2007:111
M A S T E R ' S T H E S I S
Investments in primary aluminium production
Location choice and the impact of electricity price
Thomas Ejdemo
Luleå University of Technology D Master thesis
Economics
Department of Business Administration and Social Sciences Division of Social sciences
Investments in primary aluminium production: Location choice and the impact of electricity price
Thomas Ejdemo
ABSTRACT
Primary aluminium production requires huge amounts of electricity. The cost of electricity is thus of great importance for firms in the industry. Electricity prices in the western world tends to increase over time at the same time as many producers face expiring long term power contracts, thus resulting in increased production costs and reduced competitiveness. Hence, reports from media suggest a growing tendency towards a re-location of capacity to countries capable of offering cheap electricity.
This study was carried out in order to shed further light on which factors firms in the industry consider in their choice of location for investments. The purpose of this thesis is to explore the determinants of investment in primary aluminium production and asses how these determinants affect location choice for investments. Using smelter-level data on a majority of the world’s primary producers, a count data analysis was performed in order to asses the impact of factor prices and agglomeration effects. Due to the lack of country-level data for key variables, smelters were grouped into 12 regions.
Contrary to expectations, no significant evidence was found for the importance of electricity prices. However, access to cheap labour and agglomeration effects appears to attract investments. This implies that high levels of sunk costs make firms prone to upgrading existing capacity, but may also suggest that agglomeration effects such as technical know-how and industry infrastructure are considered in the choice of location for investments in new plants.
SAMMANFATTNING
Primäraluminiumproduktion kräver enorma mängder elektricitet vilket innebär att elpriset är av yttersta vikt för aktörer i industrin. Västvärldens elpriser tenderar att stiga samtidigt som långtidskontrakt för många smältverks elförsörjning har börjat löpa ut.
Detta har lett till rapporter i media om en begynnande tendens till omlokaliseringar av produktionskapacitet från högkostnadsländer till länder kapabla att erbjuda billigare el.
Denna studie utfördes för att öka kunskapen om vilka faktorer företag inom aluminiumindustrin beaktar när de väljer var i världen investeringar ska förläggas.
Syftet med uppsatsen är att undersöka vilka faktorer som påverkar investeringar i primäraluminiumproduktion, och hur dessa faktorer påverkar var investeringar sker rent geografiskt. Svårigheter att anskaffa viktiga data på nationsnivå innebar att en regional ansats gjordes. Data på smältverksnivå analyserades med hjälp av count data regression för att analysera om och hur priser på produktionsfaktorer samt agglomerationseffekter påverkar valet av var kapacitetsinvesteringar förläggs. Den regionala ansatsen innebar att smältverken grupperades i 12 regioner.
I motsats till vad som förväntades påträffades inga signifikanta bevis för att elpriset skulle ha betydelse för var investeringar förläggs geografiskt. Resultaten indikerar att tillgång till billig arbetskraft samt agglomerationseffekter attraherar investeringar. Detta implicerar att höga uppstartningskostnader gör företag benägna att investera i uppgraderingar av befintliga anläggningar snarare än att starta upp nya smältverk, men kan också vara en indikation för att agglomerationseffekter såsom teknisk ”know-how”
och den industriella infrastrukturen kan påverka var investeringar förläggs.
Acknowledgements
I am grateful towards my supervisor for this master’s thesis, Dr. Jerry Blomberg.
Besides possessing an impressive level of patience which should not go unrecognized, Jerry has provided me with some much needed guidance and advice for this thesis.
CONTENTS
Chapter 1
INTRODUCTION...1
1.1 Background...1
1.2 Purpose...2
1.3 Scope and Method……...2
1.4 Previous studies...3
1.5 Outline...4
Chapter 2 THEORETICAL FRAMEWORK...5
2.1 Basics of location theory...5
2.2 Choice of location and spatial variations in production costs...5
2.2.1 Natural resources and availability...6
2.2.2 Weber’s basics of location theory...6
2.2.3 Economies of agglomeration...7
2.2.4 Diseconomies of agglomeration...9
2.3 Choice of location and the investment decision in process industries…………..9
Chapter 3 ALUMINIUM PRODUCTION AND INPUT COSTS...11
3.1 Aluminium production...11
3.2 Major cost components of primary aluminium smelting………….………...….12
3.3 Regional variations in input prices………...………...13
Chapter 4 MODEL, DATA AND VARIABLES………...17
4.1 Determinants of investment………17
4.2 Model specification…..………18
4.3 Smelter data and validity issues…………...………..20
Chapter 5
RESULTS, DISCUSSION AND ANALYSIS………..24
5.1 Estimated results……..………24
5.1.1 Results of the regression..………...24
5.1.2 Results of the regression when China was disregarded……...………25
5.2 Discussion and analysis………...26
5.3 Results compared to previous studies………28
Chapter 6 CONCLUDING REMARKS………29
REFERENCES………..31
APPENDIX A……….33
TABLES AND FIGURES
TABLES
Table 3.1 COC shares of major cost components of primary aluminium smelting in
2003……….12
Table 4.1 Summation of independent variables………..18
Table 4.2 Regions as defined by the CRU Group (2004)………...21
Table 4.3 Variables and descriptive statistics……….23
Table 5.1 Estimated signs………..25
FIGURES Figure 3.1 Flow chart over process chain………12
Figure 3.2 Regional average smelter electricity costs, 2003………...14
Figure 3.3 Regional average smelter alumina costs, 2003………..15
Chapter 1 INTRODUCTION
1.1 Background
Primary aluminium is produced in a highly energy intense process. The electricity requirements for primary producers are huge and accounts for a considerable portion of production costs. Primary producers are therefore concerned with securing a steady supply of cheap electricity for their plants. Economic theory1 suggests that industries are best served to locate in an area where factors in which the industry is particularly intense are cheap and abundant. This argument implies that, for primary aluminium producers, the location of plants should to some extent depend on prices and availability of electricity as well as raw materials and labour.
In many countries around the world, rising electricity prices have been affecting both industries and households. This is of particular concern to aluminium producers as increased costs for such a crucial input factor may adversely impact the ability to earn profits on a competitive market. Primary producers operating in Europe and North America are in many cases facing expiring long-term electricity contracts and anticipate cost increases to levels that are no longer economically viable.
Because of this, a re-location of production capacity to countries capable of supplying cheap electricity is for many producers a very realistic option, and some evidence supporting such a re-location can already be found. An example of this was expressed by a representative of Alcoa, the world’s biggest producer, stating to the 2005 Financial Times special report on aluminium that Alcoa seeks to build smelters in parts of the world where energy costs less, meanwhile planning the shut-down of older plants in both North America and Europe.
1 See for example Bergman and Johansson (2002)
This serves as a reminder of the 1970´s when oil crises contributed to restructuring the aluminium industry (Peck 1988). Proprietary data presented by the CRU Group (2004) indicates that the significant increase in electricity prices that came as a result of the oil crisis lead to the demise of the Japanese primary aluminium production. The CRU Group (2004) states that this capacity was replaced elsewhere by smelters operating at significantly lower production costs.
For this reason it can be interesting to conduct a study to examine in more detail which factors are important determinants of investment location in primary aluminium production.
1.2 Purpose
The purpose of this thesis is to explore the determinants of investment in primary aluminium production and asses how these determinants affect location choice for investments.
1.3 Scope and Method
This thesis will focus on primary aluminium production during the period 1990-2003.
This study does not consider secondary aluminium produced from recycled aluminium scrap, as secondary production operates under very different conditions regarding raw materials and require much less energy. A regional approach is taken due to lack of country-level data for key variables. Smelters around the world are grouped into 12 regions according to the CRU Group’s (2004) classification.
In this thesis, investments are viewed as discrete events, i.e. investments are not measured in monetary terms but are instead considered countable events that can be attributed to certain surrounding conditions. Proprietary smelter-level data on input factor-costs, factor consumption, production capacity, and smelter production during the period 1990-2003 are compiled and used to calculate regional averages for factor prices.
Investments are defined as an increase in capacity by 5 percent or more compared to the previous year. The number of investments in each region are then calculated and compiled in a set of panel data together with regional average factor prices, existing productive capacity to approximate agglomeration effects, and a dummy variable to
separate between regions. This set of panel data is then subjected to a count data regression analysis using a fixed effects negative binomial model which is estimated under maximum likelihood procedures to examine how the proposed determinants of investment affect the probability of actual investments in a region. The results are then discussed and analysed with respect to the theoretical framework and compared to similar studies, and some conclusions are drawn based on the analysis.
1.4 Previous studies
There is not a great deal of work done in the field of location choice for aluminium smelter investments. However, the aluminium industry shares common traits with other process industries in its large energy requirements and high levels of sunk costs, justifying a comparison to similar industries. This thesis draws much of its inspiration from studies concerning investment behaviour in the pulp and paper industry.
Bergman and Johansson (2002) explores the determinants of investments in the pulp and paper industry using a count data regression analysis and concludes that wage rates, installed production capacity, the price of paper and the USD/ECU exchange rate are the most important determinants.
Lundmark (2001) applies a conditional logit model to a panel of 16 European countries to estimate the relative impact of input prices, output market size and agglomeration effects on the choice of location for investment projects in the European pulp and paper industry, special attention is paid to the price of wastepaper. He concludes that the most important determinants are labour costs, market size and agglomeration effects, and that in contradiction to previous results the location decision is influenced more by market conditions for the finished paper product, than it is by input factors. In a second study, Lundmark (2003) again finds that labour costs, market size and agglomeration effects are the most important determinants of investment location in the European pulp and paper industry.
The importance of agglomeration effects in both Lundmark (2002) and Lundmark (2003), and the impact of installed production capacity found by Bergman and Johansson (2002) indicate that high levels of sunk costs tend to allocate investments to already existing plants. A common trait in all three of these studies is the attention paid
to factor input prices as assumed determinants of investment. Both Bergman and Johansson (2002), and Lundmark (2001) model investments as discrete events. These studies have inspired the work presented in this thesis, where investments are viewed in a discrete choice framework and factor input prices are assumed determinants of investment, estimated together with other assumed determinants by the use of count data regression analysis.
1.5 Outline
The thesis is structured in the following manner. Chapter 2 presents the theoretical framework which is used to explain a firm’s choice of investment location. Chapter 3 gives an outline of the aluminium production process, the energy requirements of the process, major cost shares of production and a discussion about the current situation in the industry. Chapter 4 discusses the determinants of investment which leads to the definition of the independent variables used to explain the location of investments. The model is then presented and discussed. Finally, the data is reviewed and issues concerning its validity are addressed. Chapter 5 presents the results of the regression and provides a discussion and analysis of these based on the theoretical framework in chapter 2, and compares the results to similar studies. Finally, some concluding remarks are given in chapter 6.
Chapter 2
THEORETICAL FRAMEWORK
This chapter outlines the theoretical framework which is used to explain a firm’s choice of investment location. Initially, basic neoclassical location theory is explained, followed by the introduction of spatial variations in factor costs and the impact this has on a firm’s locational choice. Spatial availability of natural resources and its impact on locational choice is reviewed, followed by an introduction to Weber’s work in the field of location theory and economies of agglomeration. The chapter proceeds by explaining the investment decision as a result of the locational choice and adds the option of investing at existing facilities. The chapter is concluded by an interpretation of the Heckscher-Ohlin theory of factor proportions and its implications on where firms choose to locate investments, focusing on the process industry for the sake of this thesis.
If no other source is referred to, this chapter builds on Dicken and Lloyd (1990).
2.1 Basics of location theory
When a firm is setting up production of a good, it is faced with the decision of where to locate its production. According to traditional neoclassic theory in its most simplified form, assuming a homogenous land-surface, this decision will be influenced by the proximity to the market (i.e. customers), and the cost of transportation, which is assumed to be directly proportional to distance.
These arguments are derived under very simplifying assumptions about both economy and land surface, but nevertheless provides the foundations of location theory. Leaving the spatial organization of customers aside, this chapter will focus on the production of a good.
2.2 Choice of location and spatial variations in production costs
When starting up production of a good, the firm requires a supply of basic inputs. In economic literature these are known as production factors, Dicken and Lloyd (1990) defines these factors as: a) Land, simply the land where the facility is placed, b)
Materials, which is the needed raw materials for production, c) Labour, the workforce required for production, d) Capital, constituting start-up-costs such as machinery, equipment and materials, and finally e) Technology, which Dicken and Lloyd (1990) simply labels as “know-how”.
All of these production factors are scarce, thus subjecting them to the laws of supply and demand, and consequently we see the existence of markets for these factors. For instance the labour market, where firms trade money for labour and households give up leisure for wage. This means the production factors give arise to production costs facing the firm.
In the basic theory outlined in section 2.1, the spatial dimension was assumed to consist only of distance. However, adopting a more realistic view it is evident that from a global perspective there are significant differences in factor prices. Labourers will not agree to the same wage in every country, land will be cheaper at some locations than others and so on, this is known as spatial variations in factor costs.
Introducing these factors into the firm’s locational choice creates an incentive for the cost-minimizing producer to seek out the location that will give him the most beneficial conditions, depending on which factors the firm is particularly intense in.
2.2.1 Natural resources and availability
Many industries require a steady supply of raw materials and energy, these production factors are harvested from natural resources such as minerals and wood for raw materials, and natural gas and the damming of rivers for energy. Knowing this, it is clear that when choosing location, firms will consider the spatial distribution of these factors. Depending on material and energy intensivity in production, spatial availability will influence the locational choice accordingly.
2.2.2 Weber’s basics of location theory
Much of the work in location theory is based on the ideas of Alfred Weber, a German economist, whose classic work ”The Theory of the Location of Industries” published in 1909 and the English translation of it in 1929 acknowledged him as a pioneer of the economics of location. Dicken and Lloyd (1990) summarize his basic outlooks on the
location of industries as being influenced by primarily regional factors such as labour cost and cost of transportation, and secondarily what Weber called agglomerating and deglomerating factors. Weber defines two types of materials used by a manufacturer, first there are Ubiquitous materials, which are materials available everywhere such as water and air, and second there are Localized materials, which are materials particular to a certain location.
He then proceeds to calculate a material index (MI) by dividing the weight of localized materials used in production, by the weight of the product, thus creating a ratio indicating whether a location near the market site should be preferred to a location near the materials site. This is Weber’s analysis of the minimum transport point, and is basically what was stated in section 2.2.1; when choosing its location, a firm will consider the spatial distribution of necessary natural resources according to the relative importance in production. A firm with a high material index would not want to locate at any other point than the raw materials source, as this would involve paying unnecessary transportation costs.
Dicken and Lloyd (1990) illustrates this with the example of heavy industrial processes in the nineteenth and early twentieth centuries. These were dependent on very high volumes of coal, and would accordingly often locate on the coalfields, furthermore, the iron and steel industries would also tend to locate at or near sources of iron ore.
Weber also extended his reasoning to incorporate labour costs, stating that with varying labour cost, the industry would deviate from its previously set location with respect to transportation costs, in proportion to the relative importance of the labour factor. This implies that the influence of an additional factor on the location decision is as high as its relative importance compared to all other factors, or in other words its cost share.
2.2.3 Economies of agglomeration
Aside from previously mentioned factors concerning a firm’s location decision, Weber saw an additional influence from what may be called external economies of scale. By operating in close proximity to other economic activities, or in other words clustering, a firm could benefit from what Weber called economies of agglomeration, meaning the individual firm could, without raising its own production, benefit from the major
clusters internal scale economies. According to Weber, these agglomeration effects would influence locational choice in a way similar to other factors such as cheap labour.
Dicken and Lloyd (1990) points out that Weber’s view of agglomeration economies is somewhat simplistic, adding that unlike cheap labour and materials, agglomeration economies are dependent on the decision of a number of firms. Spatial agglomeration among firms may be driven by desire to reduce uncertainty; a firm’s decision to locate within a cluster can be based on observing other firms in the same industry operating at the particular location, thereby drawing the conclusion that conditions must be satisfactory.
Probing the concept of agglomeration further, a distinction is made between localization economies of agglomeration, and urbanization economies of agglomeration, where localization economies of agglomeration are gained by firms either in the same or closely related industries in close spatial proximity. Urbanization economies of agglomeration are savings from large-scale operations of the agglomeration as a whole, applying to all firms in all industries at a location.
Benefits from spatial clustering includes the possibility of specialized function, meaning that certain services needed within a firm can be outsourced to a specialist within the cluster (i.e. city), creating the possibility for large-scale operations by the specialist on tasks individual firms could not perform as cost-efficient. Operating in close proximity to other firms may also increase the possibility of knowledge spill-over and innovation (Nyström, 2005).
Other advantages of the large agglomeration include reduced uncertainty of supply due to the high level of aggregated demand permitting the operation of specialized factors or merchants. This can allow firms to immobilize less capital in inventories, compared to an isolated firm operating outside the agglomeration.
Large agglomerations can also offer producers better abilities to meet sudden shifts in production activity due to the wide range of labourers, skills and floor space within the agglomeration. Finally, the large agglomeration may allow for large-scale purchasing, for example a number of spatially concentrated small firms could combine shipments of
merchandise, adding up smaller volumes in one shipment allowing for a more cost- efficient means of transportation.
These positive aspects of agglomeration summarized have historically awarded agglomeration effects a key position in studies of location, but as Dicken and Lloyd (1990) points out, although short-distance linkages remain very important in some industries and firms, for others the spatial horizon of interfirm linkages have become the world itself.
2.2.4 Diseconomies of agglomeration
At a certain point, further expansion of an urban agglomeration can result in a loss of efficiency due to for example clogged transportation routes, pollution and administrative overload. This is what Weber called deglomerative tendencies, and may result in the reversal of the spatial concentration of human activity. Another aspect of firms clustering is the possibility of increasing wages and prices on inputfactors when competing for the same resources (Nyström, 2005).
2.3 Choice of location and the investment decision in process industries
As described in previous sections of chapter 2, the locational choice is influenced by spatial variations in factor prices, and to some extent spatial agglomeration. Focusing on the process industry, it is self-explanatory that choice of location for a plant means an investment will occur. Firms may also choose to invest in already existing plants, Bergman and Johansson (2002) categorize process industry-investments into three types; the first being so called greenfield investments, this is the construction of new plants. Firms may also choose to invest in an existing plant by installing additional processing units to increase capacity. Lastly, firms may choose to invest in existing processing units, this can be done by replacing older units or aiming to increase capacity by bottleneck elimination.
Besides the previously outlined theory of location, another useful tool is the Heckscher- Ohlin theory of factor proportions, or as it’s often referred to; the factor endowment theory. According to Lundberg (1995), the main contents of the factor endowment theory are that countries are equipped with productive resources in different proportions. For instance western industrialized nations have access to a qualified
workforce, while this is a scarce resource in developing nations who instead usually has access to a large but uneducated workforce. Furthermore, the factor endowment theory states that industries use production factors in different proportions, meaning that some industries are capital intense and require large investments in for example machinery while other industries are labour intense and requires a large workforce. The Heckscher- Ohlin theorem can be expressed as; a country will specialize in manufacture of products that require input factors the country has in abundance (Lundberg 1995).
As interpreted by Bergman and Johansson (2002) in their study “Large investments in the pulp and paper industry”, the Heckscher-Ohlin theory of factor proportions implies high levels of investments at locations where factors in which the pulp and paper industry is intense are cheap and/or abundant.
Assuming this is true for all factor-intense industries, and adding this to the previously outlined theory on location, a process industry-firm is expected to choose where to locate its investments based on spatial availability and price of production factors in which the firm is intense, and agglomerating factors such as existing capacity and population density.
Chapter 3
PRIMARY ALUMINIUM PRODUCTION AND INPUT COSTS
This chapter begins with an outline of the process chain and addresses the magnitude of the energy use in the production process. The major cost shares of primary aluminium production are then presented and discussed, and the chapter is concluded by addressing the general concern the industry has been expressing regarding rising electricity prices.
3.1 Aluminium production
An outline of the process chain would have to take its starting point at the bauxite-mine.
Bauxite is a naturally occurring mineral, and the basic raw material for primary aluminium production (Schwarz, 2004).
The bauxite is washed, ground and processed into aluminium oxide, more commonly called alumina, which is the raw material input used by primary aluminium smelters.
The alumina is transferred into primary aluminium using the Hall–Héroult process, where the alumina is dissolved under high temperature in an electrolytic bath of molten cryolithe inside a carbon or graphite-lined large container, referred to as a pot (Schwarz, 2004).
Aluminium is obtained by passing a very high amperage (100-320 kA) electric current trough the electrolyte. As the current flows between a positive carbon anode and the pot acting as negative (cathode) with its thick carbon or graphite-lining, molten aluminium is deposited at the pots bottom, and is siphoned off periodically. The aluminium is now ready to be blended into a specific alloy, cleaned and ready for casting (Schwarz, 2004).
Figure 3.1 illustrates the process chain in a flow chart.
Alumina plant
Bauxite Alumina
Cast house
Primary aluminium
Smelting plant Bauxite
mine
Figure 3.1 Flow chart over process chain
3.2 Major cost components of primary aluminium smelting
Table 3.1 presents estimated world averages for the Corporate Operating Costs (COC) by major cost components for the year 2003, as reported by the CRU Group (2004). The CRU Group (2004) defines a smelter’s COC as all of the cash costs associated with primary aluminium production.
Table 3.1 COC shares of major cost components of primary aluminium smelting in 2003
Alumina Labour Power Other materials
Sustaining equipment
Net
realisations
Corporate OH
36 % 11 % 28 % 26 % 4 % -8 % 3 %
Source: CRU Group (2004)
As table 3.1 shows, alumina accounts for the largest cost share by approximately 36 % on average, followed by power costs at an approximated average of 28 %. “Other materials” include carbon anodes, bath materials etc, and together accounts for an average cost share of 26 %. Carbon anodes and bath materials are the main components of the Hall–Héroult process (see 3.1) where the raw material alumina is dissolved and processed into aluminium. The headline “Sustaining equipment” is basically a maintenance cost and is quite small together with corporate overhead costs. Net realisations represents the cost of realising the products market value, this is negative in
table 3.1 and could according to CRU arise from a plants experienced premiums or discounts relative to benchmark prices. The CRU treats various premiums as negative realisation costs to ensure that all units can be considered on a directly comparable basis.
The large cost share for power is particularly interesting. According to the International
electrolysis accounts for 96 % of electricity requirements
ith this in mind, one could assume that when choosing location for a smelter, firms
.3 Regional variations in input prices
en expressing concerns about high electricity- prices. The matter is for example addressed in the 2005 Financial Times Special Report Aluminium Institute (IAI) (2005a) the average smelter requires 15.7 kWh of electricity to produce one kilogram of aluminium from alumina. To put this figure in some perspective, the electricity required to produce one ton of aluminium would be enough to sustain the average Swedish electricity-heated home for nearly nine months (based on SCB:s statistics for 2002). According to IAI (2005b) statistics for 2003, the worlds production of primary aluminium reached a total of 21 935 000 metric tons.
Schwarz (2004) reports that
in primary production, and on average, electricity constitutes one fourth of total operating costs for primary aluminium production. This is in line with CRU’s estimated COC share of 28%.
W
will pay considerable attention to both supply and price of electricity. This assumption is supported by Schwarz (2004), stating that for smelters, although accounting for a lesser part of total operating cost than alumina, electricity price is more important when choosing location for a smelter because of much greater variation in the price compared to that of alumina. The CRU Group (2004) supports this by stating that unlike electricity, alumina is often obtained under similar contracts. Therefore according to the CRU Group (2004), the price of electricity facing a smelter could impact the smelters competitiveness to a greater extent than its alumina cost. Also, according to IAI (2005a), smelters tend to be located in areas with an abundance of energy sources, this gives additional support to the hypothesis that electricity price is a key factor in determining the location of investments in smelters.
3
Recently, the aluminium industry has be
on aluminium. The concerns seem to come from predominantly European and US based producers where long-term power contracts are expiring and the producers are faced with significantly increased electricity costs. As a result of high prices, plants have begun to shut down, no longer being able to compete on the international market, for example Norsk Hydro, ceasing operations at two of their German-based plants, blaming unability to secure favourable power contracts. Figure 3.2 illustrates regional average smelter electricity cost (US$/MWh) for 2003, for the six largest primary capacity- possessing regions in 2003.
0 5 10 15 20 25 30 35 40
China North America
CIS Latin
America
Northern Europé
Asia
US$/MWh
Figure 3.2 Regional average smelter electricity costs, 2003
Source: CRU Group (2004), own calculation
China is in a sense a c erienced an exceptional
growth in its industrial sector during the recent years. CRU data (see figure 3.2) ase of its own. The country has exp
indicates that Chinese primary aluminium smelters face quite large electricity costs, but the expansion in the primary aluminium industry has still been of seldom seen magnitude. One might assume that this expansion is not entirely driven by ideal conditions for primary aluminium smelting, but to some extent driven by public policy and very strong domestic demand. Figure 3.2 indicates that Northern American, Latin American, and Northern European smelters face similar electricity costs, but due to expiring long term power contracts, this picture could change considerably in the near future.
Regional differences for alumina prices are not of the same magnitude, as alumina usually is obtained under similar contracts (CRU Group 2004). Figure 3.3 illustrates regional average alumina costs (US$/Ton) for 2003, for the six largest primary capacity- possessing regions in 2003.
0 50 100 150 200 250 300
China North America
CIS Latin
America
Northern Europé
Asia
US$/Ton
Figure 3.3 Regional average smelter alumina costs, 2003
China again stands out, displaying a sligh ly higher average cost for alumina. Hunt
abour costs also vary across regions and constitutes are considerable cost factor.
ue to rising electricity prices around the world while at the same time many smelters
Source: CRU Group (2004), own calculation
t
(2004) suggests that the Chinese domestic supply of alumina is not yet capable of meeting the strong demand, thus forcing the Chinese smelters to rely heavily on imported alumina to sustain their production.
L
Smelters in low-wage regions tend to be quite labour-intense, but are also characterized by lower labour productivity. Therefore, labour costs per ton of primary aluminium do not vary widely (see for example Blomberg and Jonsson, 2007; King, 2001).
D
face expiring long-term power contracts, firms across the industry are starting to show tendencies towards a re-location to countries able to offer cheaper electricity. The 2005 Financial Times Special Report on aluminium reports Norsk Hydro to be working on a 570 000 tonnes/year smelter in gas-rich Quatar, meanwhile shutting plants in both
Germany and its native Norway. Alcoa, the industry’s biggest producer, is said to consider shutting down a 195 000 tonnes/year smelter in Maryland, USA, due to the electricity price being 40% higher than for the average smelter, while investing 1.1 billion US$ in its first new smelter for two decades in Iceland, reason being Iceland’s abundance in geothermal energy, allowing for cheaper electricity (Ibid). Addressing the matter of relocation, Financial Times (2005) quotes Daniel Brebner, analyst at UBS Warburg:
“There is a migration of capacity from traditional areas of aluminium manufacturing
lso, according to Financial Times (2005), Alcoa, the worlds largest aluminium
istory reveals a similar pattern as a result of the 1970´s oil crisis which lead to that are no longer economic, including some of the first smelters built, to areas with surplus energy, such as Iceland, the Middle East, Russia and Brazil”
A
company, with almost 70 % of its capacity in the US, and another 500 000 tonnes/year capacity in Europe, has adopted the outspoken strategy to build new smelters in parts of the world where energy costs less, and is reported to plan shut-down of older smelters in both North America and Europe.
H
significant increases in electricity prices. According to the CRU Group (2004), this led to the demise of the Japanese primary aluminium production in particular, and increased production cost for smelters in general. The CRU Group (2004) states that this development sparked the need for modern low cost smelters to replace lost capacity, and high cost smelters were widely decommissioned. It is only natural for an industry to seek cost reducing alternatives if possible, and with this historical development in mind and the western-world based smelters concerns about expiring long-term power contracts leading to higher energy prices, questions are raised whether a restructuring is at hand or not. For these reasons, a study on the determinants of location choice of investments in the primary aluminium industry can be of interest, possibly determining the impact electricity price has on location choice.
Chapter 4
MODEL, DATA AND VARIABLES
his chapter initially discusses the determinants of investment in the aluminium
.1 Determinants of investment
on’s (2002) interpretation of the Heckscher-Ohlin
he base raw-material is alumina, processed from bauxite, and therefore the price of
patial agglomeration among firms may attract investments to clusters based on for T
industry and defines the independent variables used to explain the location of investments in this thesis. Section 4.2 presents the fixed effects negative binomial model used to estimate factor price-influence on the location of investments. The data is reviewed and the reason for the thesis regional-approach is explained and some issues concerning validity are addressed.
4
Following Bergman and Johanss
theory of factor proportions, a high level of investments is expected in regions where factors the aluminium industry is intense in are cheap and/or abundant. Abundance in these factors will be measured in factor prices, as this will reflect transportation costs and other barriers to trade among regions (Bergman and Johansson, 2002). It is assumed that the major cost shares reported in section 3.2, table 3.1 reflects these factors.
Specifically, the costs and availability of raw materials, energy, and labour are believed to affect the location choice.
T
alumina (ALP) will be used as an independent variable. Because of the process huge electricity requirements, thus making electricity one of the most important factors, the price of electricity (ELP) will be added as an independent variable. Other necessary factors to be included are the price of labour (LAP), the price of anodes (ANP) and the price of bath materials (BAP). Section 3.2 gives brief information about major cost components and their respective cost shares.
S
example the availability of a trained workforce and specialized subcontractors. For this reason, following Lundmark's (2001) work on the European pulp and paper industry,
existing productive capacity (CAP) in this case measured as number of pots in a region is used as a proxy for agglomeration effects.
Lundmark (2001) and other similar studies (see for example Bergman and Johansson
he expected impact of each explanatory variable is, with regard to economic theory,
Table 4.1 Summation of independent variables
Variable Unit
2002) incorporate variables concerning market conditions for the finished product, such as proximity to the market. This aspect is however not taken into consideration in this thesis, as aluminium is assumed to be traded on a global market thus relaxing the importance of proximity to the customers.
T
quite intuitive. Regional increases in input prices (raw materials, energy and labour) are expected to reduce incentives for investments, while agglomeration within a region is expected to increase a firm’s willingness to invest in that particular region. Table 4.1 summarizes the independent variables which are used.
Definition
ALP Alumina price US$/Ton
ELP Electricity price US$/kWh
LAP Labour price ´000 US ployee
Bat ice
Exi ity Nr on
$/process em
ANP Anode price US$/Ton
BAP h materials pr US$/Ton
CAP sting productive capac of pots/regi
.2 Model specification
variables for smelter-investment location choice implies an 4
The proposed explanatory
underlying simple generic model (or perhaps a better wording is “line of thought“). It is assumed that a generic model which describe the suggested dependencies of input factors and agglomeration effects can be written as equation (1), where the number of investments (ninv) in a region (z) during time period ( t ) is a function of the regional average price (P) of inputs, and agglomeration effects (A).
nregionInv t f(Pinputst,At) (1)
z =
To examine if and how regional average input prices and agglomeration effects impact investment location choice a count data regression analysis is performed, where the
ents for a given region and year are counted and the impact of each explanatory easured.
ized to obtain a measure of the impact of each explanatory
odel, see Hausman et al (1984) for the full version; Let investm
variable is m
The count data regression is performed under maximum likelihood procedures. In other words, the likelihood-function of investments occurring in a given region, during a certain year, is maxim
variable. Two common econometric models for maximum likelihood estimation of count data are the Poisson regression model, and the negative binomial model. They are similar in notation, but an important difference is that the negative binomial model allows for the variance to exceed the mean, or in other words overdispersion in the data, while the Poisson does not. Due to the overdispersed count variable here (nr of investments), the Poisson model would produce underestimated standard errors for the explanatory variables. The negative binomial model is a standard model for overcoming this problem.
Hausman et al (1984) derives the fixed effects negative Binomial model as an extension
of the Poisson m λ denote the
eter, and log
Poisson param λ = Xβ, where the vector X describes the characteristics of a regressor in a given time period. If n denotes the count for unit it i in time period t , the basic Poisson probability specification is:
( ) ( ) .
it n it it
it
n e f n pr
it
itλ
λ
= −
= (2)
! n
To deal with overdispersion, Hausman et al (1984) has shown that by assuming that the son parameter follows a gamma distribution with parameters (y,δ), where
Pois λit
β
xit
e
y= and δ is common across both and , the negative binom
i t ial distribution can
be specified as:
( ) nit ( )it it
it
it e f d
n n
pr =∞∫ −λitλ it λ λ
0 !
1 = ( )
( ) ( ) ( ) it
it
n y
it it
it it
n y
n
y ⎟ + −
⎠
⎜ ⎞
⎝
⎛ + + Γ Γ
+
Γ δ
δ
δ 1
1
1 (3)
Hausman
conditioning out
et al (1984) then proceeds by making the joint probability conditional on
∑n , thereby it δ , and they are now able to specify the joint robability as:
p
(
ni nit∑nit)
pr 1,..., ( )
( ) ( )
⎟⎟
⎜⎜ Γ⎜⎛ + ⎟⎞
⎠
⎝ ∑ ∑
t t
it it
t y n
it it
This is the fixed
⎟⎟
⎟
⎠
⎞
⎜⎜
⎜
⎝
⎛
⎠
⎝
⎟⎠
⎜ ⎞
⎝
⎛ +
Γ
⎟⎠
⎜ ⎞
⎝ Γ⎛
⎟⎟⎞
⎜⎜⎛
= ∑ ∑
∏Γ Γy yΓit+nnit +1 t yit t nit 1 (4)
effects negative binomial model as presented by Hausman et al (1984), which allows for both overdispersion and individual variance to mean ratios. The model
pplied to estimate the impact of each proposed expla
probability of investments in a given region. The resulting estimated coefficients for the explanatory variables can be interpreted as either positive or negative indicators2 of how
sumption and capacity. As it turned out to be difficult to obtain included in the database, i.e. with no current active smelter instead of a more precise, country-level
ber of pots in a region serves as a measure of the installed base of capital, and it could be of interest to compare this against the regional average electricity price.
is a natory variable on the
a change in the value of a given explanatory variable for a region affects the probability of smelter-investments occurring in that particular region.
4.3 Smelter data and validity issues
Smelter data for the period of 1990-2003 was gathered from a proprietary database acquired from the CRU Group (2004) covering the strong majority of active smelters worldwide. The database covers information about smelter production, production costs, input factor-con
certain data for countries not
capacity, a regional approach will be taken study.
The data constitutes an unbalanced panel of the world divided into 12 regions. For convenience, the same regions as defined by the proprietary database obtained from the CRU Group (2004) will be used, these regions are presented in table 4.2 along with the number of pots and regional average electricity prices for the first and last years of the
e period. The num tim
2 It should be noted that no marginal effects have been calculated. This issue means that no precise measurement of the impact of an explanatory variable can be reported in this thesis.
Table 4.2 Regions as defined by the CRU Group (2004)
Number of pots Electricity price (US$/kWh) Region
1990 2003 1990 2003 Africa 1892 2928 0.024 0.013 Asia 6316 4206 0.043 0.031 Central Europe 3991 2448 0.859 1.545 CIS 14124 11800 0.010 0.013
China 3739 19879 0.038 0.036 Middle East 2137 3144 0.027 0.026
North America 18945 15399 0.021 0.019 Northern Europe 4258 4502 0.022 0.019 Oceania 3209 3908 0.021 0.016 Latin America 5949 6091 0.023 0.020 Southern Europe 3536 2964 0.033 0.024 Eastern Europe 2789 2458 0.041 0.027
Source: roup (
At fir simple ariso ears d some partial support for the suggested link between variations in ca y and icity prices n example, the regional average electricity in Afr creas approxima % % between 1990 and 2003, while the installed base of capital increased by nearly 155 % during the same er, as d ed in on 4.1 y other fac sides the price of electricity must be considere parison between the first and
st years of the time period does not hold any merit in the analysis, but is intended to
l number of smelters in the region to obtain a regional eighted average. Lastly, these variables are arranged as a set of panel data where each
CRU G 2004)
st glance, this comp n app to len
pacit electr . As a
price ica de ed by tely 54
period. Howev escrib secti , man tors be
d. This very simple com la
serve as a basis for discussion.
The data covers the period of 1990-2003. Because of the regional approach a weighted average of each factor price is calculated for each year and region. Real factor prices are calculated for each smelter and year by multiplying factor costs, measured in US$/ton aluminium, by smelter production (ton aluminium), and dividing this by used quantity of the particular factor. These factor prices are then summarized for each year and region, and divided by the tota
w
region is given a code, which is a dummy-variable to separate the regions in the econometric analysis.
The set of panel data is completed by adding the number of investments for each year and region, and the variable CAP used to approximate agglomeration effects. In this thesis, investments are viewed as discrete events. This means that investments are regarded as countable events which can either occur, or not occur. To obtain the number of investments, changes in smelter production capacity compared to the previous year was calculated for each year and region. Typical empirical studies of similar industries ften define investments as an increase in plant capacity by 5 percent or more (Bergman
des China’s aluminium industry, nd another which disregards the Chinese aluminium industry.
o
and Johansson, 2002). Following this definition, every capacity increase exceeding 5 percent was registered as an investment having occurred, and the investment frequencies were registered for each year and region.
It should be noted that China is in a sense a case of its own. The country has experienced an exceptional growth in its industrial sector during the recent years. CRU data (see page 15, figure 3.2) indicates that Chinese primary aluminium smelters face quite large electricity costs, but the expansion in the primary aluminium industry has still been of seldom seen magnitude. For this reason, estimation procedures will be performed using two different setups; one that inclu
a
Lastly, the variable “CAP” used to approximate agglomeration effects was calculated by summarizing the number of smelting pots for each year and region to serve as a measure of the installed base of capital. Descriptive statistics for the used variables are presented in table 4.3
Table 4.3 Variables an descriptive statistics
Variable Description Mean S.D Min Max
d
INV Nr of investments/region 1.52 2.70 0.00 23.00
ALP Alumina price (US$/Ton) 25.53 53.74 163.89 477.38 ELP Electricity price (US$/kWh) .11 0.29 0.01 1.55 LAP Labour price (´000
US$/process employee)
6.15 20.05 1.19 89.41
ANP Anode p .66 375.46
Bath ma 855.97 184.29 374.72 1326.64
y 8 0
2 0 2
rice (US$/Ton) 251.88 49.84 97
BAP terials price
(US$/Ton)
CAP Existing productive capacit (Nr. of pots/region)
5600.55 4956.1 1744.00 20946.0
It is anticip difficulties in
ing co varia a ricit s, h he
y of ysis. This c ve ea o
comparison e as to why a specific country was chosen before another for
ield i i th n g
effects could be distorted because of the sheer size of some regions. Furthermore, input rices are likely not entirely homogenous over an entire region such as for example
nal averages calculated from the available data will enable to test for an xplanation of why investments take place in some geographic regions before others.
ated that the regional approach, which was necessary due to
obtain untry-level data for key bles such s elect y price ampers t validit the econometric anal
can be mad
lack of ountry-le l data m ns that n
greenf nvestments. It is also recogn zed that e importa ce of ag lomeration
p
Africa. Smelters may operate under significantly different conditions even within a region, due to for example some smelters having access to their own low-cost hydro power.
However, the regional approach is still maintained to be useful in attempting to identify the existence of trends in smelter-investments. There are important differences between the specified regions, such as the availability of raw materials and energy which is believed to affect prices, and differing wage rates between for example countries in Northern Europe and Eastern Europe. This is the rationale behind the regional approach, as regio
e
