prices as metals and copper

TC/RC are separately modelled instead of employing the available market data, and their values are

projected ahead. Thus, at time t, the price of the ton of concentrate can be expressed as:

𝐶𝐶𝑡𝑡 = 𝑀𝑀𝑡𝑡 − 𝐷𝐷𝑡𝑡 − 𝑃𝑃𝑡𝑡

Though copper concentrates may contain multiple payable elements apart from copper, should

there be some gold and silver are the most common. Thus, Mt may be defined as the sum of the

amount paid for all payable elements, which is the amount paid for the copper, gold and silver

content in the concentrate. Metals in concentrates are paid at market price, though the full content is

not usually paid, as smelters are normally not capable of recovering all the content. Hence, the price

of payable metals in the concentrates is calculated as the payable content (pp) multiplied by its

market price at time t (𝑆𝑆𝑡𝑡 ).

𝑀𝑀𝑡𝑡 = 𝐶𝐶𝐶𝐶𝑡𝑡 + 𝐴𝐴𝐴𝐴𝑡𝑡 + 𝐴𝐴𝐴𝐴𝑡𝑡

𝐶𝐶𝐶𝐶𝑡𝑡 = 𝑝𝑝𝑝𝑝𝐶𝐶𝐶𝐶 × 𝑆𝑆𝑡𝑡𝐶𝐶𝐶𝐶

𝐴𝐴𝐴𝐴𝑡𝑡 = 𝑝𝑝𝑝𝑝𝐴𝐴𝐴𝐴 × 𝑆𝑆𝑡𝑡𝐴𝐴𝐴𝐴

𝐴𝐴𝐴𝐴
𝐴𝐴𝐴𝐴𝑡𝑡 = 𝑝𝑝𝑝𝑝𝐴𝐴𝐴𝐴 × 𝑆𝑆𝑡𝑡

Payable metal prices, including the market price for copper, gold and silver, are incorporated

separately with the inclusion of the Schwartz-Smith’s (2000) two-factor model, which decomposes

the log-market price of metals as the sum of two stochastic factors: 𝜒𝜒𝑡𝑡 , that comprises the short-

term deviation of prices to reflect rapid and temporary changes in market conditions, and 𝜉𝜉𝑡𝑡 ,

representing the long-term equilibrium price, thus including the long-term market dynamics in

prices, such as supply and demand changes.

70
 ln�𝑆𝑆𝑡𝑡𝑖𝑖 � = 𝜒𝜒𝑡𝑡𝑖𝑖 + 𝜉𝜉𝑡𝑡𝑖𝑖

𝑖𝑖 = 𝐶𝐶𝐶𝐶, 𝐴𝐴𝐴𝐴, 𝐴𝐴𝐴𝐴

Short-term price fluctuations are modelled to revert towards zero, following and Orstein-Uhlenbeck

process, whereas the long-term equilibrium price level is assumed to follow a Brownian motion

with drift.

𝑑𝑑𝜒𝜒𝑡𝑡𝑖𝑖 = −𝜅𝜅 𝑖𝑖 𝜒𝜒𝑡𝑡𝑖𝑖 𝑑𝑑𝑑𝑑 + 𝜎𝜎𝜒𝜒𝑖𝑖 𝑑𝑑𝑑𝑑𝜒𝜒𝑖𝑖

𝑑𝑑𝑑𝑑𝑡𝑡𝑖𝑖 = 𝜇𝜇 𝑖𝑖 𝑑𝑑𝑑𝑑 + 𝜎𝜎𝜉𝜉𝑖𝑖 𝑑𝑑𝑑𝑑𝜉𝜉𝑖𝑖

𝑑𝑑𝑑𝑑𝜒𝜒𝑖𝑖 𝑑𝑑𝑑𝑑𝜉𝜉𝑖𝑖 = 𝜌𝜌𝜒𝜒𝜒𝜒

𝑖𝑖
𝑑𝑑𝑑𝑑

𝑖𝑖 = 𝐶𝐶𝐶𝐶, 𝐴𝐴𝐴𝐴, 𝐴𝐴𝐴𝐴

Where 𝜅𝜅 is known as the short-term mean-reversion rate, 𝜎𝜎𝜒𝜒 and 𝜎𝜎𝜉𝜉 are the short-term and the

equilibrium volatilities, respectively, 𝜇𝜇𝜉𝜉 is the equilibrium drift rate, while 𝑑𝑑𝑑𝑑𝜒𝜒 and 𝑑𝑑𝑑𝑑𝜉𝜉 are two

correlated increments of standard Brownian motions with correlation parameter 𝜌𝜌𝜒𝜒𝜒𝜒 . According to

Schwartz-Smith (2000) the log of future spot prices is normally distributed, whereas the spot price

itself is log-normally distributed, being:

𝑆𝑆𝑡𝑡𝑖𝑖 = exp (𝜒𝜒𝑡𝑡𝑖𝑖 + 𝜉𝜉𝑡𝑡𝑖𝑖 )

1
Ε�𝑆𝑆𝑡𝑡𝑖𝑖 � = exp (Ε�ln (𝑆𝑆𝑡𝑡𝑖𝑖 )� + Var�ln�𝑆𝑆𝑡𝑡𝑖𝑖 ��)
2

1
ln�Ε�𝑆𝑆𝑡𝑡𝑖𝑖 �� = Ε�ln�𝑆𝑆𝑡𝑡𝑖𝑖 �� + Var�ln�𝑆𝑆𝑡𝑡𝑖𝑖 �� =
2

2 𝑖𝑖
𝑖𝑖 1 −2𝜅𝜅 𝑖𝑖 𝑡𝑡
�𝜎𝜎𝜒𝜒𝑖𝑖 � 𝑖𝑖 2 −𝜅𝜅 𝑖𝑖 𝑡𝑡
𝜌𝜌𝜒𝜒𝜒𝜒 𝜎𝜎𝜒𝜒𝑖𝑖 𝜎𝜎𝜉𝜉𝑖𝑖
= 𝑒𝑒 −𝜅𝜅 𝑡𝑡 𝜒𝜒0𝑖𝑖 + 𝜉𝜉0𝑖𝑖 + 𝜇𝜇𝜉𝜉𝑖𝑖 𝑡𝑡 + ��1 − 𝑒𝑒 � + �𝜎𝜎𝜉𝜉 � 𝑡𝑡 + 2 �1 − 𝑒𝑒 � �
2 2𝜅𝜅 𝑖𝑖 𝜅𝜅 𝑖𝑖

𝑖𝑖 = 𝐶𝐶𝐶𝐶, 𝐴𝐴𝐴𝐴, 𝐴𝐴𝐴𝐴

71
Hence, the payable amount for metals in concentrates can be rewritten using the notation of the two-

factor model as follows:

𝐶𝐶𝐶𝐶
+𝜉𝜉𝑡𝑡𝐶𝐶𝐶𝐶 �
𝐶𝐶𝐶𝐶𝑡𝑡 = 𝑝𝑝𝑝𝑝𝐶𝐶𝐶𝐶 𝑒𝑒 �𝜒𝜒𝑡𝑡

𝐴𝐴𝐴𝐴
+𝜉𝜉𝑡𝑡𝐴𝐴𝐴𝐴 �
𝐴𝐴𝐴𝐴𝑡𝑡 = 𝑝𝑝𝑝𝑝𝐴𝐴𝐴𝐴 𝑒𝑒 �𝜒𝜒𝑡𝑡

𝐴𝐴𝐴𝐴 𝐴𝐴𝐴𝐴
�𝜒𝜒 +𝜉𝜉𝑡𝑡 �
𝐴𝐴𝐴𝐴𝑡𝑡 = 𝑝𝑝𝑝𝑝𝐴𝐴𝐴𝐴 𝑒𝑒 𝑡𝑡

𝐶𝐶𝐶𝐶 𝐴𝐴𝐴𝐴 𝐴𝐴𝐴𝐴

+𝜉𝜉𝑡𝑡𝐶𝐶𝐶𝐶 � 𝐴𝐴𝐴𝐴
+𝜉𝜉𝑡𝑡𝐴𝐴𝐴𝐴 �
+ 𝑝𝑝𝑝𝑝𝐴𝐴𝐴𝐴 𝑒𝑒 �𝜒𝜒𝑡𝑡 +𝜉𝜉𝑡𝑡 �
𝑀𝑀𝑡𝑡 = 𝐶𝐶𝐶𝐶𝑡𝑡 + 𝐴𝐴𝐴𝐴𝑡𝑡 + 𝐴𝐴𝐴𝐴𝑡𝑡 = 𝑝𝑝𝑝𝑝𝐶𝐶𝐶𝐶 𝑒𝑒 �𝜒𝜒𝑡𝑡 + 𝑝𝑝𝑝𝑝𝐴𝐴𝐴𝐴 𝑒𝑒 �𝜒𝜒𝑡𝑡

Additionally, the payable copper content in concentrates is also dependent on the metal grade (𝑔𝑔),

multiplied by a recovery factor (𝑟𝑟), subject to a minimum deduction (MD).

𝑔𝑔𝐶𝐶𝐶𝐶 − 𝑀𝑀𝑀𝑀𝐶𝐶𝐶𝐶 , 𝑔𝑔𝐶𝐶𝐶𝐶 − 𝑀𝑀𝑀𝑀𝐶𝐶𝐶𝐶 < 𝑔𝑔𝐶𝐶𝐶𝐶 × 𝑟𝑟𝐶𝐶𝐶𝐶

𝑝𝑝𝑝𝑝𝐶𝐶𝐶𝐶 = � ∀ 𝑔𝑔𝐶𝐶𝐶𝐶 ≥ 𝑀𝑀𝑀𝑀𝐶𝐶𝐶𝐶
𝑔𝑔𝐶𝐶𝐶𝐶 × 𝑟𝑟𝐶𝐶𝐶𝐶 , 𝑔𝑔𝑖𝑖 − 𝑀𝑀𝑀𝑀𝑖𝑖 ≥ 𝑔𝑔𝐶𝐶𝐶𝐶 × 𝑟𝑟𝐶𝐶𝐶𝐶

𝑝𝑝𝑝𝑝𝐶𝐶𝐶𝐶 = 0 ∀ 𝑔𝑔𝐶𝐶𝐶𝐶 < 𝑀𝑀𝑀𝑀𝐶𝐶𝐶𝐶

On the other hand, precious metals in concentrates are paid only if their content is above the

minimum deduction levels, otherwise they are priced at zero. The final payable content of precious

metals is determined by multiplying the metal grade by a recovery factor.

0 , 𝑔𝑔𝑖𝑖 ≤ 𝑀𝑀𝑀𝑀𝑖𝑖
𝑝𝑝𝑝𝑝𝑖𝑖 = �
𝑔𝑔𝑖𝑖 × 𝑟𝑟𝑖𝑖 , 𝑔𝑔𝑖𝑖 > 𝑀𝑀𝑀𝑀𝑖𝑖

𝑖𝑖 = 𝐴𝐴𝐴𝐴, 𝐴𝐴𝐴𝐴

In addition, Dt is the sum of deductions applied, including copper Treatment Charges (TC) and

copper Refining Charges (RCu), as well as gold Refining Charges (RAu) and silver Refining Charges

(RAg).

𝐴𝐴𝐴𝐴
𝐷𝐷𝑡𝑡 = 𝑇𝑇𝑇𝑇𝑡𝑡 + 𝑅𝑅𝑡𝑡𝐶𝐶𝐶𝐶 + 𝑅𝑅𝑡𝑡𝐴𝐴𝐴𝐴 + 𝑅𝑅𝑡𝑡

72
Copper TC are quoted in US Dollars per tonne of concentrate, usually expressed on a dry basis,

while copper RC is quoted in US cents of dollars per pound of payable copper content. To apply

copper TC/RC discounts to concentrate pricing, we have assumed that they remain constant at

benchmark levels throughout their year of validity. TC/RC benchmark levels have been forecasted

using the Holt’s Linear Exponential Smoothing (LES) method, which delivers the best forecasts of

those analyzed in Chapter 3, treating TC and RC separately instead of observing the usual 10:1

relation. Holt’s LES model may be expressed by the combination of level, trend and forecast

updating equations, written as follows:

𝐿𝐿𝑡𝑡 = 𝛼𝛼𝑌𝑌𝑡𝑡 + (1 − 𝛼𝛼)(𝐿𝐿𝑡𝑡−1 + 𝑇𝑇𝑡𝑡−1 )

𝑇𝑇𝑡𝑡 = 𝛽𝛽(𝐿𝐿𝑡𝑡 − 𝐿𝐿𝑡𝑡−1 ) + (1 − 𝛽𝛽)𝑇𝑇𝑡𝑡−1

𝑌𝑌�𝑡𝑡+𝑘𝑘 = 𝐿𝐿𝑡𝑡 + 𝑘𝑘𝑇𝑇𝑡𝑡

In addition, gold and silver RC are applicable as long as there is gold or silver payable content in the

concentrate, gold RC being quoted in USD per troy ounce of payable gold content and silver RC in

USD per gram of payable silver content.

𝑅𝑅𝑡𝑡𝑖𝑖 = 𝑅𝑅𝑅𝑅𝑡𝑡𝑖𝑖 × 𝑝𝑝𝑝𝑝𝑖𝑖

Finally, Pt is the overall penalty for an n number of punishable elements in the concentrate which

negatively affect its quality as well as the chemical, mechanical and electrical properties of the

copper products made with that concentrate. These penalties are calculated as a function of the

excessive content of the punishable element beyond the penalty-free limit, below which the

concentrate would be exempt of penalty for that element.

𝑛𝑛

𝑃𝑃𝑡𝑡 = � 𝑝𝑝𝑖𝑖 ∀ 𝑡𝑡
𝑖𝑖

0, 𝑔𝑔𝑖𝑖 < 𝐹𝐹𝑖𝑖

𝑝𝑝𝑖𝑖 = �
(𝑔𝑔𝑖𝑖 − 𝐹𝐹𝑖𝑖 ) × 𝜌𝜌𝑖𝑖 , 𝑔𝑔𝑖𝑖 ≥ 𝐹𝐹𝑖𝑖

73
Where 𝑔𝑔𝑖𝑖 is the content of the punishable element i, Fi is the penalty-free limit content for that

element and 𝜌𝜌𝑖𝑖 is the penalty per unit of excessive content of the punishable element i beyond the

penalty-free limit. The unitary penalty is set as the nominal penalty 𝛿𝛿𝑖𝑖 , divided by the nominal

punishable increment of element i, hi.

𝛿𝛿𝑖𝑖
𝜌𝜌𝑖𝑖 = ∀ 𝑖𝑖
ℎ𝑖𝑖

4.3. Copper concentrates benchmark price forecasting

Copper concentrates daily prices have been modelled using Matlab software employing historical

daily official spot and futures data from the two main commodity markets (LME - copper; COMEX

- gold and silver) obtained from Refinitiv EIKON as well as historical copper TC/RC benchmark

levels from 2004 to 2018 to calibrate the model. Daily forecasts for each concentrate layout have

been obtained using Monte Carlo simulations performing 1000 trials for each concentrate to up to

one year ahead (2019).

This horizon has been chosen as the price of concentrates beyond a year may be of scarce interest

for market participants as most copper concentrate transaction occur in a shorter timeframe, whilst

this still allows to understand the model’s behaviour while key pricing components, such as metals

prices, evolve over time.

Payable metals prices have been modelled separately to each other, calibrating the model within the

in-sample timeframe of the datasets (2004-2018) using both spot and future data for each metal.

Metals prices forecasts are put together, along with the projected values of discounts and penalties,

to build the concentrate forecasts.

74
4.3.1. Parameters estimation

The whole set of parameters for the Schwartz-Smith two-factor model for metal prices

(𝜅𝜅 𝑖𝑖 , 𝜎𝜎𝜒𝜒𝑖𝑖 , 𝜎𝜎𝜉𝜉𝑖𝑖 , 𝜇𝜇 𝑖𝑖 , 𝜌𝜌𝜒𝜒𝜒𝜒

𝑖𝑖
) are estimated following the process proposed by Barlow et al. (2004), which

running the Kalman filter (Kalman, 1960) recursively, like the original approach by Schwartz and

Smith (2000), maximizing the log-likelihood function calculated with the parameters obtained from

the previous Kalman filter estimation over the model’s state-space form. The Schwartz-Smith two-

factor model used to obtain the market prices of payable metals in the concentrates is expressed in

its state-space form by the transition equation and the measurement equation, being the first:

𝑥𝑥𝑡𝑡 = 𝑐𝑐 + 𝐺𝐺𝑡𝑡−1 + 𝜔𝜔𝑡𝑡 , 𝑡𝑡 = 1, … , 𝑛𝑛𝑇𝑇

where,

𝑥𝑥𝑡𝑡 ≡ [𝜒𝜒𝑡𝑡 , 𝜉𝜉𝑡𝑡 ], 𝑎𝑎 2 × 1 𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 𝑜𝑜𝑜𝑜 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣;

𝑐𝑐 ≡ �0, 𝜇𝜇𝜉𝜉 ∆𝑡𝑡�, 𝑎𝑎 2 × 1 𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣;

−𝜅𝜅∆𝑡𝑡
𝐺𝐺 ≡ �𝑒𝑒 0� , 𝑎𝑎 2 × 2 𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚;
0 1

With 𝜔𝜔𝑡𝑡 being a 2 × 1 vector of serially uncorrelated, normally distributed disturbances with

E[𝜔𝜔𝑡𝑡 ] = 0 and Var[𝜔𝜔𝑡𝑡 ] = 𝑊𝑊 ≡ 𝐶𝐶𝑜𝑜𝑜𝑜[(𝜒𝜒∆𝑡𝑡 , 𝜉𝜉∆𝑡𝑡 )], ∆𝑡𝑡 the length of the time period and 𝑛𝑛𝑇𝑇 the

number of time periods in the dataset.

Being the measurement equation:

𝑦𝑦𝑡𝑡 = 𝑑𝑑𝑡𝑡 + 𝐹𝐹′𝑡𝑡 𝑥𝑥𝑡𝑡 + 𝑣𝑣𝑡𝑡 , 𝑡𝑡 = 1, … , 𝑛𝑛𝑇𝑇

where 𝑦𝑦𝑡𝑡 ≡ [ln 𝐹𝐹𝑇𝑇1 , … , ln 𝐹𝐹𝑇𝑇𝑇𝑇 ] is a 1 × 𝑛𝑛 vector of log future prices with time maturities

𝑇𝑇1 , 𝑇𝑇2 , … , 𝑇𝑇𝑛𝑛 ; 𝑑𝑑𝑡𝑡 ≡ [𝐴𝐴(𝑇𝑇1 ), … , 𝐴𝐴(𝑇𝑇2 )] is a 𝑛𝑛 × 1 vector, 𝐹𝐹𝑡𝑡 ≡ [𝑒𝑒 −𝜅𝜅𝑇𝑇1 1, … , 𝑒𝑒 −𝜅𝜅𝑇𝑇𝑛𝑛 1] is a 𝑛𝑛 × 2

matrix and 𝑣𝑣𝑡𝑡 is a 𝑛𝑛 × 1 vector of serially uncorrelated, normally distributed disturbances with

E[𝑣𝑣𝑡𝑡 ] = 0 and 𝐶𝐶𝐶𝐶𝐶𝐶[𝑣𝑣𝑡𝑡 ] = 𝑉𝑉.

75
The parameters estimation procedure has been run independently for each payable metal using spot

and futures data from 2004 to 2018. The different maturities of contracts considered for the

parameters estimation for each metal are shown in Table 25. The maximum number possible of

maturities of contracts have been considered to calibrate the models in order to increase the metal’s

price forecasts reliability and longer-term accuracy. Furthermore, according to Schwartz-Smith

(2000), the model’s state variable may ultimately be estimated within a range of around ±8% if only

spot data are used, whereas if multiple future contracts with different maturities are used the

uncertainty will be reduced.

Table 25
Contracts considered for parameters estimation.
Exchange Contracts
Cu LME Cash, 3M, 15M, 27M, 63M
Au COMEX Cash, 2M, 3M, 6M, 9M, 12M, 15M, 18M, 20M
Ag COMEX Cash, 2M, 3M, 6M, 9M, 12M, 15M, 18M, 20M

4.4. Metals and Copper TC/RC Forecasts

In-sample daily metals prices from 2004 to 2018 along with their estimation using

Schwartz and Smith’s (2000) model with parameters adjusted through the Kalman filter

maximum likelihood recursive method proposed by Barlow (2004) are shown in Figure

6.

Figure 6. Observed and Schwartz and Smith’s two-factor model estimation for 2004-

2018 daily metals prices.

76
The optimum parameters estimations obtained using historical daily data of spot and futures prices

for each metal are shown in Table 26, along with the different error estimations for the measurement

equation with the future contracts employed with each metal at every maturity. Values of the log

likelihood function have been maximized to achieve the parameters that render the best forecast for

each commodity, thus enhancing the model fitting to historical data and forecasting capacity of

metal prices and of concentrates prices consequently.

Table 26
Parameters values for Schwartz and Smith’s model for Copper, Gold
and Silver obtained through the Kalman Filter and maximum
likelihood using futures and spot prices data from 2004 to 2018.
Copper Gold Silver
𝜅𝜅 0.2427 0.0440 0.0455
𝜎𝜎𝜒𝜒 0.2418 0.4369 0.4139
𝜇𝜇 0.0958 0.0606 -0.2718
𝜎𝜎𝜉𝜉 0.3121 0.4660 0.4230
𝜌𝜌𝜒𝜒𝜒𝜒 -0.2889 -0.8852 -0.5549
4 5
𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 5.5948 × 10 1.3311 × 10 1.2112 × 105
s1 8.8710 × 10−5 1.155 × 10−5 5.038 × 10−6
s2 1.3519 × 10 −4
2.401 × 10 −7
1.017 × 10−5
s3 1.000 × 10−7 2.757 × 10−6 1.114 × 10−5
s4 0.0040 1.865 × 10 −6
8.453 × 10−6
s5 1.047 × 10−6 2.549 × 10−6
s6 2.044 × 10−4 6.944 × 10−5
s7 0.002 5.695 × 10−4
s8 0.0055 0.0236

On the other hand, the LES model’s forecasts for 2019 are 84.1247 USD/dmt for TC and 8.4125

USc/lb for RC. As it may be seen, the usually observed 10:1 historical relation in TC/RC levels

slightly differs in the values obtained, as TC and RC have been modeled individually. The LES

model has been calibrated following the procedure explained in Chapter 3, minimizing the Mean

Absolute Percentage Error (MAPE) for the one-year-ahead forecast, achieving a value of 4.12%.

77
4.5. Copper concentrates benchmark price forecasts and errors

The five different copper concentrates, whose layouts have been defined in Chapter 2, have been

used to test the model’s fitting errors and forecasting accuracy. As previously explained, these five

concentrates have been used to be able to check the model’s behaviour in common conditions,

pricing regular clean concentrates, as well as complex concentrates, having high levels of

punishable elements, or concentrates with meaningful levels of gold or silver.

Figure 7. Estimations of concentrates 1 and 2 prices compared to objective price for the

in-sample period (2004 - 2018).

Model price estimations for the in-sample period (2004 – 2018) for concentrates 1 and 2 compared

to their objective price calculated with spot data are shown on Figure 6. In addition, fitting errors for

the five concentrates considered are shown on Table 27. Error values have been measured with

respect to their market price calculated using historical spot metals prices during the same historical

period.

78
Table 27
Fitting errors for concentrates price estimation (2004 – 2018).
C1 C2 C3 C4 C5
MD -4.0332 -0.9911 -1.8450 -3.6765 -5.3456
MAD 10.2647 10.1856 10.1189 10.7307 14.1090
MSE 279.4339 236.0696 243.5428 299.6983 525.5189
RMSE 16.7163 15.3646 15.6059 17.3118 22.9242
MPE -0.0018 −4.0298 × 10 −5
−4.0160 × 10−4 -0.0014 -0.0017
MAPE 0.0071 0.0041 0.0046 0.0064 0.0069

For the 2019 out-of-sample forecasts of concentrates prices we have performed 1000 simulations

for each of the five concentrates during the period considered. The first forecast obtained for every

concentrate is shown on Figure 8 compared to the objective concentrates prices for the same year,

as well as to the averages of the 1000 simulations for the one-year ahead daily prices forecasts for

each concentrate considered, which represent the expected values for the concentrates’ prices by the

model over time.

Figure 8. 2019 Model's forecasts for concentrates, expected prices and objective spot

prices.

79
 Short-term and longer-term average forecasting error measures achieved for each concentrate may

also be seen on Table 28. Forecasting error measures have also been calculated by comparing

forecasted values with concentrates prices calculated for the year 2019 using SPOT prices data of

payable metals.

Table 28
Errors for concentrates price forecasting.
C1 C2 C3 C4 C5 Average
MD -1.452 4.320 2.816 -0.365 -1.554 0.753
MAD 21.448 36.414 32.476 25.591 30.707 29.327
MSE 699.021 2006.750 1598.975 996.447 1433.446 1346.928
1 day
RMSE 21.448 36.414 32.476 25.591 30.707 29.327
MPE -0.001 0.001 0.001 0.000 -0.001 0.000
MAPE 0.016 0.012 0.013 0.015 0.015 0.014
MD -26.908 -12.971 -16.123 -25.393 -35.894 -23.458
MAD 45.171 62.898 57.549 50.514 63.286 55.883
MSE 3403.150 6694.986 5595.886 4278.982 6694.282 5333.457
5 days
RMSE 45.171 62.898 57.549 50.514 63.286 55.883
MPE -0.020 -0.004 -0.006 -0.015 -0.018 -0.013
MAPE 0.033 0.021 0.023 0.030 0.032 0.028
MD -13.209 3.010 -0.883 -10.518 -16.861 -7.692
MAD 73.552 120.444 107.985 86.825 104.925 98.746
MSE 9419.109 25761.431 20670.874 13223.143 19216.173 17658.146
1 month
RMSE 73.552 120.444 107.985 86.825 104.925 98.746
MPE -0.010 0.001 0.000 -0.006 -0.009 -0.005
MAPE 0.053 0.040 0.042 0.050 0.052 0.048
MD 41.757 64.676 58.815 48.841 59.422 54.702
MAD 134.769 220.363 197.857 159.356 192.387 180.946
3 MSE 31407.277 84012.894 67726.073 43933.814 64014.626 58218.937
months RMSE 134.769 220.363 197.857 159.356 192.387 180.946
MPE 0.027 0.020 0.022 0.026 0.027 0.025
MAPE 0.092 0.070 0.075 0.087 0.090 0.083
MD -2.881 4.668 3.158 -1.298 -3.251 0.079
MAD 183.731 301.582 270.094 217.130 262.186 246.945
6 MSE 60179.870 159490.231 128294.522 83681.170 122385.032 110806.165
months RMSE 183.731 301.582 270.094 217.130 262.186 246.945
MPE -0.004 0.001 0.001 -0.002 -0.004 -0.002
MAPE 0.128 0.097 0.103 0.121 0.126 0.115
MD -100.988 -23.680 -41.838 -90.147 -132.711 -77.873
MAD 267.058 424.151 378.967 309.632 378.378 351.637
MSE 137882.891 322525.588 260476.025 181401.640 274961.672 235449.563
1 year
RMSE 267.058 424.151 378.967 309.632 378.378 351.637
MPE -0.076 -0.008 -0.016 -0.052 -0.068 -0.044
MAPE 0.193 0.133 0.142 0.175 0.188 0.166

80
MAPE values for concentrates price forecasts barely exceed 1.5% for the one-day ahead horizon,

and 5% for the one-month ahead. On the most frequent horizons for copper concentrates trading,

the one-month ahead and the three-month ahead, average MAPE values of the model varies from

4.8% for the first and 8.3% for the latter, thus giving reasonable estimations over the final price that

market participants would pay for concentrates, as well as a clearer picture of the level of

commodity-price-related risk their taking, their short-term liquidity needs or their hedging

necessities, among other concerns they face. Additionally, non-existing references in literature on

copper concentrate pricing, making this the first research to address the lack of an appropriate model

for this, impedes comparing the values of errors achieved with any pre-existing benchmark.

The influence of high gold or silver grade is clearly a driving factor as may be understood by the

prime paid for C2 and C3 in comparison to the rest of the concentrates. We also find that the

presence of high levels of gold and silver in the concentrate act as a stabilizing factor for concentrates

prices as precious metals tend to stay less volatile than copper. This has a direct impact in the

model’s accuracy when forecasting the future prices of these kind of concentrates. Thus, MAPE

values for C2 and C3 for the three-month horizon stay between 7.0-7.5%, whereas it goes above

8.5% for C4 and 9.0% for C1 and C5. This effect is amplified as the forecasting horizon is extended,

reaching differences in MAPE values of 2.4 percentage points between C3 and C5 and 6.0

percentage points between C1 and C2 for the one-year horizon.

Taking the full scope of concentrates layout considered and using MAPE as the benchmark to

compare the model’s forecasting ability for a particular concentrate, the group of those presenting

higher levels of precious metals are forecasted more accurately by the model (C2 and C3). On the

other hand, the price of those with lower levels of precious metals are tougher to forecast by the

model (C4, C5 and C1). By the contrary, RMSE values for the different concentrates indicate a

81
larger dispersion of residuals for C2 and C3, as well as the opposite for the others, at every horizon

studied, indicating that the influence of high levels of gold and silver, whilst stabilizing the average

forecast, may be tending to introduce a higher variance in individual predictions.

In addition, high values of copper may even compensate the amounts of penalties due to punishable

elements in the concentrate and make them commercially feasible as is the case of C5, grading over

8 percentage points more of copper than C4. Nevertheless, the difference in value between highly

complex concentrates due to high levels of payable elements, such as C4, and clean concentrates

with an average content of payable elements, such as C1, might not be sufficient to compensate the

complexity and additional costs incurred in treating the most complex ones at specialized smelters

versus the cleanest. In this case, this difference in value is an average of 24.4% more for C4 than C1

for the one-year ahead forecast.

4.2.1. Effects of discounts and punishable elements over the price of concentrates

The value of the discounts as well as the penalties due to punishable elements and the total amount

of deductions applied expressed in USD per dry metric tonne of concentrate for the year 2019 (one-

year ahead forecast) may be seen in Table 28 for each of the concentrates considered. According

to our results, the impact of punishable elements on the overall deduction, and thus on the final price

paid for concentrates is considerable, as they just represent 8.35% of the total amount subtracted for

C1, which is regarded as a clean concentrate, while it is a 30.43% for C5, with a net increment of

54.02 USD/dmt, over four times as much as for the previous case.

Table 28
Discounts and Penalties in USD/dmt of concentrate for the 2019 forecast.
C1 C2 C3 C4 C5
Cu TC 84.12 84.12 84.12 84.12 84.12
Cu RC 47.52 50.26 49.33 50.69 65.70
Au RC 0.00 8.98 6.04 1.53 0.52
Ag RC 0.00 0.67 2.43 1.06 0.63
Penalties 12.00 9.30 14.83 55.93 66.02
Total 143.64 153.34 156.76 193.33 216.99

82
Notwithstanding, the copper content in C5 is noticeably higher than for C1 which presumably to

some extent offsets the increase in penalties and in copper RC. On the other hand, regarding C4, the

penalties have increased by 366% compared to the first concentrate, now accounting for 28.93% of

the overall deductions, while the copper content has barely increased from 26.62% to 28.33%. This

could even make C4 less valuable than C1 if it were not for the important difference in gold and

silver content which helps boost its price.

Finally, the discounts’ relative weight with respect to payable metals content fluctuates over time

for a specific concentrate layout as metal prices and discounts themselves also do. The forecasted

relative weight of the overall discounts on the payable metal content of each of the concentrates is

represented by the boxplots in Figure 9.

Figure 9. Discounts’ relative weight in payable metal content in concentrate.

The mean value of the discounts’ relative weights with respect to the payable metals content for

each concentrate is shown by the middle line of the boxplots, while the lower and upper limits

83
represent the 25th and 75th percentiles, respectively. Also, the whiskers stretch to the furthermost

values in the forecasted timeseries that have not been considered as outliers.

84
 Chapter 5: Managing a High Uncertainty Scenario through a Real

Option Assessment: Evidence from a Copper Concentrate Trader

5.1. Introduction

Most commodity traders participate to some extent in the extraction or processing stages of the

commodities supply chain. However, independent traders, also known as trading houses, neither

produce nor process the commodities that they trade with, and yet they play a major role in

commodity markets. Independent traders buy and sell a commodity seeking to take advantage of

potentially profitable arbitrage opportunities in time and space, connecting primary producers to

processors. Their activity normally involves bespoke agreements to ensure supply, while mastering

logistics and simultaneously merging financial markets with physical transactions to mitigate risks

and ensure margins.

As a subset of commodity traders, metal traders focus on the most universally used industrial and

precious metals, not limiting themselves to the refined form of the metals, but also engaging in the

trading of unrefined forms, semi-processed or raw ores. For the case of copper, copper concentrate

is the main form of unrefined copper and the primary product sold by most copper miners, being

massively traded by independent trading houses.

Traders acquire copper concentrate from miners, taking the other side of the trade from smelters

and refiners, which process concentrates to obtain copper cathodes. Their role is predominant in the

global copper market. An example is Glencore, which in 2020 traded 3.4 million tons of metal

copper and concentrates, almost as much as the mining production of Peru and China combined,

the second and third leading copper producers in the world respectively. Meanwhile, its main

85
competitor, Trafigura, traded 11.1 million tons of copper, zinc, lead and nickel concentrates as well

as alumina and cobalt hydroxide just in 2020 alone (Trafigura Beheer B.V., 2021).

Consequently, in this chapter we analyse the copper concentrate trading business resorting to the

Real Options methodology to assess the high volatility and uncertainty which dominates in this area

of commodity markets. The results obtained indicate that this method can be applied by copper

concentrate traders to separate the price uncertainty from their business operative and financial

planning for the future. This provides them with a reliable tool to lower the level of uncertainty

affecting their long-term goals, as well as keeping the risk that they are taking under control. It also

gives them higher managerial flexibility.

5.2. Related Work

Traditionally, commodity trading houses have developed their activities in utter discretion, having

little incentive to disclose their most sensitive information, if any at all. Hence, details related to the

conditions of the agreements that traders reach with either primary producers or their clients, the

pricing mechanisms for concentrates, as well as for most of the commodities they trade with, or the

structure of some usual financial tools that they employ such as the pre-payment and pre-finance

agreements, are widely unknown except by those directly involved in commodity trading.

This lack of understanding of the behaviour of commodity traders has not been straightforwardly

addressed in the academic literature either. This has mainly focused on studying commodities

pricing methods, such as Brennan and Schwartz (1985), Schwartz and Smith (2000), Liu et al.

(2017), and Dehghani and Bogdanovic (2018), who analyse different alternatives to forecast copper

prices as well as assessing the value of commodity-related projects. Jia and Kang (2022) formulate

86
two claims about spot and future price predictions by investigating six industrial metals traded in

the London Metal Exchange (LME).

Furthermore, a significant number of works have instead centred on evaluating the feasibility of

commodity-related investments, such as Sarkis and Tamarkin’s (2005), which applies Real Options

analysis to a particular petroleum project that would generate greenhouse gases emissions permits.

Many works have employed Real Options as their main tool to assess a certain project’s feasibility

and its long-term value. Real Options are quite suitable to determine the value of projects which rely

on commodities, either energy, mineral, agricultural or soft commodities, and which require the

estimation of financial statements. This method also helps to draw more accurate financial plans as

it allows considering a stretch of different possible scenarios.

Many researchers have employed Real Options specifically focused on mining projects, as

summarised by Savolainen (Savolainen, 2016), such as Inthavongsa et al. (Inthavongsa et al., 2016),

who employ a real-option approach to analyse a hypothetical gold mine project. Similar research

by Zhang et al. (2015) evaluated a gold mine with price uncertainty as well as analysing its optimal

long-term production strategy. This has been extensively examined in the literature (Ajak and

Topal, 2015; Aminrostamkolaee et al., 2017; Guj and Chandra, 2019; Hazra et al., 2019; Siña and

Guzmán, 2019).

On the other hand, another branch of research has focused on valuing companies that extract,

process and trade with commodities without considering the trading aspect of their business as a

standalone element, or even as a key aspect for their profitability or for their company value (Kaiser,

2013; 2015). In this sense, Misund et al. (2015) valued a series of US and non-US oil and gas

companies considering the primary factors affecting production. Sabet and Heaney (2017)

87
examined the relation between oil and gas firms share prices, and the return, volatility and drilling

activity of crude oil and natural gas, neither having taken into consideration the trading side.

Moreover, Amram and Kulatilaka (1998) focused on using Real Options to manage the uncertainty

of investments. Barton and Lawryshyn (2011) fit managerial cash flows estimates to a continuous

cash flow process with changing growth and volatility parameters. Damodaran (2012) had in large

part laid the basis for most of the currently-used investment valuation tools.

This chapter focuses on the copper concentrate trading business. Here the degree of uncertainty is

notably higher than for other commodities because of the absence of a price reference for copper

concentrates, which prevents market participants from discovering price outright, opposite to what

happens with copper cathodes for instance. It also very seriously hinders their ability to hedge their

exposure to the commodity. In contrast to all the previous research, this research analyses copper

concentrate trading as a single and independent element of commodity markets. This is done

through its highest exponent, represented by independent commodity traders, tackling the

remarkable void in the literature on commodity traders’ operations, financials, and inner mechanics,

particularly regarding copper concentrates.

The copper concentrate trading business model is reviewed in our research through the valuation of

a standardised copper concentrate trading house, which neither produces nor processes any copper

concentrate, using classical discounted cash flows valuation techniques and Real Options. The main

factors affecting the activity of copper concentrate trading are considered stochastically, resorting to

publicly available historical data on spot prices for copper, gold and silver from COMEX and the

London Metal Exchange (LME), and forecasted five years ahead to assess the trader’s business

future performance.

88
The methodology laid out in this chapter constitutes a first approach to commodity traders’

dynamics, albeit a full-scale analysis would involve taking into account long-term hedging practices

with futures and options. Nonetheless, this procedure enables assessing the impact that strategical

decisions have on the future financial performance of any commodity trader, which is also subdued

by commodities price fluctuations, among other factors, if an appropriate hedge is not implemented.

Additionally, even though we have focused on copper concentrate trading, the process developed

in this paper may well be applicable to commodity traders regardless of the commodity or

commodities that they are trading with. This is because its main utility is to provide continuous

managerial flexibility under very volatile contexts, such as those which unfold in commodity

markets.

5.3. Materials and Method

To assess the valuation of the copper concentrate trading company a five-year timeframe has been

set from the years 2021 to 2025. 2020 being the last full fiscal year completed by the trader. The

reference economical and financial data employed to develop the projections have been based on a

commodity trader’s 2020 annual income statement and balance sheet. Given the complexity of

copper concentrate physical trading, the starting information has been limited to exclusively

portraying the copper concentrate trades carried out by the company, not the remaining possible

associated deals that it might have conducted. On the other hand, exchange rate fluctuations in the

long run have not been accounted for to model the behaviour and profitability of the trader’s activity,

thus assuming a perfectly currency-hedge trader, with null foreign currency exposure.

5.3.1. Concentrate Price

The trader’s activity has been modelled employing a standard concentrate whose chemical layout

has been detailed in Table 29. The concentrate’s specifications were set in accordance with the

89
statistical classification described in Delbeke and Rodriguez (2014), fitting the concentrate used

within the 60th percentile of this classification, containing low levels of gold and silver.

Table 29
Copper Concentrate Specifications. (Source: Own elaboration
based on Delbeke and Rodriguez (2014).
% Ppm g/DMT
Cu 27.25%
Au 1.50
Ag 30
As 0.002% 20
Sb 0.002% 20
Bi 0.003% 30
Cd 0.005% 50
Cl 0.005% 49
F 0.021% 210
Pb 0.040% 400
Hg 0.000% 0
Ni 0.004% 40
Co 0.009% 90
Se 0.002% 20
Te 0.001% 10
Zn 0.220% 2200
Humidity 8.00%

The expected annual prices for the copper concentrate in each period, t, comprising the years from

2021 to 2025, have been obtained by averaging out 500 individual daily price paths for the copper

concentrate over this timespan. The selling price for the concentrate on each day d, expressed in

United States Dollars per dry metric ton (USD/DMT), is determined by the price model developed

in Chapter 4.

To obtain the concentrate daily price paths over the forecasting horizon we have employed a slightly

simplified, less computing-demanding procedure than the one exposed in Chapter 4. Hence 500

individual daily price paths for Copper, Gold and Silver have been first obtained, following the

process shown in Figure 10. The Anderson – Darling test has been used to find out the distribution

best fitting the historical data on daily spot prices returns from 2010 to 2020 of each metal. The

90
Anderson – Darling test has also been applied to obtain the distribution best fitting the historical data

on annual TC benchmark levels from 2004 to 2021.

Cu price simula�on

Au price
simula�on Concentrate price paths
Input Distribu�on calcula�on
data fi�ng (2021 – 2025)
Ag price simula�on

Addi�onal parameters
TC levels
determina�on
simula�on
(RC, Penal�es, …)

Figure 10. Concentrate price paths calculation process. (Source: Own elaboration).

Daily returns for Copper, Gold and Silver, 𝑠𝑠𝑑𝑑𝑘𝑘 , for the N working days in each year from 2021 to

2025, thus 𝑠𝑠𝑑𝑑𝑘𝑘 = �𝑠𝑠1𝑘𝑘 , … , 𝑠𝑠𝑁𝑁𝑘𝑘 � , as well as annual values for copper Treatment Charges benchmark

levels, 𝑇𝑇𝑇𝑇𝑡𝑡 , from 2022 to 2025, have been forecasted by performing 500 independent Monte Carlo

simulations for each metal and for TC using the corresponding best-fitting statistical distribution

shown in Table 2. Historical annual benchmark values from 2010 to 2021 used in Chapter 3 have

been used to estimate expected annual TC benchmark levels. Additionally, forecasts and

distribution fittings have been obtained using Oracle’s Crystal Ball add-in for Microsoft Excel.

Table 30
Statistical Distribution of Historical
daily returns of metal prices. (Source:
Own elaboration).
Distribution
Cu Logistic (0; 0.01)
Au t – Student (0; 0.01)
Ag t – Student (0; 0.01)
TC Weibull (66.34; 3.39)

To find an adequate number of independent simulations to achieve reliable forecasts for both metal

prices daily returns and TC benchmark levels, while also avoiding an unnecessarily high number of

91
simulations, we have pursued the stabilisation of the simulated timeseries’ standard deviation rate

of decrease as the number of simulations is incremented as seen in Figure 11.

Cu Au
0.30% 0.0000% 0.35% 0.0000%

0.25% -0.0020% 0.30% -0.0020%

RATE OF DECREASE OF SIGMA

RATE OF DECREASE OF SIGMA

0.25% -0.0040%
0.20% -0.0040%
0.20% -0.0060%
SIGMA

SIGMA
0.15% -0.0060%
0.15% -0.0080%
0.10% -0.0080%
0.10% -0.0100%
0.05% -0.0100% 0.05% -0.0120%

0.00% -0.0120% 0.00% -0.0140%

25 50 75 100 125 150 175 200 225 250 275 300 325 350 375 400 425 450 475 500 25 50 75 100 125 150 175 200 225 250 275 300 325 350 375 400 425 450 475 500
NUMBER OF SIMULATIONS NUMBER OF SIMULATIONS

Ag
0.80% 0.0000%
0.70%
-0.0050%

RATE OF DECREASE OF SIGMA

0.60%
-0.0100%
0.50%
SIGMA

0.40% -0.0150%
0.30%
-0.0200%
0.20%
-0.0250%
0.10%
0.00% -0.0300%
25 50 75 100 125 150 175 200 225 250 275 300 325 350 375 400 425 450 475 500
NUMBER OF SIMULATIONS

Figure 11. Rate of decrease of sigma for different number of simulations for each metal

price forecasts.

Metals prices each day d, 𝑆𝑆𝑑𝑑𝑘𝑘 , have been projected alternatively to the method presented in Chapter

4, in which a basket of future contracts was the necessary input for the model to deliver a valid

forecast for the metal prices. To deliver a more purposeful approach that stays in line with the main

objective of valuating the activity of the trader, rather than giving extremely precise and reliable

price forecasts, a purely statistical method has been implemented.

This method, which sacrifices some precision in favour of simplicity, sets the closing price for each

metal on December 31, 2020, as the initial reference, 𝑆𝑆0𝑘𝑘 , for the forecasted timeseries of metal

prices. The 500 simulated independent timeseries of daily returns of each metal have been used to

determine the expected accumulated annual returns over the 2021 – 2025 timespan. The expected

92
accumulated annual returns to date, 𝑎𝑎𝑑𝑑𝑘𝑘 , for each metal have been employed to draw individual

expected price paths for copper, gold, and silver.

𝑠𝑠𝑑𝑑𝑘𝑘 = �𝑠𝑠1𝑘𝑘 , … , 𝑠𝑠𝑁𝑁𝑘𝑘 �

𝑎𝑎𝑑𝑑𝑘𝑘 = ��1 + 𝑠𝑠𝑑𝑑𝑘𝑘 �

𝑑𝑑

𝑆𝑆𝑑𝑑𝑘𝑘 = 𝑆𝑆0𝑘𝑘 × �1 + 𝑎𝑎𝑑𝑑𝑘𝑘 �

𝑘𝑘 = 𝐶𝐶𝐶𝐶, 𝐴𝐴𝐴𝐴, 𝐴𝐴𝐴𝐴; 𝑡𝑡 = 1, … ,5

∀ 𝑑𝑑 ≤ 𝑁𝑁 × 𝑡𝑡

Additionally, annual expected benchmark values for copper Treatment Charges (TC), the main

deductions along with Refining Charges (RC) that smelters apply to determine copper concentrates

prices, have only been projected from 2022 to 2025, as the benchmark level for 2021 is already

known by the end of 2020. On the other hand, copper RC levels have not been projected

independently as a nominal 10:1 relation throughout the historical data employed has been

observed, following the explanation in Chapter 3.

Hence, the same relation for copper TC/RC has been used to project RC values based on projections

of TC over the 2022 – 2025 period. On the other hand, the remaining parameters for copper

concentrate pricing, shown in Table 31, are set at constant levels over the full timeframe applying

the same unitary values previously used and calculated for the concentrate’s specific layout.

Table 31
Constant parameters for Concentrate Pricing
(2021 – 2025). (Source: Own elaboration).
Value
Gold RC 10.13 USD/DMT
Silver RC 12.15 USD/DMT
Penalties 12.00 USD/DMT

93
Lastly, individual forecasted price paths for each metal in the concentrate have been merged with

the projections of copper TC/RC benchmark levels as well as with the remaining discounts to

deliver 500 daily price paths for the copper concentrate over the 2021 – 2025 period. Daily price

paths for the concentrate have not been used to model the trader’s financial performance over the

projected period, instead the concentrate’s expected annual average price for each year have been

employed. To obtain the expected annual average price for the concentrate, the daily price paths

have been transformed into annual average price paths. The 500 annual average price paths have

been averaged out to give a single expected annual price for the concentrate. Price paths have been

converted into EUR/DMT, as that is the company reference currency, employing a fixed USD/EUR

rate of 1.10.

5.3.2. Copper concentrate trading companies valuation model

The value of the trader within the established five-year horizon is based upon the approach of

fundamental analysis that matches risk and return (Gajek and Kuciński, 2017; Li and Mohanram,

2019). This can be expressed as the sum of the value of the company’s trading business, 𝑉𝑉0,

developed in accordance with the initial planning, along with the added value that the company may

reach due to variations from this initial business plan, 𝑉𝑉 ′ . Thus, the company’s initial valuation, 𝑉𝑉0,

which is also deemed to be the most likely, is expressed through the discount of Expected Equity

Free Cash Flows (EFCF) along with the business’ Terminal Value, 𝑇𝑇𝑇𝑇𝑛𝑛 , at the end of the timeframe

considered, 𝑛𝑛 = 5.

𝑉𝑉 = 𝑉𝑉0 + 𝑉𝑉 ′
𝑛𝑛
𝐸𝐸𝐹𝐹𝐹𝐹𝐹𝐹𝑡𝑡 𝑇𝑇𝑇𝑇𝑛𝑛
𝑉𝑉0 = � + 𝑛𝑛 = 5
(1 + 𝐾𝐾) 𝑡𝑡 (1 + 𝐾𝐾)𝑛𝑛
𝑡𝑡=1

On the other hand, the potential added value alternatives, 𝑉𝑉 ′ , are analysed through Real Options,

expressed as the sum of the expected company value if each option is exercised multiplied by its

respective exercise probability as schematised in Figure 12. 𝑢𝑢1 𝑣𝑣 and 𝑢𝑢2 𝑣𝑣 are the company’s values

94
if either the first grow option or the second is exercised, whereas the 𝑑𝑑1 𝑣𝑣 is the company’s value

should the abandon option be exercised.

Growth – Step 1

p1
u1v
Grow

Growth – Step 2

p
u2v
p2

V’

1-p

d1v

Abandon

Figure 12. Real Options valuation scheme. (Source: Own elaboration).

𝑉𝑉 ′ = 𝑝𝑝[𝑝𝑝1 𝑢𝑢1 𝑣𝑣 + 𝑝𝑝2 𝑢𝑢2 𝑣𝑣] + (1 − 𝑝𝑝) [𝑑𝑑1 𝑣𝑣]

5.3.3. Equity free cash flows

Equity free cash flows (Damodaradan, 2006; Yaari et al., 2016; Aharon et al., 2019; Smith and

Pennathur, 2019) in each projected period, 𝐸𝐸𝐹𝐹𝐹𝐹𝐹𝐹𝑡𝑡 , are obtained through the application of the

Market Strategy, the Financial Strategy, the Capital Strategy, the Financial Policy and the Working

Capital Strategy.

𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑡𝑡 = 𝑆𝑆𝑆𝑆𝑡𝑡 + 𝑑𝑑𝑡𝑡 − 𝐼𝐼𝑡𝑡 + 𝐹𝐹𝑡𝑡

Where both the self-financing at the end of each period, 𝑆𝑆𝑆𝑆𝑡𝑡 , and the dividends perceived by

shareholders during that period, 𝑑𝑑𝑡𝑡 , are fixed by the Financial Strategy, in which the distribution of

95
returns or Earnings Before Interest Taxes Depreciation and Amortisation (EBITDA) is materialised.

Seeking the preserving of the trader’s financial capabilities, as well as minimising the necessity to

resort to external capital, the Financial Strategy has set dividends payments to zero, 𝑑𝑑𝑡𝑡 = 0,

throughout the company’s entire lifespan. On the other hand, the trader’s self-financing at each

period t, 𝑆𝑆𝑆𝑆𝑡𝑡 , has been calculated as the generated EBITDA, minus the financial expenses, 𝑓𝑓𝑡𝑡 , and

the amount of taxes paid by the company, 𝑇𝑇𝑇𝑇𝑡𝑡 .

𝑆𝑆𝑆𝑆𝑡𝑡 = 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑡𝑡 − 𝑓𝑓𝑡𝑡 − 𝑇𝑇𝑇𝑇𝑡𝑡

On the other hand, EBITDA (Rozenbaum, 2019), is obtained through the Market Strategy. This

sets up the trader’s operative over the projected timeframe 2021 – 2025 by projecting the number

of standard containers the trader is expected to sell in each period, expressed in Twenty-foot

Equivalent Units (TEU), as well as the annual increase rate of sales of containers. The sold mass of

containers in each period, expressed in Dry Metric Tons (DMT), 𝑉𝑉𝑡𝑡 , is a function of the nominal

loading capacity of each container, 27.6 MT/TEU, as well as the humidity of the concentrate.

𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑡𝑡 = (𝑉𝑉𝑡𝑡 × 𝐶𝐶𝑡𝑡 ) − 𝐶𝐶𝐶𝐶𝑡𝑡 − 𝐶𝐶𝐶𝐶𝑡𝑡 − 𝑂𝑂𝑂𝑂𝑡𝑡

The EBITDA in each period is hence calculated as the net value of sales, this being the mass of dry

tons of copper concentrates sold multiplied by their forecasted price in €/DMT, 𝑉𝑉𝑡𝑡 × 𝐶𝐶𝑡𝑡 , minus the

costs of acquiring the concentrates, 𝐶𝐶𝐶𝐶𝑡𝑡 , the costs of employees, 𝐶𝐶𝐶𝐶𝑡𝑡 , and the operative expenses,

𝑂𝑂𝑂𝑂𝑡𝑡 . These costs and expenses that the trader faces in its business operation have been set in

accordance with the observed data from the trading company taken as a reference, the cost of the

concentrates being indirectly determined by fixing a gross margin over the sales value.

Next, by the application of the Capital Strategy, particularly through the Investment Policy, the

company’s investment needs, or Capital Expenditures (CAPEX), for each period t have been

outlined, these being defined as the increases of the non-current assets, ∆𝑁𝑁𝑁𝑁𝑁𝑁, minus their

decrements, ∇NCA𝑡𝑡 , plus the change in the net working capital, ∆∇𝑁𝑁𝑁𝑁𝑁𝑁𝑡𝑡 .

96
 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 = ∆𝑁𝑁𝑁𝑁𝑁𝑁𝑡𝑡 − ∇NCA𝑡𝑡 ± ∆∇𝑁𝑁𝑁𝑁𝑁𝑁𝑡𝑡

Additionally, as part of the Investment Policy the Working Capital Strategy has been designed. This,

given the nature of physical trading, and in a simplified form, establishes that payments to suppliers

are executed only once the copper concentrates purchased are served at the trader’s designated port

for loading. On the other hand, the collection of payments from customers is estimated to be made

once the cargo has been delivered at the destination port. The trader’s business model has been

standardised, restricting its activity to fetch concentrates from different origins in South America

and delivering to smelters in China.

Finally, through the Financial Policy, the indebting capacity of the company has been projected, this

being described as the increases of the non-current liabilities, ∆𝑁𝑁𝑁𝑁𝑁𝑁𝑡𝑡 , minus their decrements,

∇𝑁𝑁𝑁𝑁𝑁𝑁𝑡𝑡 , plus capital increases, 𝐶𝐶𝐶𝐶𝑡𝑡 .

𝐹𝐹𝑡𝑡 = ∆𝑁𝑁𝑁𝑁𝑁𝑁𝑡𝑡 − ∇𝑁𝑁𝑁𝑁𝑁𝑁𝑡𝑡 + 𝐶𝐶𝐶𝐶𝑡𝑡

5.3.4. Terminal value

The Terminal Value depicts the expected income the trader should obtain beyond the year set as a

limit for the planification horizon, year n, this being 2025. It can be established assuming as a

constant the equity free cash flow from that moment on, 𝐸𝐸𝐹𝐹𝐹𝐹𝐹𝐹𝑛𝑛 , discounted at a factor K.

𝐸𝐸𝐹𝐹𝐹𝐹𝐹𝐹𝑛𝑛
𝑇𝑇𝑇𝑇𝑛𝑛 =
𝐾𝐾

5.3.5. Discount factor

The discount factor for future cash flows, K, has been determined employing the formulation in the

Capital Asset Pricing Model, based on Markowitz (1952) and Markowitz (1959), and later

developed in subsequent works by Sharpe (1964) and Makwasha et al. (2019). This is a function of

the perceived risk-free rate, 𝑅𝑅𝑓𝑓 , plus the market risk premium, denoted as the difference between

97
the demanded return by investors, 𝑅𝑅𝑚𝑚 , with the risk-free rate, multiplied by a factor that correlates

the risk of the company with the market, 𝛽𝛽.

𝐾𝐾 = 𝑅𝑅𝑓𝑓 + �𝑅𝑅𝑚𝑚 − 𝑅𝑅𝑓𝑓 � × 𝛽𝛽

The 𝛽𝛽 employed has been determined qualitatively, given the inconvenience of calculated betas for

non-publicly traded companies, which cannot be compared to one that is and concerning which

there are enough historical data. For the specific case of the copper concentrates trader, there is no

availability of data from a similar company that could be employed for the calculation of a generic

beta. It is also unfeasible to fall back on data from large publicly traded corporations as these have

a vertically integrated structure, the extractive, processing, trading, and supplying steps of the

commodities supply chain being developed altogether.

Consequently, the MASCOFLAPEC method for qualitative betas described in Fernandez and

Carabias (2007) has been used, adjusting the weighing to the ones shown in Table 32. The answers

to two separate surveys undertook independently by executive staff from the reference trader, where

each question could be answered grading from 1 to 5 according to the interviewed person’s

perceived level of risk from low to remarkably high, have been averaged out. The overall result of

each questionnaire has also been normalised by employing a factor of 0.5 in both surveys, thus

ensuring that the values of the qualitative betas fall within a reasonable range from 0.5 to 2.5.

98
 Table 32
MASCOFLAPEC weighing. (Source: Own elaboration).
Weight
M Management 10.0%
A Assets; Business; Industry/Product 25.0%
S Strategy 3.0%
C Country risk 15.0%
O Operative leverage 10.0%
F Financial leverage 15.0%
L Liquidity of investment 5.0%
A Access to sources of funds 5.0%
P Partners 2.0%
E Exposure to other risks 5.0%
C Cash flow stability 5.0%

5.3.5. Real options

The different long-run managerial alternatives have been considered as the business faces

uncertainties that might either play in its favour or against it. Should market conditions become

favourable enough, the trader could opt for expanding its supply capacity at differentiated steps

whether conditions are more or less advantageous. On the other hand, in the event of a critical

market downturn the trader would choose to terminate its operations rather than continuing with the

base case. The expected value of the trading company as a function of the possible modifications of

the original business plan has been assessed using the Real Options method (Savolainen, 2016;

Aminrostamkolaee et al., 2017; Berk and Podhraski, 2018; Yeh and Lien, 2020).

To value the different alternatives to be considered, mean annual concentrate prices at each year

between 2021 – 2025 have been calculated based on previous simulations. As shown in Figure 13,

each of the 500 simulated timeseries of daily concentrate prices have been averaged annually so

each of the 1310 components is funnelled to just 4 elements. Mean annual prices are hence taken to

populate a discrete distribution that allows for real option modelling under the concentrate price

uncertainty.

99
 Input
data
(see. Figure 1)

Mean annual
prices adapter
(2021 – 2025)

Discrete distribu�on Mean

Company
adjusted to available
expected value
mean concentrate prices
annual
distribu�on prices
(2021 – 2025)

Figure 13. Global average concentrate prices and Company Expected Value calculation

workflow. (Source: Own elaboration).

To assess the different scenarios the trader’s management would face throughout the company’s

lifetime, three different kinds of real options have been considered: Abandonment, Continuity and

Growth. The abandonment option has been first considered to model the company’s terminal value

if adverse market conditions occur. The option is set to be exercised if the company’s expected

value falls below the opportunity cost, which has been assumed to be the company’s value estimated

through the DCF method previously developed.

On the other hand, a growth option with two differentiated incremental steps has been modelled in

which supply capacity would be first increased by 25 percentage points, or ultimately by 50

percentage points, rounded to give complete containers. These increases in the trader’s supply

capacity are meant to cope with sudden increases in copper concentrate prices. The increases in

prices are considered as demand-driven, hence underpinning the expansion of the trader’s supplying

capacity as the whole concentrates’ availability is assumed to be sold.

100
Both the two increase steps have been chosen over narrower alternatives, where smaller increments

were considered and more steps were possible, since the differences in the company value become

meaningful as steps are wider. In addition, provided that the defined growth options are exercised

according to copper concentrate prices, smaller increments pose lower exercise thresholds. These

would usually carry an inappropriately high exercise probability linked to the volatility of copper

concentrates prices.

As any expansion in the trader’s supply capacity is assumed to get immediately sold out, the trader

would therefore to a certain extent be willing to pay a higher price for the concentrates acquired

from producers, whilst it is also keen on marketing them to its customer at a less optimal price

aiming to maximise sales. Hence, for each supply capacity increasement step, a selling penalty is

borne by the trader, whilst the gross margin on the concentrates’ acquisition costs also shrinks,

indicating the traders wish to purchase the required concentrate tonnage. The two increase steps for

the trader’s supply capacity are displayed in Table 33 as a percentage of the capacity fixed at the

baseline scenario, where the corresponding selling price penalisation and gross margin decrease

amount is also shown.

Table 33
Trader’s supply capacity increasement steps (Selling price penalty/Gross margin
decrease). (Source: Own elaboration).
25% 50%
Selling price penalty -250 USD/DMT
Step 1
Gross margin decrease -125 bp
Selling price penalty -500 USD/DMT
Step 2
Gross margin decrease -250 bp

The exercise thresholds for the abandonment option have been determined by running 2000

independent Monte Carlo simulations for the price of the concentrate for each year, while the

remaining prices for the following and previous years have been left at the global annual average

prices. The minimum price levels are marked by the discount factor employed. To achieve price

101
thresholds, the company valuation at each price level when the Monte Carlo simulation is carried

out is compared to the company baseline valuation defined by the DCF method.

To determine the exercise threshold for each step of the growth option, 280 trials have been run with

Crystal Ball’s Opt Quest at incremental steps of 10 USD/DMT each for the threshold value,

performing 2000 Monte Carlo simulations per trial. The minimisation of the mean of the company’s

expected value using the discounted EFCF method is sought. This is restricted to being higher than

the base case set by the DCF valuation. In addition, each step’s threshold and expected company

mean value has been limited down to the levels reached in the previous steps, while the first step’s

threshold has been limited to the 2025 global annual average price.

Finally, the exercise probabilities, along with the company’s expected value in either case have been

estimated carrying out 2000 Monte Carlo simulations. The inverse cumulative distribution

functions for each year between 2021 – 2025 for copper concentrate prices, as shown in Figure 14,

have been employed to model Real Options. Each year’s discrete distribution of concentrate prices

has been taken directly from the results of the preceding simulations, thus disposing of 500 possible

concentrate prices for each year within the valuation horizon.

102
 Figure 14. ICDF for annual Copper Concentrates prices used for ROV. (Source: Own

elaboration).

5.4. Results and Discussion

5.4.1. Concentrate price forecasts

Five sample paths for copper concentrate price, determined as explained in subchapter 5.3.1., are

shown in Figure 15 out of the 500 paths simulated. Respectively, the annual average prices for each

of the five sample paths are shown in Figure 16. Additionally, each year’s expected annual average

price for each year used to perform the trader’s valuation through the DCF method are shown in

Figure 17, being the average of the 500 annual average prices. The average daily price path is also

shown in Figure 17.

103
 $5,000.00
$4,500.00
Concentrate price (USD/DMT)
$4,000.00
$3,500.00
$3,000.00
$2,500.00
$2,000.00
$1,500.00
$1,000.00
$500.00
$-
2021 2022 2023 2024 2025
Year

Figure 15. Five sample price paths for copper concentrate 2021 – 2025. (Source: Own

elaboration).

$4,500.00
Annual average concentrate price

$4,000.00

$3,500.00
(USD/DMT)

$3,000.00

$2,500.00

$2,000.00

$1,500.00

$1,000.00

$500.00

$-
2021 2022 2023 2024 2025 2026
Year

Figure 16. Annual average price paths for copper concentrates 2021 – 2025. (Source:

Own elaboration).

104
 $3,200.00

Concentrate price (USD/DMT) $3,100.00

$3,000.00

$2,900.00

$2,800.00

$2,700.00

$2,600.00

$2,500.00
2021 2022 2023 2024 2025
Year

Figure 17. Average price path for 500 forecasts and expected annual average price for

copper concentrate 2021 – 2025. (Source: Own elaboration).

The expected annual average prices for the copper concentrate used to model the trader’s activity

are also shown in Table 34.

Table 34
Expected Annual Average Price for Copper Concentrate 2021 – 2025. (Source:
Own elaboration).
2021 2022 2023 2024 2025
Concentrate Price
2,732.89 2,814.79 2,899.35 2,965.68 3,061.59
(USD/DMT)
Concentrate Price
2,484.45 2,558.90 2,635.77 2,696.07 2,783.27
(EUR/DMT)

5.4.2. Equity free cash flows

The trader’s expected Equity Free Cash Flows (EFCF) have been estimated following a self-

financing approach, where dividends have not been paid out to shareholders, leaving any possible

external source for financing the company’s activities untapped. The initial levels of self-financing

achieved, shown in Table 7 alongside the forecasted levels for Equity Free Cash Flows, may only

be feasible thanks to the relational capital that the project’s promotors count on. This is provided

from an initial and stable market share, knowledge of the commodity processes, as well as of their

prices and trading mechanics.

105
 Table 35
Expected Annual Average Price for Copper Concentrate 2021 – 2025. (Source:
Own elaboration).
2021 2022 2023 2024 2025
Self-financing 125,619.27 85,091.33 116,812.90 159,045.64 215,089.70
Dividends - - - - -
CAPEX 118,176.64 71,679.89 59,914.91 76,030.16 99,049.76
Financing - - - - -
EFCF 7,442.63 13,411.44 56,897.99 83,015.47 116,039.94

Expected EFCF levels achieved by the trader each year between 2021 to 2025 are subject to the

annual forecasted price for the concentrate, along with the different strategies and policies

preliminarily set by the company. The projected tonnage sold for each year has been assumed to

increase at a constant level of around 25%, rounded to full containers. The value of sales for each

year, on the other hand, varies at a different pace as it also depends on the forecasted fluctuation of

concentrate prices. The price of concentrate has been set per dry metric ton of concentrate (DMT);

that is, excluding the humidity. Nonetheless, humidity is to be accounted for as transportation costs

are measured on the global tonnage carried, expressed in metric tons (MT).

The trader’s anticipated EFCF levels are expected to increase on a yearly basis, although the rate of

increase presents some deceleration beyond the year 2022 as this shows the effect of the trader’s

normalising its creation of stockpiles of concentrates in comparison with the first two years of

operation. This is also the primary factor driving EBITDA up in 2021, as net concentrates stockpiles

variations in the year levels up to 116,746.69 €, with EBITDA being 166,742.36 €. On the other

hand, the net stockpiles variations in 2022 are less than a third of the increase shown in 2021, whilst

EBITDA, excluding stockpile variations, increased by 58.03%, dropping 32.41% if concentrate

stockpile variations are added in.

Additionally, EBITDA, and thus EFCF levels, are driven by the steady increase in sales during the

whole timespan considered. On the other hand, the cost of the concentrates sold has been projected

106
as a fraction of their sale price, as traders usually apply higher discounts to miners to determine the

price paid for the concentrate that they acquire. The gross margin between the acquisition price and

the sale price for the given concentrate has been observed to be at around 8.75%2. In addition, the

efficiency and scalability of commodity traders help them to optimise structural costs, thus

Personnel Costs have been assumed to remain constant at observed levels despite the projected

increase in the trader’s sales volume. This is because the additional workload caused by the increase

in sales is assumed to remain sufficiently manageable by the trader’s current staff.

Also, Operative Expenses increase at a constant pace of 1.35% annually. Hence, higher business

revenues with marginal increments of main expenses, and maybe slight fluctuations in margins, are

to be expected throughout the trader’s lifetime. This is clearly shown in Figure 18 and Figure 19,

where the net increases (decreases) of revenues, EBITDA and self-financing are shown along with

their relative yearly variations.

€1,000,000.00

€800,000.00
Relativa variations (EUR)

€600,000.00

€400,000.00

€200,000.00

€-
2022 2023 2024 2025

€(200,000.00)
Year

Revenues Variation EBITDA Variation Self-financing Variation

Figure 18. Revenue, EBITDA, Self-financing variations. (Source: Own elaboration).

2. As observed from non-publicly available and confidential economical and financial data employed as reference.

107
 50.00%

40.00%

30.00%
Relative variation (%)

20.00%

10.00%

0.00%
2022 2023 2024 2025
-10.00%

-20.00%

-30.00%

-40.00%
Year

Revenues Relative Variation EBITDA Relative Variation

Self-financing Relative Variation

Figure 19. Revenue, EBITDA, Self-financing relative variations. (Source: Own

elaboration).

As can clearly be seen, revenues experience a somewhat steadily increase while EBITDA and self-

financing behave likewise, reaching a certain stabilisation of their respective rates of increase around

2024 even though operative expenses, excluding the costs of the concentrates sold, maintain their

previously described control rise. Further detail on the trader’s EBITDA construction as well as its

planned Market Strategy may be found in Table 36 and 37.

Financial expenses have also been assumed to remain constant throughout the entire projected

timespan, as most financial resources needed by the company are self-supplied. Just a small short-

term recurring loan is available, with a maximum available amount of 40,000.00 EUR, of which

10,000.00 EUR are on average disposed of yearly. This loan has a fixed interest rate of 4.50% with

an annual fee of 2.00%, being fully reimbursed at the end of every period. The trader lacks any other

outstanding debt, either short- or long-term.

108
Table 36
Market Strategy 2021 – 2025 (EUR, unless otherwise stated).
2021 2022 2023 2024 2025
Sales (TEU) 23 29 36 45 56
Sales (DMT) 584.02 736.37 914.11 1142.64 1421.95
Operating Revenue 1,450,955.90 1,884,292.08 2,409,391.48 3,080,640.54 3,957,667.29
Cost of Concentrate 1,334,212.32 1,732,682.37 2,215,532.39 2,832,772.91 3,639,234.29
Cost of Employees 47,160.00 47,160.00 47,160.00 47,160.00 47,160.00
Other Operative Expenses 19,587.90 25,437.94 32,526.78 41,588.65 53,428.51

Table 37
Expected Income Statements 2021 – 2025 (EUR).
2021 2022 2023 2024 2025
Revenues 1,450,955.90 1,884,292.08 2,409,391.48 3,080,640.54 3,957,667.29
Cost of Concentrates Sold -1,334,212.32 -1,732,682.37 -2,215,532.39 -2,832,772.91 -3,639,234.29
Cost of Employees -47,160.00 -47,160.00 -47,160.00 -47,160.00 -47,160.00
Other Operative Expenses -19,587.90 -25,437.94 -32,526.78 -41,588.65 -53,428.51
Stockpile Variations 116,746.69 33,693.34 40,828.24 52,191.86 68,191.77
EBITDA 166,742.36 112,705.10 155,000.54 211,310.85 286,036.26
Amortisations -6,000.00 -6,000.00 -6,000.00 -6,000.00 -6,000.00
Provisions - - - - -
Financial Expenses -1,250.00 -1,250.00 -1,250.00 -1,250.00 -1,250.00
EBT 159,492.36 105,455.10 147,750.54 204,060.85 278,786.26
Taxes -39,873.09 -26,363.78 -36,937.63 -51,015.21 -69,696.57
Net Result 119,619.27 79,091.33 110,812.90 153,045.64 209,089.70

109
On the other hand, projected annual CAPEX values have been calculated as the variations of the

asset’s book value, along with the net yearly requirements for working capital. Here current assets

and liabilities are accounted for, including accounts payable, stockpile variations and tax payments.

The trader’s free cash flows are greatly decreased by heightened levels of CAPEX, mainly inflated

due to working capital necessities. Unfortunately, this is a major disadvantage of almost any

physical commodity trading operation, as payments to commodities suppliers, miners in the case of

copper concentrates, need to be made earlier, whereas collecting accounts pending for the sale of

the same concentrates to customers is deferred more or less time depending on the conditions agreed

for shipping the commodities, usually Cost Freight (CFR) or Cost Insurance Freight (CIF) at the

destination port. Thus, prior to collecting payments from customers, the concentrates must be

shipped to the customers, usually at large distances, and several tests and proceedings need to be

conducted. To reflect this effect, we have assumed the average accounts payable days to be 5, while

accounts receivable days are 30 and days inventory outstanding are 25.

5.4.3. Discounted Cash Flows Valuation

Projected free cash flows have been discounted at a rate of 10.69% yearly. The employed discount

factor, K, has been calculated using the parameters shown in Table 38, the risk-free rate being the

10-year yield of US treasuries, while the demanded return by investors is the mean daily annual

returns of the MSCI World Index from 2010 to 2020. Also, the value of 𝛽𝛽, which has been

determined following the MASCOFLAPEC method previously explained under the chapter of

Discount Factor, has been found to be 1.470.

Table 38
Values for the calculation of the Discount Factor, K. (Source:
Own elaboration).
Reference Value
𝑅𝑅𝑓𝑓 US T-BILL 10Y 0.50%
𝑅𝑅𝑚𝑚 MSCI World Index 7.43%
𝛽𝛽 MASCOFLAPEC 1.470

110
Additionally, the trader’s expected cumulative discounted EFCF throughout the whole timespan

considered from 2021 to 2025 amounts to 184,776.60 €, according to Table 39, this being the most

likely valuation horizon. Also, the EFCF obtained by the trader in the last year of operations, 2025,

is projected to be 116,039.90 €, which gives a TV for the company of 1,803,983.94 €. Equity free

cash flows are estimated to remain constant from the year 2025 on, as further assumptions beyond

this point on concentrates prices, market conditions or the company operation itself are hardly

quantifiable.

Table 39
Discounted Equity Free Cash Flows 2021 – 2025 (EUR). (Source: Own elaboration).
2021 2022 2023 2024 2025
EFCF 7,442.63 13,411.44 56,897.99 83,015.47 116,039.94
Discount Factor 1.1069 1.2252 1.3561 1.5010 1.6614
Discounted EFCF 6,724.02 10,946.65 41,957.14 55,305.84 69,842.93
Most likely
184,776.60
Valuation Horizon
Terminal Value 1,803,983.94
Company Value 1,988,760.54

Finally, the overall Company Value, calculated as the sum of the TV and the discounted cumulative

EFCF obtained from 2021 to 2025, or the most likely cumulative value, is 1,988,760.54 €. This, if

compared to the TV of the trader’s equity shown on the expected Balance Sheets for the year 2025,

811,658.83 €, is 145% higher, giving a positive outlook for the trader’s business and therefore for

its growth perspectives.

5.4.4. Real options valuation

The optimised exercise concentrates price thresholds for the growth option’s capacity increase steps

can be seen in Table 40, while each year’s optimum exercise price thresholds for the abandonment

option are shown in Table 41. Also, as no scale-down or reversal option has been considered, the

increases in the trader’s supplying capacity or the closure of the company are thus made irreversibly

once a certain price threshold has been exceeded. Additionally, so long as the market price for the

111
concentrate stays within the limits marked by the abandonment option price thresholds and the price

threshold of the first step of the growth option, the trader’s operative will remain unaltered.

Table 40
Concentrate exercise price for growth option increase steps
(USD/DMT). (Source: Own elaboration).
Growth 25% Growth 50%
Exercise Price 3,810 5,780

Table 41
Concentrate exercise price for abandonment option in each year (USD/DMT).
(Source: Own elaboration).
2021 2022 2023 2024 2025
Exercise Price 2,732 2,811 2,895 2,965 3,057

5.4.5. Abandonment option

The trader would consider halting its activity in the event of the price of copper concentrates

dropping below the threshold levels for each year. In this case, the sales of the company in the

remaining time until the end of 2025 after the triggering event occurred are expected to be zero.

Also, the abandonment option has been supposed to be exercised at the end of the corresponding

year, thus both the exercise year’s incomes and costs are fully realised. Nonetheless, subsequent

annual cash flows from the exercising option are assumed to be zero, as no remaining stock still

possessed by the company would be sold and neither would other assets be liquidated afterwards.

Table 42
Discounted Equity Free Cash Flows 2021 – 2025 (EUR). (Source: Own elaboration).
2021 2022 2023 2024 2025
Exercise Probability 52.60% 53.95% 57.00% 56.05% 59.55%
Company Mean
Expected Value 6,724.02 17,670.68 59,627.82 114,933.67 184,776.60
(EUR)
Option Value (EUR) 3,536.83 9,533.33 33,987.86 64,420.32 110,034.47

In Table 42 the exercise probabilities of the abandonment option in each year are shown along with

the mean expected value of the company and the option value itself if the option were exercised at

112
the end of that year. The company mean expected value when the option is to be exercised is

calculated as the sum of the discounted equity free cash flows up to that moment, excluding the

business’ terminal value, as it will not resume operations later. Thus, the intrinsic option value can

be calculated as the sum of the realised option value at each execution opportunity, hence being

221,512.81 €.

5.4.6. Growth Option

On the contrary, should the price of the concentrate exceed the exercise threshold levels, the trader

would automatically expand its capacity in accordance, albeit assuming certain rises in the price it

pays to acquire the concentrate from suppliers. The option exercise probability is 54.55%, existing

44.10% of just expanding capacity by 25% as the second threshold would not be surpassed.

Nevertheless, there is a 10.45% probability of the trader increasing its capacity by 50%, as the price

of concentrate would soar pass the mark of 5,780 USD/DMT at some point between 2021 and

2025.

Table 43
Growth Option Probabilities and Value. (Source: Own elaboration).
Growth 25% Growth 50%
Exercise Probability 44.10% 10.45%
Mean Expected Value (EUR) 1,800,234.94 530,913.92
Option Value (EUR) 793,903.61 55,480.50

In Table 43 the results achieved for the two steps of the growth option can be seen. An expectancy

of 793,903.61 € is estimated if the first step of the growth option is triggered, while an additional

expectancy of 55,480.50 € would be added if the second step is too. Consequently, the total value

of the growth option is assumed to be 849,384.11 €.

The time effect has not been taken into consideration for the growth option to make the valuation

more straightforward. However, it is worth mentioning that the results present a quite meaningful

113
level of dispersion certainly due to this factor, as it would not have the same impact on the company

value to exercise the option in an earlier stage than later. This is because the trader would enjoy a

higher selling tonnage and, presumably, higher incomes. But it is also true that the contrary might

well happen, mainly when the triggering event takes place early in the company’s life. As

concentrate prices have much more time to freely fluctuate and hence fall back below the growth

threshold, the business model that is developed when the growth step is taken, which contemplates

the worsening of the buying and selling prices of the copper concentrate, can easily become

inefficient, as it is notably more price sensitive.

5.4.7. Company Value

The company value resides not only in the management’s projections which allow setting a baseline

valuation scenario, but also in the potential behind the developing opportunities the company could

come across along the way. These opportunities need to be assessed to attain a fair valuation of the

company, hence accounting for the initial plan as well as for the evolution potential. As for the

copper concentrate trading company, the DCF method brings forward the most likely value of the

company, this being kept within the original volume of tons sold per year and its forecasted average

annual price from 2021 to 2025. Moreover, the potential within the business model is portrayed by

the Real Options Valuation method. This allows considering price uncertainty, which is indeed the

main source of risk and of returns the trader is exposed to.

114
 P = 10.45%
50% 2,044,241.04 €

25% 2,782,664.15 €
P = 44.10%

GROWTH OPTION

DCF 1,988,760.54 € BASELINE

ABANDON OPTION

P = 52.60%
2021 1,992,297.37 €

P = 53.95%
2022 1,998,293.87 €

P = 57.00%
2023 2,022,748.40 €

P = 56.05%
2024 2,053,180.86 €

P = 59.55%
2025 2,098,795.01 €

Figure 20. Real Options valuation alternative for Copper Concentrate Trader. (Source:

Own elaboration).

As shown in Figure 20 the company would develop its business model reaching a total value

slightly under 2.0 M€ if everything goes as initially planned. This implies that concentrate prices,

and therefore metal prices and annual benchmark TC/RC levels, stay within the range delimited by

the exercise thresholds of both the growth and the abandonment options. Also, so long as reference

prices remain within limits, the minimum demanded return to the trader’s business would be met,

making it undesirable for investors to move away to look for higher-yielding investment

opportunities elsewhere. On the other hand, if the opportunity cost is reached because the company

fails to deliver the returns required, the abandonment option shall be executed by investors, who

would then seek for investments that yielded at least what concentrate trading was meant to. In such

a case, as time passes, investors would have accumulated more cash flows from their investment in

the trader, thus increasing the value of the abandonment option over time.

115
When it comes to capacity growth, the company valuation is increased as the growth option is

exercised in either of its steps. However, the potential improvement for the company valuation if

the trader’s supply capacity is increased by 50% instead of 25% is not as much for the first step as

for the second. This, nonetheless, can still be interpreted as a strategical decision by the company’s

management that, while optimising the moment to take the increase decision, perceives that

expanding the supply capacity would eventually lead to a higher market share which, in the end,

could be traduced as a competitive advantage as well as higher survival odds for the company in

the long run.

116
 Chapter 6: Conclusions

Throughout this research we have developed a copper concentrates benchmark price model that

provides market participants involved in copper concentrates production, processing or trading of a

neutral, unbiased price reference. The methodology proposed resorts to copper, gold and silver spot

and futures data, as well as well-known forecasting techniques, to circumvent the problem of

lacking a market-wide reference for copper concentrates.

Despite not having access to most confidential contract terms of copper concentrate supply

agreements, the findings of this research suggest that some of the developments may well be of

application by market participants. For instance, the ability to forecast with a sufficient degree of

accuracy the TC/RC benchmark levels can have a great financial importance for either copper

miners and smelters, as both can manage to set in advance more advantageous TC/RC for their

contracts if considering the information delivered by the model.

Furthermore, the practical usability and potential advantages of using the benchmark price model

for copper concentrates by a copper concentrates trader has also been analysed using Real Options,

which additionally shows how having good forecasts of copper concentrates prices can substantially

improve the trader’s financial performance, while also allows to take to action when prices are

expected to fall.

6.1. Implications of developing forecasts for Copper TC/RC

Copper TC/RC are a keystone for pricing copper concentrates which are the actual feedstock for

copper smelters. The potential evolution of TC/RC is a question of both economic and technical

significance for miners, as their value decreases the potential final selling price of concentrates.

117
Additionally, copper miners' revenues are more narrowly related to the market price of copper, as

well as to other technical factors such as ore dilution or the grade of the concentrates produced.

Smelters, on the contrary, are hugely affected by the discount which they succeed in getting when

purchasing the concentrates, since that makes up the largest part of their gross revenue, besides other

secondary sources. In addition, eventual differences between TC/RC may give commodity traders

ludicrous arbitrage opportunities. Also, differences between short and long-term TC/RC

agreements offer arbitrage opportunities for traders, hence comprising a part of their revenue in the

copper concentrate trading business, as well copper price fluctuations and the capacity to make

economically optimum copper blends.

As far as we are aware, no rigorous research has been carried out on the behaviour of these

discounts. Based on historical data on TC/RC agreed upon in the LME Copper Week from 2004 to

2017, three potentially suitable forecasting models for TC/RC annual benchmark values have been

compared through four measures of forecasting accuracy at different horizons. The models chosen

in this research were chosen, firstly, due to their broad implementation and proven capacity in

commodity prices forecasting that they all share and, secondly, because of the core differences in

terms of price behaviour with each other.

Our research contributes by delivering a formal tool for smelters or miners to make accurate

forecasts of TC/RC benchmark levels. The level of errors attained indicates the LES model may be

a valid model to forecast these crucial discounts for the copper market. In addition, our work further

contributes to helping market participants to project the price of concentrates with an acceptable

degree of uncertainty, as now they may include a fundamental element for their estimation. This

would enable them to optimise the way they produce or process these copper concentrates. Also,

118
the precise knowledge of these discounts’ expected behaviour contributes to let miners, traders and

smelters alike take the maximum advantage from the copper concentrate trading agreements that

they are part of. As an additional contribution, this work may well be applied to gold or silver RC,

which are relevant deduction concentrates when these have a significant amount of gold or silver.

6.2. Accuracy of TC/RC benchmark levels forecasts

Focusing on the MAPE values achieved, those obtained by the LES model when TC and RC are

treated independently have been significantly lower than for the rest of the models. Indeed, one-

year-ahead MAPE measures for TC values for the GBM model (20%) almost triple those of the

OUP model (7.66%), in contrast with the significantly lower values from the LES model

(0.0046%). This gap tends to be narrowed when TC/RC values are forecasted at longer horizons,

when most measures of error become more even. The GBM and OUP models have proven to

deliver better accuracy performance when the TC/RC values are projected monthly and then

averaged to obtain annual benchmark forecasts. Even so, the LES model remains the most accurate

of all with MAPE values of 10% at two-year-ahead forecasts, with 18% and 31% for TC for OUP

and GBM, respectively.

Despite TC and RC being two independent discounts applied to copper concentrates, they are both

set jointly with an often 10:1 relation as our data reveals. This relation also transcends to simulation

results and error measures, hence showing a negligible discrepancy between the independent

forecasting of TC and RC, or the joint forecasting of both values, keeping the 10:1 relation. This is,

for instance, the case of the five-year-ahead OUP MAPE values (0.2715/0.2719) which were

obtained without observing the 10:1 relation in the data. A similar level of discrepancy was obtained

at any horizon with any model, which indicates that both values could be forecasted with the same

accuracy using the selected model with any of them and then applying the 10:1 relation.

119
Our findings thus suggest that both at short and long-term horizons TC/RC annual benchmark levels

tend to exhibit a pattern which is best fit by an LES model. This indicates that these critical discounts

for the copper trading business do maintain a certain dependency on past values. This would also

suggest the existence of cyclical patterns in copper TC/RC, possibly driven by many of the same

market factors that move the price of copper.

6.3. The importance of having a benchmark on copper concentrates prices

We have developed a copper concentrates benchmark price model which aims to address the

absence of market-wide accepted references that can orientate market participants to set a fair price

for the copper concentrates they trade with. The relevance of a reliable reference to set the price of

copper concentrates is key as its trade involve most copper mines, smelters and commodity traders

worldwide, which are currently forced to employed indirect references and incomplete information

to find out the just price of these.

The methodology followed helps to deliver estimations on copper concentrates pricing, enabling

comparing physical trades agreements to a comprehensible benchmark based on both official,

publicly available metal spot and future prices and main deductions forecasts. The model has been

formulated in a way that makes it suitable for pricing any concentrate available in the market as

there is not a typical concentrate layout, though we have standardized our simulations by carrying

them out with five sample concentrates, being all representative of the complete spectrum of layouts

from a statistical standpoint.

In addition, our model offers a deep insight into the different components of the price of concentrates

by giving further understanding of the future role that each of the elements involved in pricing the

120
concentrates will have as the market conditions evolve. Thus, this tool brings new possibilities for

optimizing mining production, using the expected value of the concentrate and the expected role of

penalties to target the ores that would maximize profits according to the expectations provided by

the model. Also, as the model provides greater foresight on the behavior of penalties regarding the

final price of concentrates, blending activities, mainly carried out by miners and traders, may be

optimized to obtain the maximum return possible for more complex concentrates that require some

level of blending before being sold.

6.4. Applying the models and methodologies to anticipate future business performance

A copper concentrate trading company has been valued, computing the uncertainties related to

copper concentrate price fluctuations over time through the modelling of copper concentrate prices.

Alternative valuations for the trading company have been determined employing Real Options,

where the growth or abandon conditions for the trader’s activity have been linked to the copper

concentrate price in each year.

Our research focuses on the trading of copper concentrates as a standalone element, being separated

from the extraction or the transformation stages, which fills a gap in literature. The outcome suggests

that the methodology laid out in this paper may well be employed by commodity traders, regardless

of the commodities they trade with, to anticipate potentially profitable opportunities or avoid

unnecessary risks.

As far as we know, this research manages for the first time to separate the trading element from the

extraction and processing factors in commodity markets, tackling the existent void in the literature

by shedding light on trading houses’ financials. At the same time it provides further insight into

physical commodity trading and, more specifically, on how copper concentrate traders run their

121
businesses. The results of our research indicate that the methodology employed can be applied for

estimating the future performance of copper concentrate traders through the valuation of their

company.

The outcome achieved suggests that the methodology explained in this research may well be

employed primarily, but not exclusively, by copper concentrate traders to anticipate potentially

profitable opportunities that may arise in the foreseeable future. It may also add flexibility to their

strategical decisions or help to mitigate upcoming threats for their businesses.

To better understand the physical commodities trading business and the uncertainties associated

with it, a copper concentrate trading company has been valued, using copper concentrate trading as

a sample of the richer and more complex physical commodities trading business model. The

uncertainties under which copper concentrate trading is carried out have been reckoned to properly

address the valuation problem with precision. An acceptably reliable outcome is also obtained, such

as prices and discount fluctuations, resorting to real market data to reflect the conditions a copper

concentrate trader is submitted to, as well as estimating its most likely evolution.

6.5. Limitations and future lines

As a limitation of this research, mainly with what regards to TC/RC forecasts, we should point out

the timespan of the data considered, compared to those of other forecasting works, on commodity

prices for example, which use broader timespans. For our case, we have considered the maximum

available, and sufficiently reliable data on TC/RC benchmark levels, starting back in 2004, as there

is no reliable data beyond this year.

122
In addition, we have used four measures of error which are among the most frequently used to

compare the accuracy of different models. However, other measures could have been used at an

individual level to test each model’s accuracy. Once TC/RC annual benchmark levels are able to be

forecasted with a certain level of accuracy, future research should go further into this research

through the exploration of the potential impact that other market factors may have on these

discounts.

Furthermore, a similar problem appears when analysing trading companies’ activities through Real

Options, since publicly available data from companies engaged in concentrates trading, processing

or production is fairly limited as well. These companies tend to keep utter discretion in relation to

the economic and technical details of their transactions. Nonetheless, the reliability of our

methodology is very dependable on the accuracy of the individual forecasting techniques employed

for each of the components that build up the price of concentrates. In this sense, one of the limitations

of our research is the fact that our copper concentrate benchmark price model’s forecasting accuracy

decreases for further-out forecasting horizons, mainly beyond six months or one year, although it

doesn’t have a serious impact in the practical usefulness of the model, as copper concentrate

transactions usually occur in shorter time periods.

The accuracy of the copper concentrate benchmark price model is also a function of the concentrate

layout, even though it is related to the accuracy of Schwartz and Smith’s two-factor model and that

of the Linear Exponential Smoothing model for TC/RC forecasting, having different levels of

accuracy for concentrates with only one payable element than for those with more than one. Future

research should strive to contrast the model’s outcomes with data from market participants if they

became available to evaluate potential ways of improvement if needed.

123
Our research has centred on copper concentrate trading to understand the usefulness of our

methodology in a context of high uncertainty with low available information and where price

volatility tends to be on the upper side. Nonetheless, the main limitation lies on making a well-

balanced approach to the trader’s business model without adding excessive complexity to the

analysis. Thus, as a first approach, long-term supply contracts normally priced below broader

market conditions, in combination with hedging practices, mostly to secure contango margins

(higher prices in further-out future contracts), have not been included in this research. Future

research should also consider these common practices of commodity traders to help differentiate

more from a common retailer and to define more precisely the specificities of commodity trading.

In addition, further extensions of this research should be adapted to not only contemplate

abandonment as the only alternative if market conditions become adverse, hence introducing a

scale-down or diversification options. These different Real Options approaches may help give a

clearer picture of the strategical decisions implemented by commodity traders. They would more

probably rather diversify their commodity portfolio or simply reduce their trading operations to a

minimum level if market conditions are not favourable, instead of shutting down their operations

completely. This is provided there is a low level of their fixed costs along with the importance of

keeping their relational capital.

On the other hand, the access to useful and reliable data on copper concentrate traders is notably

constrained, as most major traders are still in private hands and therefore are not required to disclose

much of the information regarding their operations or assets ownership. Deeper access to data from

larger and more diversified commodity traders would thus provide more accurate results. This is

one of the most relevant lines of development for this research. Additionally, the inclusion of several

frequent activities in copper concentrates trading which are key to maximise returns, such as

124
blending, would certainly provide a richer perspective. Nonetheless, as the market becomes more

volatile and unstable, also given the situation stemming from the COVID-19 pandemic, the

usefulness of the real option technique has become more interesting, as it can also be adapted to

different sectors and business models. This enables companies across the board to anticipate key

managerial decisions without having to take them in a rush and under pressure due to a critical

market downturn.

Finally, Environmental, Social and Governance criteria (ESG) have not been taken into account in

this research, as we have focused in purely technical and financial challenges within copper

concentrate markets. However, future research should consider including ESG impact on copper

concentrate market, as copper concentrate prices might ultimately get affected by restrictions

imposed onto the importation of copper concentrate with high levels of certain punishable, highly

contaminating elements. The effects of ESG policies over the copper concentrates trading business

would also be another topic of interest for potential future research, analysing how observing the

elemental practices of ESG would impact the traders’ normal workflow and financials.

125
 References

Abdel Sabour, S.A., Dimitrakopoulos, R.G., Kumral, M., 2008. Mine design selection

under uncertainty. Trans. Institutions Min. Metall. Sect. A Min. Technol. 117, 53–

64. https://doi.org/10.1179/174328608X343065

Aharon, D.Y., Kroll, Y., Riff, S., 2019. Degree of free cash flow leverage. Rev.

Account. Financ. 18, 346–365. https://doi.org/10.1108/RAF-03-2018-0061

Ajak, A.D., Topal, E., 2015. Real option in action: An example of flexible decision

making at a mine operational level. Resour. Policy 45, 109–120.

https://doi.org/10.1016/j.resourpol.2015.04.001

Aminrostamkolaee, B., Scroggs, J.S., Borghei, M.S., Safdari-Vaighani, A.,

Mohammadi, T., Hossein Pourkazemi, M., 2017. Valuation of a hypothetical

mining project under commodity price and exchange rate uncertainties by using

numerical methods. Resour. Policy 52, 296–307.

https://doi.org/10.1016/j.resourpol.2017.04.004

Amram, M., Kulatilaka, N., 1998. Real Options: Managing Strategic Investment in an

Uncertain World (Financial Management Association Survey and Synthesis).

Harvard Business School Press, Boston, MA.

AQSIQ, 2017. Announcement of the Ministry of Commerce of the Ministry of

Environmental Protection of the General Administration of Quality Supervision,

Inspection and Quarantine on Promulgating the Limits of Toxic and Harmful

Elements in Imported Copper Concentrates ( [WWW Document]. URL

http://www.aqsiq.gov.cn/xxgk_13386/jlgg_12538/zjgg/2017/201712/t20171208_5

08547.htm (accessed 6.10.19).

126
Asad, M.W.A., 2007. Optimum cut‐off grade policy for open pit mining operations

through net present value algorithm considering metal price and cost escalation.

Eng. Comput. 24, 723–736. https://doi.org/10.1108/02644400710817961

Aurubis AG, 2015. Aurubis Copper Mail December 2015 - No. 131.

Aurubis AG, 2016. Aurubis Copper Mail February 2016 - No. 132.

Barlow, M., Gusev, Y., Lai, M., 2004. CALIBRATION OF MULTIFACTOR

MODELS IN ELECTRICITY MARKETS. Int. J. Theor. Appl. Financ. 07, 101–

120. https://doi.org/10.1142/S0219024904002396

Barton, K., Lawryshyn, Y., 2011. Integrating Real Options with Managerial Cash Flow

Estimates. Eng. Econ. 56, 254–273.

https://doi.org/10.1080/0013791X.2011.601403

Berk, A.S., Podhraski, D., 2018. Superiority of Monte Carlo simulation in valuing real

options within public–private partnerships. Risk Manag. 20, 1–28.

https://doi.org/10.1057/s41283-017-0025-9

Black, F., Scholes, M., 1973. The Pricing of Options and Corporate Liabilities. J. Polit.

Econ. 8, 637–654.

Blas, J., Farchy, J., 2021. The World for Sale.

Botín, J., Del Castillo, M.F., Guzman, R., 2012. A Real Option Application to Manage

Resk Related to Intrinsic Variables of a Mine Plan: A Case Study on

Chuquicamata U.G Project. J. South. African Inst. Min. Metall. 113, 583–592.

Boulamanti, A., Moya, J.A., 2016. Production costs of the non-ferrous metals in the EU

and other countries: Copper and zinc. Resour. Policy 49, 112–118.

https://doi.org/10.1016/j.resourpol.2016.04.011

127
Brennan, M.J., Schwartz, E.S., 1985. Evaluating Natural Resource Investments. J. Bus.

58, 135–137.

Brown, R.G., 1963. Smoothing, forecasting and prediction of discrete time series.

Prentice-Hall quantitative methods series.

Buchan, D., Errington, C., 2016. Commodities demystified: A guide to trading and the

global supply chain.

Buncic, D., Moretto, C., 2015. Forecasting copper prices with dynamic averaging and

selection models. North Am. J. Econ. Financ. 33, 1–38.

https://doi.org/10.1016/j.najef.2015.03.002

Chatfield, C., 1978. The Holt-Winters Forecasting Procedure. Appl. Stat. 27, 264.

https://doi.org/10.2307/2347162

Chen, W., Wang, M., Li, X., 2016. Analysis of copper flows in the United States: 1975–

2012. Resour. Conserv. Recycl. 111, 67–76.

https://doi.org/10.1016/j.resconrec.2016.04.014

Choudhury, A., Jones, J., 2014. Crop Yield Prediction Using Time Series Models. J.

Econ. Econ. Educ. Res. 15, 53–68.

Coutie, G. et al., 1964. Short-Term Forecasting. ICT Monogr. No. 2, Edinburgh Oliver

Boyd.

Crundwell, F.K., 2008. Finance for Engineers. Springer London, London.

https://doi.org/10.1007/978-1-84800-033-9

Damodaradan, A., 2006. Security Analysis for Investment And Corporate Finance.

Wiley Finance, New Jersey.

Damodaradan, A., 2012. Investment Valuation: Tools and Techniques for Determining

128
 the Value of Any. Wiley Finance, New Jersey.

Dehghani, H., Ataee-pour, M., Esfahanipour, A., 2014. Evaluation of the mining

projects under economic uncertainties using multidimensional binomial tree.

Resour. Policy 39, 124–133. https://doi.org/10.1016/j.resourpol.2014.01.003

Dehghani, H., Bogdanovic, D., 2018. Copper price estimation using bat algorithm.

Resour. Policy 55, 55–61. https://doi.org/10.1016/j.resourpol.2017.10.015

Delbeke, K., Rodriguez, P.H., 2014. Copper Concentrates Environmental and Human

Health Hazard Classification.

Díaz-Borrego, F.J., Miras-Rodríguez, M. del M., Escobar-Pérez, B., 2019. Looking for

Accurate Forecasting of Copper TC/RC Benchmark Levels. Complexity 2019, 1–

16. https://doi.org/10.1155/2019/8523748

Dixit, A.K., Pindyck, R.S., 1994. Investment Under Uncertainty, Investment Under

Uncertainty. Princeton University Press. https://doi.org/10.2307/1909571

Drouven, D.B., Faust, E., 2015. Aurubis DVFA Analyst Conference Report 2015.

Duan, H., Lei, G.R., Shao, K., 2018. Forecasting Crude Oil Consumption in China

Using a Grey Prediction Model with an Optimal Fractional-Order Accumulating

Operator. Complexity 2018, 1–12. https://doi.org/10.1155/2018/3869619

EY, 2017. EY Mining and Metals Commodity Briefcase.

Fernandez, P., Carabias, J.M., 2007. The Danger of Using Calculated Betas (El Peligro

De Utilizar Betas Calculadas). SSRN Electron. J. DI 685, 1–18.

https://doi.org/10.2139/ssrn.897700

Foo, N., Bloch, H., Salim, R., 2018. The optimisation rule for investment in mining

projects. Resour. Policy 55, 123–132.

129
 https://doi.org/10.1016/j.resourpol.2017.11.005

Freeport-McMoRan, 2018. Freeport-McMoRan 2018 Annual Report.

Gajek, L., Kuciński, Ł., 2017. Complete discounted cash flow valuation. Insur. Math.

Econ. 73, 1–19. https://doi.org/10.1016/j.insmatheco.2016.12.004

Gardner, E.S., 2006. Exponential smoothing: The state of the art—Part II. Int. J.

Forecast. 22, 637–666. https://doi.org/10.1016/j.ijforecast.2006.03.005

Geman, H., Smith, W.O., 2013. Theory of storage, inventory and volatility in the LME

base metals. Resour. Policy 38, 18–28.

https://doi.org/10.1016/j.resourpol.2012.06.014

Ghaffari, H., Huang, J. (John), Jones, K., Gray, J.H., Hammett, R., Parolin, R.W.,

Kinakin, D., Lechner, M.J., Parkinson, J.G., Pelletier, P., Brazier, N., Allard, S.,

Lipiec, T., Ramirez, M., 2016. Seabridge Gold KSM Project, Prefeasibility Study

Update and Preliminary Economic Assessment.

Glöser, S., Soulier, M., Tercero Espinoza, L.A., 2013. Dynamic Analysis of Global

Copper Flows. Global Stocks, Postconsumer Material Flows, Recycling Indicators,

and Uncertainty Evaluation. Environ. Sci. Technol. 47, 6564–6572.

https://doi.org/10.1021/es400069b

Gómez-Valle, L., Martínez-Rodríguez, J., 2013. Advances in pricing commodity

futures: Multifactor models. Math. Comput. Model. 57, 1722–1731.

https://doi.org/10.1016/j.mcm.2011.11.015

Guj, P., Chandra, A., 2019. Comparing different real option valuation approaches as

applied to a copper mine. Resour. Policy 61, 180–189.

https://doi.org/10.1016/j.resourpol.2019.01.020

130
Guzmán, J.I., Silva, E., 2018. Copper price determination: fundamentals versus non-

fundamentals. Miner. Econ. 31, 283–300. https://doi.org/10.1007/s13563-017-

0130-y

Hamid, M.F.A., Shabri, A., 2017. Palm oil price forecasting model: An autoregressive

distributed lag (ARDL) approach. p. 030026. https://doi.org/10.1063/1.4982864

Hammarstrom, J.M., Zientek, M.L., Parks, H.L., Dicken, C.L., U.S. Geological Survey

Global Copper Mineral Resource Assessment Team, 2019, United States

Geological Survey, Hammarstrom, J.M., Zientek, M.L., Parks, H.L., Dicken, C.L.,

2019. Assessment of undiscovered copper resources of the world, 2015, Scientific

Investigations Report. https://doi.org/https://doi.org/10.3133/sir20185160

Harrison, P.J., 1967. Exponential Smoothing and Short-Term Sales Forecasting.

Manage. Sci. 13, 821–842. https://doi.org/10.1287/mnsc.13.11.821

Hazra, T., Samanta, B., Dey, K., 2019. Real option valuation of an Indian iron ore

deposit through system dynamics model. Resour. Policy 60, 288–299.

https://doi.org/10.1016/j.resourpol.2019.01.002

Heaney, R., 2002. Does knowledge of the cost of carry model improve commodity

futures price forecasting ability? Int. J. Forecast. 18, 45–65.

Hloušková, Z., Ženíšková, P., Prášilová, M., 2018. Comparison of Agricultural Costs

Prediction Approaches. Agris on-line Pap. Econ. Informatics 10, 3–13.

https://doi.org/10.7160/aol.2018.100101

International Copper Study Group, 2017. The World Copper Factbook 2017.

International Copper Study Group, 2022. The World Copper Factbook 2022.

Inthavongsa, I., Drebenstedt, C., Bongaerts, J., Sontamino, P., 2016. Real options

131
 decision framework: Strategic operating policies for open pit mine planning.

Resour. Policy 47, 142–153. https://doi.org/10.1016/j.resourpol.2016.01.009

Issler, J.V., Lima, L.R., 2009. A panel data approach to economic forecasting: The bias-

corrected average forecast. J. Econom. 152, 153–164.

https://doi.org/10.1016/j.jeconom.2009.01.002

Issler, J.V., Rodrigues, C., Burjack, R., 2014. Using common features to understand the

behavior of metal-commodity prices and forecast them at different horizons. J. Int.

Money Financ. 42, 310–335. https://doi.org/10.1016/j.jimonfin.2013.08.017

Jacques, S., Simondet, A., 2016. Traders or Commodity Finance Banks. Freight and

Spreads 1–15.

Jaunky, V.C., 2013. A cointegration and causality analysis of copper consumption and

economic growth in rich countries. Resour. Policy 38, 628–639.

https://doi.org/10.1016/j.resourpol.2013.10.001

Jia, J., Kang, S.B., 2022. Do the basis and other predictors of futures return also predict

spot return with the same signs and magnitudes? Evidence from the LME. J.

Commod. Mark. 25, 100187. https://doi.org/10.1016/j.jcomm.2021.100187

Johansson, J., 2007. Boliden Capital Markets Day 2007.

Jovanović, I., Stanimirović, P., Živković, Ž., 2013. Environmental and Economic

Criteria in Ranking of Copper Concentrates. Environ. Model. Assess. 18, 73–83.

https://doi.org/10.1007/s10666-012-9327-1

Jovanović, I.M., Stanimirović, P.S., 2012. A Blending Problem in Copper Production.

Environ. Model. Assess. 17, 495–503. https://doi.org/10.1007/s10666-012-9309-3

Kaiser, M.J., 2013. Oil and Gas Company Production, Reserves, and Valuation. J.

132
 Sustain. Energy Eng. 1, 220–235. https://doi.org/10.7569/JSEE.2013.629527

Kaiser, M.J., 2015. Coal Production and Processing Technology, Coal Production and

Processing Technology. CRC Press. https://doi.org/10.1201/b19352

Kalekar, P., 2004. Time series forecasting using Holt-Winters exponential smoothing.

Kanwal Rekhi Sch. Inf. Technol. 1–13.

Kalman, R.E., 1960. A New Approach to Linear Filtering and Prediction Problems.

Trans. ASME-Journal Basic Eng. 82, 35–45.

Katanga Mining Limited, 2018. Management’s Discussion and Analysis for the years

ended December 31,2018 and 2017.

KGHM Polska Miedz, 2018. The Management Board’s Report on the Activities of

KGHM Polska Miedz S.A. and of the KGHM Polska Miedz S.A. Group in 2018.

Khashei, M., Bijari, M., 2010. An artificial neural network (p,d,q) model for timeseries

forecasting. Expert Syst. Appl. 37, 479–489.

https://doi.org/10.1016/j.eswa.2009.05.044

Khashei, M., Bijari, M., 2011. A novel hybridization of artificial neural networks and

ARIMA models for time series forecasting. Appl. Soft Comput. 11, 2664–2675.

https://doi.org/10.1016/j.asoc.2010.10.015

Kriechbaumer, T., Angus, A., Parsons, D., Rivas Casado, M., 2014. An improved

wavelet-ARIMA approach for forecasting metal prices. Resour. Policy 39, 32–41.

https://doi.org/10.1016/j.resourpol.2013.10.005

Li, K., Mohanram, P., 2019. Fundamental Analysis: Combining the Search for Quality

with the Search for Value. Contemp. Account. Res. 36, 1263–1298.

https://doi.org/10.1111/1911-3846.12466

133
Li, Y., Hu, N., Li, G., Yao, X., 2012. Forecasting Mineral Commodity Prices with

ARIMA-Markov Chain, in: 2012 4th International Conference on Intelligent

Human-Machine Systems and Cybernetics. IEEE, pp. 49–52.

https://doi.org/10.1109/IHMSC.2012.18

Liu, C., Hu, Z., Li, Y., Liu, S., 2017. Forecasting copper prices by decision tree

learning. Resour. Policy 52, 427–434.

https://doi.org/10.1016/j.resourpol.2017.05.007

Makridakis, S., Andersen, A., Carbone, R., Fildes, R., Hibon, M., Lewandowski, R.,

Newton, J., Parzen, E., Winkler, R., 1982. The accuracy of extrapolation (time

series) methods: Results of a forecasting competition. J. Forecast. 1, 111–153.

https://doi.org/10.1002/for.3980010202

Makridakis, S., Hibon, M., 1979. Accuracy of Forecasting: An Empirical Investigation.

J. R. Stat. Soc. Ser. A 142, 97. https://doi.org/10.2307/2345077

Makwasha, T., Wright, J., Silvapulle, P., 2019. Panel data analysis of multi-factor

capital asset pricing models. Appl. Econ. 51, 6459–6475.

https://doi.org/10.1080/00036846.2019.1619019

Marathe, R.R., Ryan, S.M., 2005. On The Validity of The Geometric Brownian Motion

Assumption. Eng. Econ. 50, 159–192.

https://doi.org/10.1080/00137910590949904

Markowitz, H., 1952. Portfolio Selection. J. Finance 7, 77–91.

https://doi.org/10.1111/j.1540-6261.1952.tb01525.x

Markowitz, H.M., 1959. Portfolio Selection: Efficient Diversification of Investments.

Yale University Press, New Haven, CT.

McClain, J.O., 1974. Dynamics of Exponential Smoothing with Trend and Seasonal

134
 Terms. Manage. Sci. 20, 1300–1304. https://doi.org/10.1287/mnsc.20.9.1300

McClain, J.O., Thomas, L.J., 1973. Response-Variance Tradeoffs in Adaptive

Forecasting. Oper. Res. 21, 554–568. https://doi.org/10.1287/opre.21.2.554

McDonald, R., Siegel, D., 1986. The Value of Waiting to Invest. Q. J. Econ. 101, 707.

https://doi.org/10.2307/1884175

Meersman, S., Rechtsteiner, R., Sharp, G., 2012. The Dawn of a New Order in

Commodity Trading.

Metal Bulletin TC/RC Index [WWW Document], 2017. . Met. Bull. URL

https://www.metalbulletin.com/Article/3319574/Copper/PRICING-NOTICE-

Specification-of-Metal-Bulletin-copper-concentrates-TCRC-index.html

Misund, B., Osmundsen, P., Sikveland, M., 2015. International Oil Company Valuation:

the Effect of Accounting Method and Vertical Integration. Pet. Account. Financ.

Manag. J. 34, 1–20.

Montgomery, D.C., Johnson, L.A., 1976. Forecasting and Time Series Analysis. New

York.

Mudd, G.M., Weng, Z., Jowitt, S.M., 2013. A Detailed Assessment of Global Cu

Resource Trends and Endowments. Econ. Geol. 108, 1163–1183.

https://doi.org/10.2113/econgeo.108.5.1163

Murty, K.G., 2020. International Series in Operations Research & Management Science

Models for Optimum Decision Making Crude Oil Production and Refining.

Nakazato, Y., 2017. Sumitomo Fiscal Year 2016 Report.

Nau, R., 2014. Forecasting with moving averages. Fuqua Sch. Business, Duke Univ. 1–

28.

135
Norgate, T., Jahanshahi, S., 2010. Low grade ores – Smelt, leach or concentrate? Miner.

Eng. 23, 65–73. https://doi.org/10.1016/j.mineng.2009.10.002

OEC, 2023. Copper Ores and Concentrates Trade [WWW Document]. URL

https://oec.world/en/profile/hs/copper-ores-and-concentrates

Organisation for Economic Co-operation and Development, World Bank Group,

International Monetary Fund, World Bank Group, 2017. Addressing the

Information Gaps on Prices of Minerals Sold in an Intermediate Form.

Pinheiro, C., Senna, V., 2016. Price Forecasting Through Multivariate Spectral

Analysis: Evidence for Commodities of BM and Fbovespa. Brazilian Bus. Rev. 13,

129–157. https://doi.org/10.15728/bbr.2016.13.5.6

Pirrong, C., 2013. The Economics of Commodity Trading Firms.

Robins, N., 2012. The Corporation that Changed the World, Second Edi. ed. Pluto

Press, New York.

Rozenbaum, O., 2019. EBITDA and Managers’ Investment and Leverage Choices.

Contemp. Account. Res. 36, 513–546. https://doi.org/10.1111/1911-3846.12387

Sabet, A.H., Heaney, R., 2017. Real options and the value of oil and gas firms: An

empirical analysis. J. Commod. Mark. 6, 50–65.

https://doi.org/10.1016/j.jcomm.2017.05.001

Samis, M., Davis, G.A., Laughton, D., Poulin, R., 2005. Valuing uncertain asset cash

flows when there are no options: A real options approach. Resour. Policy 30, 285–

298. https://doi.org/10.1016/j.resourpol.2006.03.003

Sánchez Lasheras, F., de Cos Juez, F.J., Suárez Sánchez, A., Krzemień, A., Riesgo

Fernández, P., 2015. Forecasting the COMEX copper spot price by means of

136
 neural networks and ARIMA models. Resour. Policy 45, 37–43.

https://doi.org/10.1016/j.resourpol.2015.03.004

Saramak, D., 2011. Optimal Production of Electrolytic Copper Determined by the

Concentration and Distribution of Copper Concentrates to Smelters on the

Example of KGHM. Arch. Metall. Mater. 56. https://doi.org/10.2478/v10172-011-

0067-z

Sarkis, J., Tamarkin, M., 2005. Real Options Analysis for “Green Trading”: The Case

of Greenhouse Gases. Eng. Econ. 50, 273–294.

https://doi.org/10.1080/00137910500227208

Savolainen, J., 2016. Real options in metal mining project valuation: Review of

literature. Resour. Policy 50, 49–65.

https://doi.org/10.1016/j.resourpol.2016.08.007

Schachler, J., 2017. Aurubis AG Interim Report 2017. Frankfurt.

Schlesinger, M.E., Davenport, W.G., 2011. Extractive Metallurgy of Copper — Fifth

Edition, 5th ed, Mineral Resources Engineering. Elsevier B.V.

Schwartz, E., Smith, J.E., 2000. Short-Term Variations and Long-Term Dynamics in

Commodity Prices. Manage. Sci. 46, 893–911.

https://doi.org/10.1287/mnsc.46.7.893.12034

Seabridge Gold, 2010. Appendix B, Smelting and Refining Terms - KSM Project

Prefeasibility Study and Preliminary Economic Asessment.

Shafiee, S., Topal, E., 2010. An overview of global gold market and gold price

forecasting. Resour. Policy 35, 178–189.

https://doi.org/10.1016/j.resourpol.2010.05.004

137
Shao, Y.E., Dai, J.-T., 2018. Integrated Feature Selection of ARIMA with

Computational Intelligence Approaches for Food Crop Price Prediction.

Complexity 2018, 1–17. https://doi.org/10.1155/2018/1910520

Sharma, R.K., 2016. Forecasting Gold price with Box Jenkins Autoregressive Integrated

Moving Average Method. J. Int. Econ. Vol. 7, 32–60.

Sharpe, W.F., 1964. Capital Asset Prices: A Theory of Market Equilibriym Under

Conditions of Risk. J. Finance 19, 425–442. https://doi.org/10.1111/j.1540-

6261.1964.tb02865.x

Shaw, A., 2015. AngloAmerican Copper Market Outlook. Nanning (China).

Siña, M., Guzmán, J.I., 2019. Real option valuation of open pit mines with two

processing methods. J. Commod. Mark. 13, 30–39.

https://doi.org/10.1016/j.jcomm.2018.05.003

Smith, D.D., Pennathur, A.K., 2019. Signaling Versus Free Cash Flow Theory: What

Does Earnings Management Reveal About Dividend Initiation? J. Accounting,

Audit. Financ. 34, 284–308. https://doi.org/10.1177/0148558X17724051

Soderstrom, U., 2008. Boliden Capital Markets Day 2008. Copper Smelter Revenue

Stream.

Svante, N., 2011. Boliden Capital Markets Day September 2011. Smelters.

Sverdrup, H.U., 2016. Modelling global extraction, supply, price and depletion of the

extractable geological resources with the LITHIUM model. Resour. Conserv.

Recycl. 114, 112–129. https://doi.org/10.1016/j.resconrec.2016.07.002

Tartakovsky, D.M., Broyda, S., 2011. PDF equations for advective–reactive transport in

heterogeneous porous media with uncertain properties. J. Contam. Hydrol. 120–

138
 121, 129–140. https://doi.org/10.1016/j.jconhyd.2010.08.009

Teck, 2012. Teck Modelling Workshop 2012.

Teck, 2015. Teck Modeling Workshop 2015.

The London Metal Exchange, 2019. Rules and Regulations.

Thompson, M., Barr, D., 2014. Cut-off grade: A real options analysis. Resour. Policy

42, 83–92. https://doi.org/10.1016/j.resourpol.2014.10.003

Trafigura Beheer B.V., 2021. Trafigura Metals and Minerals [WWW Document]. URL

https://www.trafigura.com/products-and-services/metals-and-minerals/ (accessed

4.22.21).

Tresierra Tanaka, Á., Carrasco Montero, C.M., 2016. Valorización de opciones reales:

modelo Ornstein-Uhlenbeck. J. Econ. Financ. Adm. Sci. 21, 56–62.

https://doi.org/10.1016/j.jefas.2016.07.001

U.S. Geological Survey, 2017. Copper, Mineral Commodity Summaries.

https://doi.org/10.3133/fs20143004

U.S. Geological Survey, 2023. Mineral commodity summaries 2023, Mineral

commodity summaries 2023. https://doi.org/10.3133/mcs2023

Uhlenbeck, G.E., Ornstein, L.S., 1930. On the Theory of the Brownian Motion. Phys.

Rev. 36, 823–841. https://doi.org/10.1103/PhysRev.36.823

United Nations, 2016. International Trade Statistics Yearbook Vol. II.

Wets, R.J.B., Rios, I., 2015. Modeling and estimating commodity prices: copper prices.

Math. Financ. Econ. 9, 247–270. https://doi.org/10.1007/s11579-014-0140-2

Willbrandt, P., Faust, E., 2013. Aurubis DVFA Analyst Conference Report 2013.

Willbrandt, P., Faust, E., 2014. Aurubis DVFA Analyst Conference Report 2014.

139
Xiong, T., Li, C., Bao, Y., Hu, Z., Zhang, L., 2015. A combination method for interval

forecasting of agricultural commodity futures prices. Knowledge-Based Syst. 77,

92–102. https://doi.org/10.1016/j.knosys.2015.01.002

Yaari, U., Nikiforov, A., Kahya, E., Shachmurove, Y., 2016. Finance methodology of

Free Cash Flow. Glob. Financ. J. 29, 1–11.

https://doi.org/10.1016/j.gfj.2015.05.003

Yeh, I.-C., Lien, C.-H., 2020. Evaluating real estate development project with Monte

Carlo based binomial options pricing model. Appl. Econ. Lett. 27, 307–324.

https://doi.org/10.1080/13504851.2019.1616049

Zhang, K., Nieto, A., Kleit, A.N., 2015. The real option value of mining operations

using mean-reverting commodity prices. Miner. Econ. 28, 11–22.

https://doi.org/10.1007/s13563-014-0048-6

Zhang, W., Han, G., Feng, Y., Lloret, J., 2017. IRPL: An energy efficient routing

protocol for wireless sensor networks. J. Syst. Archit. 75, 35–49.

https://doi.org/10.1016/j.sysarc.2017.03.006

140
 Appendix: Published Papers

• Díaz, F.J.; Miras, M.M.; Escobar, B. (2019): “Looking for accurate forecasting of Copper

TC/RC Benchmark Levels”. Complexity, 2019 –ID8523748. JCR, Impact factor 2019: 2.462

(Q2 Mathematics, Interdisciplinary Applications -28/106).

https://doi.org/10.1155/2019/8523748

141
• Díaz, F.J.; Escobar, B.; Miras, M.M. (2021): “Estimating copper concentrates benchmark

prices under dynamic market conditions”. Resources Policy, 70, 101959. JCR, Impact factor

2021: 5.634 (Q1 Environmental Studies - 20/125).

https://doi.org/10.1016/j.resourpol.2020.101959

142