Porter’s Five Forces and Their Effect on Copper Price Exposure of Companies in the Copper Industry

(1)

MSc International Business and Management MASTER THESIS

Porter’s Five Forces and Their Effect on Copper Price

Exposure of Companies in the Copper Industry

Gergana Todorova VALKOVA

Student number: 1834398

G.Valkova @student.rug.nl

Supervisor: Dr. Nanne Brunia

Co-accessor: Dr. Wim Westerman

(2)

Porter’s Five Forces and Their Effect on Copper Price

Exposure of Companies in the Copper Industry

Abstract

(3)

I. Introduction

In June 2009 the Fitch rating agency published a report about the impact of the credit crunch on metal mining companies. Between April and June 2009 the companies Rio Tinto, Teck and Anglo American which are active not only in mining but in the subsequent copper handling processes have raised nearly 22 billion USD at the capital markets. (Metals and Mining Latin American Special Report, 2009) The high degree of indebtedness of the copper industry companies brings concerns about the volatility of their cash flows and how they can ensure the due service of their financial obligations. As copper is a major component of their costs and revenues, managing the exposure of their cash flows to copper price changes is crucial in ensuring the necessary level of liquidity. Moreover in the last decade the trend of the copper prices at the London Metal Exchange (the price of refined copper with 99% copper content) changed dramatically. From the steady level of 1500 up to 2000 USD per ton between 1998 and 2004, the cash seller and settlement price of copper started a dramatic increase to reach 9000 USD per ton in June 2006 and May 2008. Later on the financial crisis managed very quickly to bring copper prices to back to the 3000 USD per ton level at the end of 2008. (Figure 1)

Figure 1. Daily copper seller and settlement spot price at the London Metal Exchange (January 1998–June 2009)

(5)

The commodity price exposure is the elasticity of company value to the changes in commodity prices. The academic research on the determinants of commodity price exposures focuses predominantly the effect of financial hedging, financing and production costs, amount of reserves and tax rates. (Tufano, 1998; Haushalter, Heron and Lie, 2001; Jin and Jorion, 2006; Hong and Sarkar, 2008) On the other hand, the characteristics of market competition also deeply influence commodity price exposures. The market competition determines the profitability that the companies can achieve in competing with their rivals and the prices at which they sell and purchase commodity products. (Bartram, 2005) The relation between competition and exposure has already been empirically confirmed in the studies on exchange rate exposures by Williamson (2001), Bodnar, Dumas and Marston (2003) and Dekle (2006). In comparison, in the literature on commodity price exposures only the study by Bartram (2005) mentions the effect of market competition but only as an explanation of the share of insignificant exposures to commodity price in the sample.

The five forces model by Michael Porter is by far the most well-known instrument to describe market competition. It includes the influence of the bargaining power of buyers and suppliers, the rivalry with existing and new competitors and the threat from substitute products. (Porter, 1980, p.4) Having in mind the major importance of copper price exposure to companies in the copper industry, I formulate the research question of my master thesis as follows:

What is the effect of Porter’s five forces on the copper price exposure of companies in the copper industry?

(6)

MINING SMELTING REFINING

FABRICATING

OTHER PLANTS WIRE ROD PLANT

BRASS MILL

FOUNDRY MINING AND PRODUCTION

Figure 2. Primary metal industries with copper input and/ or output according to the classification of the International Copper Study Group

Source: World Copper Factbook 2007 by The International Copper Study Group, www.icsg.org

In order to give an answer to the research question, I apply a two-level regression model to determine the magnitude of copper exposure and then test the relation between the estimated copper exposure and its competition related determinants. I use a pooled dataset of annual exposures of 30 listed companies form the copper industry between 2000 and 2005. The sample is highly international including firms from 17 different countries.

The empirical tests in the research will provide managers with information about the magnitude of copper price exposure and to which copper markets they should consider expansion in order to lower this exposure. The more general managerial implications relate to minimizing the volatility of cash flows by managing the exposure to copper price changes. The limited volatility of cash flows will enable managers to plan the necessary liquidity for meeting financial obligations’ deadlines and enjoy the positive reaction of investors when external capital is needed.

(7)

II. Literature review

In order to answer the research question with the help of existing literature, I consider two clusters of studies - on commodity price exposures and exchange rate exposures. The existing researches on commodity price exposure predominantly focus on national samples of gold mining and oil and gas extracting companies. (Tufano, 1998; Haushalter, 2000; Faff and Hillier, 2004; Jin and Jorion, 2006; Hong and Sarkar, 2008) When considering the relevance of these studies to other industries, I turn to the results of the study by Faff and Hillier (2004). They attempt to apply the model by Tufano (1998) about the determinants of gold price exposures to other commodity extraction and production companies. The successful generalizability of the model is attributed to the similar operating and market risks for the companies involved in commodity extraction and production. Consequently, the assumption about the similarity of risks makes the conclusions of the specific industries studies applicable to companies from the copper industry.

The studies on exchange rate exposure form another useful body of knowledge. The researchers on commodity price exposure often estimate the exposure magnitude using the market model by Jorion (1990) for the exchange rate exposure of US multinationals. This model is an easily applicable two-factor regression and requires publicly available data about returns on company stocks. Moreover the researches on exchange rate exposure offer detailed discussion on various aspects of exposure estimation and exposure determinants. (Bodnar and Wong, 2003; Dominguez and Tesar, 2006; Bartram and Bodnar, 2007)

2.1. Copper exposure – definition, sign and magnitude

Bartram (2005) defines commodity price exposure as the effect of unexpected changes of commodity prices on the value of the firm. Consequently copper price exposure is the elasticity of the value of the firm to the unexpected changes of the copper prices. Companies form the copper industry determine the price of mined, smelted, refined and fabricated copper on the basis of the average copper quotation at a chosen commodity exchange for a past period of time and adjustment for the processing charges. (Metal Bulletin Monthly, 2007) The price of copper at the commodity exchanges refers to refined metal with 99% copper content.1 Therefore the companies in the value chain before copper refining sell the metal at the price

1

(8)

formed at the commodity exchange with a reduction for the oncoming processing. Similarly, the fabricating companies offer copper products at the commodity exchange quotation with an addition for the fabricating charges.

Bartram (2005) analyzes the problem about the sign of commodity price exposure. Based on a mutli-industry sample, he concludes that commodity price exposure tends to be negative for companies with the commodity as an input factor and positive when the commodity is an output product. The net result of both effects is not addressed by Bartram (2005) but receives attention by Petersen and Thiagarajan (2000). The latter study focuses on two gold mining companies and examines the elasticity of their sales revenues, operating costs and cash flows to the change of gold prices. When the elasticity of the costs and revenues with respect to the commodity price are almost equal, the commodity price exposure of the cash flow is eliminated.

To determine the effect of the copper price changes on the value of the company, I follow a line of reasoning similar to Bodnar and Marston (2002) about exchange rate exposure. The economic meaning of the copper price exposure is a coefficient that shows with what percentage the value of the company changes for 1% change of copper prices. The value of the company is estimated as the present value of company cash flows. On the other hand, the cash flow is equal to the after-tax profit of the company. The profit of the company is simply the difference between total revenues and total costs, which is corrected with the corporate tax rate. I assume that all costs and revenues except the ones for copper input and output remain constant for a period t. The amount with which the company profit changes after a change in copper prices during the period t can be estimated in two ways:

• as the product of the company profit and coefficient representing copper price

exposure;

• as the difference between the amount of revenues from copper output and the

amount of costs for purchasing copper input. I combine the two approaches in the following equation:

1 2

(1 T h TR)( h TC)

βπ

= − − (1)

(9)

h1 – the ratio of revenues from copper output to total revenues, h2 – the ratio of costs for purchasing copper input to total costs.

If π1 is the pre-tax profit of the company, r – the pre-tax profitability of the company, then

using the relations

π

=(1−T)

π

₁ ,

π

₁ =TR TC− and r =

π

₁ TR to rewrite equation (1), the result is:

β

=h1+(h1−h2)(1 r−1) (2)

Equation (2) leads to the conclusion that copper price exposure depends on the ratio of revenues from copper output to total revenues, the balance between the costs for copper input and the revenues from copper output and the company profitability. If the shares of copper costs and revenues in total costs and revenues remain unchanged, higher profitability leads to lower exposure. If a company has a positive profitability, a higher share of copper input costs in total costs (h2) can drive the copper price exposure (β) to a negative value whereas a higher share of copper output revenues in total revenues (h1) guarantees positive exposure. If a company in the copper industry has mining activities, its costs for copper purchasing are zero. Therefore the sign of its copper price exposure becomes positive. The empirical researches confirm the conclusion about companies with mining activities and demonstrate predominantly positive commodity price exposures for companies in the mining and extraction industries. Tufano (1998) examines a sample of 48 North American gold miners and reports positive gold price exposures. Jin and Jorion (2006) find again only positive exposures of the 38 oil and gas companies they research.

(10)

extensive study on commodity exposures suggests a similar situation reporting that the commodity price exposures are not more significant as a financial risk factor than the exchange rate and the interest rate exposures.

2.2. Strategies to manage copper price exposure

The managers of companies from the copper industry can take different actions regarding copper price exposures. First, they have to make a decision whether to accept the exposure without attempting to change it. The acceptance is a possible approach when the competition on the input or output market is very tight and the company needs to maintain its market share. The managers decide to lower the profit margin of the company by accepting higher prices for purchasing copper input or lower prices for selling copper output. In this case the quantities of purchased or sold copper remain unchanged. This strategy is referred to as the

price change absorption. (Hoogenboom, 2006)

According to the production flexibility strategy (Tufano, 1998) the companies from the mining industry can manage exposure by operating only in times of high enough commodity prices. When the commodity prices fall below the level that covers the fixed costs of the firm, the company stops operating. Tufano (1998) describes this strategy in relation to gold mining companies but Geman (2005, p.171) claims that production flexibility is not so common among metal miners and producers.

When Bartram (2005) tries to explain the low share of companies with significant commodity exposures, he offers the argument about the price pass-through strategy. To outline the effect of the strategy I use the study by Bodnar, Dumas and Marston (2002) on exchange rate pass-through. I apply it not only to in the direction to customers but also to suppliers of the sample companies. When the copper quotations at the commodity exchange rise, the companies in the copper industry have the incentive to negotiate higher prices with their customers. When the copper quotations at the commodity exchange fall, the same companies turn to their suppliers and attempt to agree on lower prices. This price renegotiating behavior is attempted by all the copper companies along the added value chain. The companies try to keep their profit margin unchanged and in this way to minimize the volatility of their cash flow.

(11)

companies and find that only 35% of the researched 241 mining companies use commodity forwards, swaps or options in 2000-2001. The use of commodity derivatives is more typical for companies in the oil and gold mining industries as these markets have experienced long history of price turmoil. In comparison, copper hardly reached levels above 1000 USD per ton in the period 1970-2004 while the 9000 USD level was achieved for the first time in 2006. (ICSG World Copper Factbook, 2007, p.63) Some companies in the researched sample like the North American Newmont even eliminate their portfolio of financial derivatives and claim that this step has led to a better company performance. (Annual report 2007, p.1)

2.3. Michael Porter’s five forces as determinants of the copper price exposure – hypotheses development

The five forces model by Michael Porter is a well-known business concept which is applied in studying industry competition. (Figure 3) The combination of the bargaining power of buyers and suppliers, the rivalry with existing competitors and the threat of new entrants and of substitute products are specific to the industry and determine the intensity of competition. (Porter, 1980, p.6)

Figure 3. Forces driving industry competition

(12)

The industry competition affects the copper price by affecting profitability and the use of price pass-through strategy. The intensity of competition deeply influences the company profitability with the strongest force/combination of forces determining the profit margin. (Porter, 2008) According to Equation (2) the company profitability has a negative effect on copper price exposure as the higher the profitability of a company leads to a smaller exposure. The relation between exposure and competition is empirically proven by Williamson (2001) in a research of exchange rate exposure of Japanese and US automobile companies. The results of the study confirm that the exposure changes over time follow the changes of the market share of rival companies.

The industry competition determines the ability of the company to pass price changes to its customers and suppliers. By conducting price pass-through the companies from the copper industry manage the amount of their copper costs and revenues. The companies with a strong ability to renegotiate the pricing terms with their customers and suppliers have a less volatile profit margin and smaller exposure. The negative relation between price pass-through and exposure is empirically confirmed in the study by Bodnar, Dumas and Marston (2002) on exchange rate exposure of companies from eight Japanese industries.

2.3.1. Bargaining power of suppliers and customers

In his model about industry competition Porter (1980, p24-29) lists the situations in which the suppliers and buyers have high bargaining power. They refer to the concentration of companies on the market, the presence of substitute products, the product homogeneity, the vertical integration, the transaction costs for changing vendors, the composition of sales revenues and operating costs, the product impact on the quality of buyers’ output, etc. In comparison, Crook and Combs (2006) offer a more general definition of bargaining power - the ability of a company to coerce other companies to accept exchange conditions they would otherwise refuse to. This ability is the necessary factor for successful pass-through of price changes and ensures that the companies “squeeze” a larger share of industry profitability. (Porter, 2008)

(13)

power. Fewer companies on the supply side than on the demand side, differentiated products, the lack of product substitutes, high switching costs for changing vendors, the threat of forward integration to the buyer business, high share in buyers’ costs for the product and high product impact on the quality of buyers’ output determine a high bargaining power for the suppliers. (Porter, 2008) As determining the switching costs requires knowledge about the exact contract terms between companies, I will focus on the remaining conditions for bargaining power to analyze the mining, smelting, refining and fabricating markets.

If we have a look at the figure with copper products in the different markets, we can conclude that their variety rises from the mining to the fabricating market and onwards. (Figure 4)

Figure 4. Copper containing products along the added value chain

Source:International Copper Study Group – The World Copper Factbook 2007 http://www.icsg.org/ images/stories/pdfs/2007worldcopperfactbook.pdf, p.7

Taking Figure 4 into consideration, I use the following supply and demand characteristics of the fabricating market to determine whose bargaining power dominates:

• Based on the large number of industries that are customers of the copper

(14)

• The companies on the fabricating demand side enjoy the high standardization of

fabricated copper products as the fabricating output complies with technical standards.

• The customers at the fabricating market produce more technologically complex

products with a higher profit margin. (Agostini, 2006; ICSG World Copper Factbook, 2007, p.7) The more complicated production technology used by the customers of fabricated copper products acts a barrier to forward integration for copper fabricating companies.

• The copper component in the products manufactured by the buyers of fabricated

copper is less relatively important for the quality of the final product as the other components also have an influence.

• The fabricated copper products can be more easily substituted for the same

industrial applications with products from other substances or even with entirely different technologies. (Stuckey and White, 1993; Geman, 2005)

On the whole the conditions that lead to buyers’ bargaining power tend to outnumber the ones that stimulate suppliers’ bargaining power on the fabricating market. I assume that all conditions have equal weights in determining the type of bargaining power that dominates on the market. Therefore I consider the fabricating market as influenced by the bargaining power of buyers. The power of the buyers creates an obstacle for suppliers to pass copper price changes and to increase their profit. Therefore the first research hypotheses states that

Hypothesis 1. The stronger the companies in the copper industry experience the bargaining power of the buyers on the copper fabricating market, the larger copper price exposures they have.

As far as the other end of the value added chain is concerned, the mining market has the following characteristics:

• The high relative importance to the quality of buyers’ products and the low

substitutability of the copper ore are two of the conditions that create suppliers’ bargaining power on the mining market.

• The copper share in costs for companies on the demand side and the share in

(15)

• The companies supplying mined copper build their power mostly on the entry

barriers for new companies which deter backward integration and new entrants. The access to the mining market is not determined by market mechanisms but by the policy of the host governments about control over natural resources. (Shapiro, Russell and Pitt, 2007) The government controls the number of companies with mining activities and contributes to the higher concentration of mining supply among fewer firms as compared to the smelting and refining markets.

Based on the characteristics of the mining market I consider it a market dominated by the bargaining power of the suppliers. The bargaining power of the suppliers helps the companies in copper mining to pass through price changes and to increase their profitability. The second research hypothesis claims that

Hypothesis 2. The stronger the companies in the copper industry experience the bargaining power of the suppliers on the copper mining market, the smaller copper price exposures they have.

2.3.2. Vertical integration and the threat from existing and new rivals

Porter (2008) pays special attention to the type of rivalry between companies in the industry. The rivalry may be based on prices or on different product features and additional services. The rivalry among companies in resource producing/extracting industries is mostly price competition. (Porter, 2008) The products in these industries are standardized according to a set of technical requirements and the product differentiation is low. Besides the fixed costs in resource producing/ extracting industries are high relative to marginal ones and the companies are motivated to increase the number of their customers even though selling at prices lower than the average on the market. (Porter, 2008) When the price competition is intensive, the most successful strategy to maximize profitability according to Porter (1985, p.13) is cost leadership. The strategy of cost leadership involves production at the lowest costs and selling at the average prices on the market.

(16)

integration is an asset in coping not only with existing rivals but it also provides a barrier against market entry of new companies. The costs’ optimization by vertical integration is a result of lowering the transaction costs for selling and purchasing, information processing and legal expenses. (Stuckey and White, 1993) In case the vertical integration is effective, the vertically integrated activities help companies to achieve economies of scope by optimizing the costs for transportation and handling of the input products for each production process. (Balakrishnan and Wernerfelt, 1986) This reasoning follows the transaction cost approach in taking management decisions about vertical integration and is widely used in existing literature (Mahoney, 1992; Argyres, 1996; Broedner, Kinkel and Lay, 2009)

On the whole, the higher extent of vertical integration contributes to maximizing company profitability and therefore lowers the exposure to copper prices. Following this line of reasoning, the third research hypothesis states that:

Hypothesis 3. The companies in the copper industry with higher extent of vertical integration have lower copper price exposures.

2.3.3. The threat from substitute products and having substitutes on the product portfolio

According to Porter (1985, p.23) a substitute product is one that performs the same or a similar function to the industry product but by a different means. When the customers can easily switch to the substitute product and in this way get better price-performance trade-off, the profitability in the industry diminishes. (Porter, 2008)

(17)

copper price changes as the threat of substitute products limits the price of copper output. The lower profitability and the less plausible price pass-through imply higher exposure to copper prices and the fourth research hypothesis states that

Hypothesis 4. Offering copper substitutes leads to higher copper price exposures for the companies in the copper industry.

2.3.4. Other influencing factors (control variables)

In the section about the control variables I would like to consider other factors on company level that influence copper exposure. These factors diminish the exposure by intensifying the use of strategies other than price pass through.

In existing literature researchers consider the effect of company indebtedness on commodity exposure in two directions. Tufano (1998) studies the influence of financial leverage (measured as the ratio of the book value of debt and market value of equity) on gold exposure under the condition of flexible production volume. On a sample of 48 North American gold miners he finds significant and positive relation between the two variables. The indebtedness of companies increases their fixed costs and therefore diminishes the share of copper costs in total costs. According to Equation (2) the lower share of copper costs in total costs leads to a higher copper exposure.

On the other hand, Haushalter (2000) reports a positive relation between financial leverage (measured as the ratio of total debt to total assets) and the fraction of financially hedged production for the sample of 100 US oil and gas companies. This result is confirmed in a later study on price uncertainty and corporate value on a sample of oil producing companies by Haushalter, Heron and Lie (2001). Companies that experience high financial distress or underinvestment (lack of sufficient cash flows to invest) have stronger incentives to get involved into exposure management activities and use them efficiently. (Haushalter, Heron and Lie, 2001) Following this line of reasoning, the companies in the copper industry with a high degree of indebtedness tend to have lower copper exposures.

Another company-level factor that influences the intensity of financial hedging is the

company size. The studies on exchange rate exposure point out that the hedging with financial

(18)

(19)

III. Methodology

3.1. Model description and justification

To test the set of research hypotheses, I apply the two level regression model used by Dominguez and Tesar (2006) in their study on exchange rate exposures. The first level regression (the market model) is a two-factor OLS one and determines the magnitude of the copper price exposure as follows:

(3);

where Ri,t is the total daily return on company i stocks for the period t (adjusted for dividend payment and stock splits), RM,t is the total daily return on the market portfolio for the period t

(again adjusted for dividend payment and stock splits), ∆ct – the unexpected daily change in the copper price for the period t, β2,i – the estimate of copper price risk exposure of company i

The second-level regression tests the effect of Porter’s five forces and the control variables on the copper price exposure as follows:

(4);

where β_2,i is the absolute value of the copper price exposure of company i and Xi represents

the exposure determinants - Porter’s five forces and the control variables as estimated for the company i.

For the second level regression I use a pooled data base (both time-series and cross-sectional) which includes the estimated copper price exposures for a period of one year and the annual values of the independent and control variables for the period between 2000 and 2005.

Since the copper price exposure is defined as the elasticity of firm value to copper prices, using a regression model to estimate it seems most logical. (Stulz and Williamson, 1997) Besides this model requires publicly available company data and is applied in previous commodity exposure studies, which allows for comparability of the results. An alternative approach to estimate commodity price exposures is to define it as the elasticity of cash flows to the changes in commodity prices. (Stulz and Williamson, 1997; Petersen and Thiagarajan,

, 0, 1, , 2, ,

i t i i M t i t i t

R =β +β R +β ∆ +c ε

2,i 0 1Xi i

(20)

2000) The cash flow model is more suitable for small study samples and case study research as it requires detailed data about costs and revenues. Bartram (2007) conducts a comparison between the results using the market model and the cash flow model for the exchange rate exposure of a sample of 6917 USA companies. In the short run both models render relatively similar outcome but over longer time horizons cash flow exposures significantly diverge from equity exposures. Bartram (2007) concludes that the estimates based on accounting information are less successful proxies of exposure than the market data as they fail to reflect the company value estimation by the market.

The inclusion of the market portfolio returns in the first level regression is suggested by Jorion (1990) in the study of the exchange rate exposure of US multinational companies. Therefore the estimated copper exposures are residual and their value is the difference between the total company exposure and the exposure of the market portfolio. Then a zero exposure doesn’t mean that the company is unexposed to the changes of copper prices but that it is exposed at the same level as the market portfolio.

According to Bodnar and Wong (2003) the changes of company value are exposed to two effects – the impact of exchange rate changes and the impact of other macroeconomic factors. Therefore following the same line of reasoning when estimating the copper price exposures of companies in the copper industry, the effect of the remaining influencing factors on company, industry and country level need to be isolated. As Stulz and Williamson (1997) state, the market factors like commodity prices, exchange and interest rates tend to be correlated and incorporating them all in one regression is burdened with multicollinearity. The market portfolio acts as a variable that controls for the effect of all the other variables relevant to the company value.

Regarding the exposure determinants analysis, the application of a linear regression model leads to the assumption that the relations between the variables are linear. Another approach that allows testing the assumption about linearity of relations is the use of t-tests for quartiles of the sample (Haushalter, Heron and Lie, 2002; Bartram, 2007) although it can render biased results for a small sample of companies with diverse activities.

(21)

collinearity between the independent variables. (Dougherty, 2007, p.408-409) Pooled databases are used in several studies about commodity price exposures, as well. Tufano (1998) makes use of a data panel with quarterly company observations and quarterly gold exposures estimated with daily frequency. Haushalter, Heron and Lie (2002) employ a pooled database with annual estimations of oil exposures on the basis of daily data and annual data about company-level factors. Jin and Jorion (2006) examine oil exposure factors in a similar way but with monthly data frequency to determine oil betas.

Tufano (1998) and Haushalter, Heron and Lie (2002) suggest that the use of daily frequency

data to estimate exposures suffers the bias of infrequently traded stocks. It can be eliminated

by different adjustment techniques (Dimson, 1979; Fowler and Rorke, 1983) but the easiest way is to have a sample of large-size companies whose stocks are frequently traded on financial markets and demonstrate consistent price trends. The longer return horizon, e.g. using monthly stock returns and price changes, results in higher explanatory power of the regression (higher R square) as found by Bodnar and Wong (2003) about exchange rate exposures. But when the exchange rate exposures are estimated for short periods of time (like one year), the monthly frequency of data renders limited number of observations and biased results. (De Jong, Ligterink and Macrae, 2006)

3.2. Operationalization of the variables in the first and second level regressions

The operationalization of the variables in the first level regression is presented in Table 1.

Table 1. Variable operationalization of the first level regression

Variables Measurement

1 Dependent company stock return total daily return on company stocks reported in

Datastream*

2 Independent unexpected changes of

the copper price

return from overnight holding of one ton copper based on London Metal Exchange spot prices

3 Control market stock return daily return on the Morgan Stanley Capital

International World Investible portfolio equity index

*The company stock returns, the market stock returns and the unexpected changes of the copper price are estimated on the basis of data in US dollars. The data which is not originally reported in USD in Datastream is recalculated using the exchange rates by Global Trade Information Service.

To calculate the copper price exposures, I opt for the international market return index (world investible) provided by Morgan Stanley Capital International (MSCI). A global index

, 0, 1, , 2, ,

(22)

is used by Faff and Hillier (2004) to examine oil exposures of five international industry portfolios. Since the research sample is international, the use of market return indices on country level will make comparability of company stock returns more difficult. The market return indices in the developing countries are inclined to suffer inefficiency as well as the bias of under/overvaluation. Turning to the exchange exposure literature, I find that Chue and Cook (2008) also use a global index by MSCI to estimate the exchange rate exposure of 931 emerging market companies. On the other hand, Dominguez and Tesar (2006) find that the use of international indices of market returns leads to lower explanatory power of the regression model (lower R square). Therefore I rely on a robustness check with different indices of market returns to determine the applicable index.

Table 2. Comparison of copper exposure results estimated with different Morgan Stanley

Capital International market indices for the period 2000 to 2005 and the regression model

Nr of companies

Data

frequency Market index Time period

Nr (share) of companies with significant β2,i at 10% level Nr (share) of companies with significant β2,i at 5% level

β2,i mean R2 mean

30 daily MSCI investible

world 2000-2005 26 (86%) 25 (83%) 0.183 0.095

country 2000-2005 17 (57%) 16 (53%) 0.125 0.209

sector world 2000-2005 15 (50%) 15 (50%) 0.107 0.129

The robustness check includes estimating the coefficients of regression (3) using Morgan Stanley Capital International indices for a global market, national market and world mining industry portfolio. For the whole period between 2000 and 2005 the global MSCI index really results in lower R2 compared to world industry and national MSCI indices. At the same time the MSCI investible world ensures the highest share of significantly exposed companies as well as the highest average exposure magnitude. (Table 2)

Returning to the definition of copper price risk exposure, I need to underline that it is explained as the effect of unexpected commodity price changes on firm value. The relevant question here is whether the spot prices at the commodity exchanges can be used as a proxy for unexpected price changes. Pindyck (2001) claims that a part of commodity price fluctuations is predictable on the basis of seasonal changes pattern but this is more typical for

, 0, 1, , 2, ,

(23)

commodities with a cyclical production like agricultural ones. Geman (2005, p.23) considers the role of futures prices and information about stock exchanges’ inventories for spot price forecasting. The futures prices and the inventory data give information about the current and future balance of the demand and supply on the market and therefore influence the company decisions about purchasing, selling and storing. On the other hand, on the basis of empirical research French (1986) concludes that futures price have very little predicting power as far as copper market is concerned. The predicting power of futures prices even declines in times of low inventory levels (Fama and French, 1988). Single events like trade unions’ strikes in the largest copper mines in South America can have an immediate effect on the copper market by a copper output decrease. (Kelly and Krishnan, 2007) Therefore, the volatility of copper spot prices can be regarded as predominantly unanticipated by the companies in the copper industry.

The operationalization of the variables in the second level regression is set out in the following table:

Table 3. Variable operationalization of the second-level regression

Variables Measurement

1 Dependent Copper price exposure Beta regression coefficient of the effect of changes in the

copper prices on the company stock returns

Interaction dummy mining – market share

at output market

The product of :

1) dummy mining - a dummy variable taking the value of 1 if the output market is the mining one, otherwise 0 and 2) market share at the output market - the annual share in world production capacity at the output market

Bargaining power

Interaction dummy fabricating – market share at output market

The product of :

1) dummy fabricating - a dummy variable taking the value of 1 if the output market is the fabricating one, otherwise 0 and

2) market share at the output market - the annual share in world production capacity at the output market

Mismatch of capacity

Natural logarithm of the difference between the maximum and minimum annual production capacity for the production processes in which the company is involved

Dummy 2 processes Dummy variable taking the value of 1 if the copper output is

handled through two production processes; otherwise 0

handled through three production processes; otherwise 0 Extent of

vertical integration

handled through four production processes; otherwise 0

2 Independent

Substitutes in the product portfolio Annual ratio of aluminum to copper revenues

Company indebtedness The annual ratio of the book value of long-term debt to the

sum of book value of debt and book value of equity

3 Control

Company size Natural logarithm of the total assets of the company in USD

on annual basis

2,i 0 1Xi i

(24)

The proxy the effect of bargaining power employed in the current research is based on the resource dependence theory of bargaining power (Pfeffer and Salancik, 1978). The share of controlled resources in the specific output market measures the bargaining power of the companies in the copper industry. This measurement is also based on the observation that production capacity and supply are almost equal in the copper industry.2 The output market is the one related to the production process (mining, smelting refining or fabricating) which is the final one for the added value chain of the companies. For the companies with mining as an output market the larger capacity in mining involves higher suppliers’ bargaining power. For the companies with an output market in fabricating the larger capacity in fabricating means that they experience a greater effect of buyer’s bargaining power. The focus on the output market is determined by the assumption that all the manufactured products are subject to bargaining with customers who are independent of the supplying company.

Regarding the measurement of bargaining power, Cachon (2003) mentions that companies’ bargaining power is a concept that is easy to understand, but difficult to quantify. Cool and Henderson (1998) develop a complex supplier and customer bargaining power estimation on the basis of Likert-scale questions about number of companies, costs for switching vendors, vertical integration, product impact on costs and revenues, product differentiation, etc. This measurement encompasses all the conditions for bargaining power according to Porter’s model (1980). The complex estimation method by Cool and Henderson (1998) requires a well developed questionnaire in which the concepts in the questions are unambiguously understood by all managers.

The estimation of extent of vertical integration is a combination of a metric and a dummy variable. The lower capacity mismatch means that a greater part of the company’s products is processed through several production processes. The dummy variables give additional information about the number of the production processes included in the added value chain of the company. The more processes the companies’ production involves and the less the capacity mismatch among them, the higher extent of vertical integration the company enjoys. The measurement approach of the vertical integration is an extension of the estimation by

2

(25)

Davis and Duhaime (1992) who use the number of secondary business segments according to the SSIC2 classification3.

The proxy for substitutes in the product portfolio follows the variable describing the composition of the product portfolio by Haushalter (2000) for the sample of oil and gas companies.

Regarding the operationalization the control variables, the adopted measures are in line with existing literature. Haushalter, Heron and Lie (2001) suggest debt ratio composition as the ratio of book value of long-term debt and book value of both debt and equity. Again Haushalter, Heron and Lie (2001) use the natural logarithm of total assets as a proxy for company size.

3

(26)

IV. Data description

4.1. Sampling frame

The selection of the researched sample is based on the copper processing classification introduced by the International Copper Study Group (ICSG) as a specialized inter-government institution devoted to various copper researches. The ICSG annually publishes data about the 20 world largest mining, smelting, refining and fabricating sites and the companies that operate them. The latest publicly available report includes 51 companies and dates back to 2006. Previous studies on commodity exposure rely mostly on the internationally accepted Standard Industry Classification (SIC) (Haushalter, 2000; Jin and Jorion, 2006) but this type of industry reporting fails to distinguish between smelting and refining activities. The 2003 revision of the SIC does not even include a separate code for copper mining. The companies in the sample need to be listed so that I can obtain data about their stock returns. The requirement about public companies limits considerably the sample size. Some large government-owned companies like the Chilean Codelco, which alone operates the second world largest copper mine, get excluded. Datastream provides data on stock prices of 30 companies which are involved in the four copper handling processes. Due to data constraints, the pooled database for the second level regression consists of 142 complete company observations. The researched period of time is determined by the availability of data on company capacity in the mining, smelting, refining and fabricating subsectors for the period 2000 to 2005 in the ICSG Directory of Copper Mines and Plants. Initially the sample seems not so extensive but if compared with other studies, the size seems quite sufficient. Bartram (2005) includes between 4 and 14 companies in the longitudinal study of copper price risk exposure for each three-year period between 1987 and 1995. Tufano (1998) examines 48 North American gold-mining companies in the period January 1990 - March 1994.

4.2. Data sources

(27)

about the total assets of the companies, the long-term debt, the revenues from copper and aluminum production is collected from the corporate annual reports available on the Internet.

The reasons for missing data in the research database are mainly the following:

• the unavailability of the company annual reports for the period 2000-2005;

• the unavailability of data about the revenues from the copper and aluminum

production;

• the unavailability of Datastream data about the stock prices of the sample

companies for the whole period 2000-2005;

• mergers and acquisitions among the sample companies.

When the annual sales per metal are reported only in terms of volume but not in value, the copper and aluminum revenues are calculated using the average annual LME cash seller and settlement prices.

4.3. Descriptive statistics

The research sample is highly nationally diverse and includes companies from 17 different countries both developed and developing.4 Japan and Great Britain take part in the sample with the highest number of companies – eight and three respectively. 17% of the companies in the research sample have mining as an output market. 37% offer their final output at the fabricating market. Nearly half of the sample companies finish their added value chain with the refining market. There are no firms in the research sample that offer their final output at the smelting market. (Table 4)

As far as the extent of vertical integration of the sample companies is concerned, 73% of them are involved in two and more than two copper handling processes. 40% of all companies are involved in mining, smelting and refining at the same time. There are no companies involved separately in smelting and refining which suggests a strong interconnection between the two production processes. (Table 4)

4

(28)

Table 4. Process profiles of the sample companies

Company processes Number of companies

Share (%) in the sample

Mining only 5 0.17

Smelting and refining 2 0.07 Fabricating only 3 0.10 Mining, smelting and refining 12 0.40 Smelting, refining and fabricating 3 0.10 Mining, smelting, refining and fabricating 5 0.17

Table 5. Descriptive statistics of the dependent, independent* and control variables in the second-level regression

Determinants of the copper price exposure Copper

price exposure**

Bargaining power Vertical integration Substitutes in product portfolio Control variables β2,i Interaction mining dummy and the market share at the output market

Interaction fabricating dummy and the market share at the output market

Capacity mismatch AL/CU revenues Company indebtedness Company size mean _0.162 _0.001 _0.008 _5.899 _0.139 _0.230 _8.587 median _0.141 _0.000 _0.000 _6.031 _0.000 _0.207 _8.329 max _0.775 _0.011 _0.053 _7.535 _2.419 _0.555 _10.863 min _-0.771 _0.000 _0.000 _3.100 _0.000 _0.003 _6.453 st.dev. _0.181 _0.003 _0.013 _0.803 _0.440 _0.133 _1.115 Number of observations 166 180 180 180 153 173 173 *The dummy variables about the number of the production processes are not presented in the table as the data is already

outlined in the table with the process profiles. (Table 4) ** 27% of the copper price exposures are significant at the 1% level, 36% - at the 5% level and 43% - at the 10% level. 97% of the significant exposures at the 10% level are positive.

(29)

To test for a multicollinearity problem with the independent and control variables in the two level regression model, I analyze the Pearson correlation coefficients. I compare the absolute value of the correlation coefficients with 0.800 as suggested by Cooper and Schindler (2006, p.577-578). The value of 0.800 is the threshold value for excluding a variable due to a multicollinearity problem. Table 6 outlines the correlation between the unexpected changes of the copper price and the market stock returns from the first level regression. For the period 2000-2005 the correlation is positive and significant between 2002 and 2004. In 2002 the unexpected changes of the copper price and the market stock returns are correlated with the highest coefficient of 0.272 (significant at the 0.001 level)

Table 6. Correlation between the independent and control variables in the first level regression

Year

2000 2001 2002 2003 2004 2005 Pearson correlation coefficients -

unexpected change of the copper prices and market stock return

0.017 0.124 0.272 0.144 0.152 -0.017 Sig. (2-tailed) 0.781 0.045 0.000 0.020 0.014 0.783 Observations 260 261 261 261 262 262

Significant correlations at 0.01 and 0.05 level in bold

Table 7 presents the matrix of correlation coefficients between the dependent, independent and control variables of the second level regression. All significant correlation coefficients at the 0.01 and 0.05 levels do not exceed 0.800 in absolute value and I will retain all variables in the model of the second level regression. The strongest positive correlation exists between the four process dummy and the company indebtedness (0.422 significant at the 0.01 level). The two and three process dummies are linked with the strongest negative correlation (-0.425 significant at the 0.01 level). I will take into consideration the existing correlations among the independent variables when analyzing the results of the second-level regression.

(30)

Table 7. Correlation matrix copper price exposure and exposure determinants on annual basis for pooled database

Pearson correlation coefficients C o p p e r p ri c e E x p o s u re In te ra c ti o n m in in g d u m m y m a rk e t s h a re a t o u tp u t m a rk e t In te ra c ti o n f a b ri c a ti n g d u m m y m a rk e t s h a re a t o u tp u t m a rk e t C a p a c it y m is m a tc h D u m m y t w o p ro c e s s e s D u m m y t h re e p ro c e s s e s D u m m y f o u r p ro c e s s e s A l/ C U r e v e n u e s C o m p a n y s iz e C o m p a n y in d e b te d n e s s

Copper price exposure 1.000 -0.159* -0.072 0.080 -0.010 0.034 0.144 0.063 0.084 -0.125 Sig. (2-tailed) 0.041 0.359 0.307 0.901 0.661 0.065 0.453 0.286 0.116 Interaction mining

dummy - market share at output market

1.000 -0.205** -0.421** -0.130 -0.341** -0.127 -0.090 0.266** 0.120 Sig. (2-tailed) 0.006 0.000 0.083 0.000 0.090 0.266 0.000 0.121 Interaction fabricating

dummy - market share at output market

1.000 0.287** -0.256** -0.194** 0.217** -0.155 -0.387** -0.336** Sig. (2-tailed) 0.000 0.001 0.009 0.003 0.055 0.000 0.000 Capacity mismatch 1.000 0.011 0.079 0.205** 0.184* 0.192* 0.043 Sig. (2-tailed) 0.879 0.292 0.006 0.022 0.012 0.579 Dummy two processes 1.000 -0.425** -0.158* 0.405** -0.063 0.273**

Sig. (2-tailed) 0.000 0.035 0.000 0.410 0.000 Dummy three

processes 1.000 -0.415** -0.100 0.118 -0.272** Sig. (2-tailed) 0.000 0.217 0.123 0.000 Dummy four processes 1.000 -0.137 -0.084 0.422**

Sig. (2-tailed) 0.092 0.274 0.000 Al/CU revenues 1.000 0.351** 0.156 Sig. (2-tailed) 0.000 0.062 Company size 1.000 0.109 Sig. (2-tailed) 0.160 ** Correlation is significant at the 0.01 level (2-tailed).

(31)

V. Results

5.1. First level regression (determining the magnitude of copper price exposure)

Table 8 presents the copper price exposures of the sample companies including the exposures’ significance level, sign, mean and the average R2 statistic of the regression. The first level regression renders significant copper exposure coefficients at the 0.05 significance level about 17% to 50% of the sample companies every year in the period 2000-2005. The share of companies with significant exposures demonstrates an upward trend for the whole period of 2000-2005 following at the rise of copper prices. The finding that the copper price exposures change their value and significance over time confirms the research results by Tufano (1998) and Bartram (2005).

Table 8. Annual copper exposures of the sample companies estimated using Morgan Stanley Capital International World Investible index, daily data and the following regression model Nr of companies Time period Nr (share) of companies with significant β2,i at 10% level Nr (share) of companies with significant β2,i at 5% level Nr of companies with significant positive β2,i at 10% level Nr of companies with significant positive β2,i at 5% level β2,i mean R 2 mean 24 2000 7 (29%) 4 (17%) 6 3 0.127 0.043 26 2001 7 (28%) 5 (19%) 7 5 0.109 0.084 28 2002 15 (54%) 11 (39%) 14 11 0.187 0.112 29 2003 12 (41%) 11 (38%) 12 11 0.193 0.095 29 2004 15 (52%) 12 (41%) 15 12 0.150 0.132 30 2005 16 (53%) 15 (50%) 16 15 0.192 0.137

An increase of 1% of the returns on holding one ton of copper overnight results in an increase of 10.9 to 19.3% of the stock returns of the sample companies for every year in the period 2000-2005. The exposures tend to be predominantly positive and the average R2 statistic of the regressions is relatively small. The significant negative exposures in the sample are experienced by Sumitomo Corporation (Japan) in 2000 and Teck (Canada) in 2002. Both companies demonstrate positive profitability for the years of negative copper price exposure. The negative sign of the exposure is then explainable with the considerably higher share of copper costs in total costs as compared to the share of copper revenues in total revenues

, 0, 1, , 2, ,

(32)

Since 22 of the companies in the sample have mining activities in their added value chain, we should bear in mind that their input costs for the mining process are not affected by the quotations at the commodity exchanges. As the share of copper input costs in the total costs of the company for the mining process is zero, then starting the value-added chain with mining can be a possible explanation of the positive sign of the copper price exposures. But as the vertical integration of the companies is characterized with high capacity mismatch, the input costs for the processes after mining still get influenced by the commodity exchange quotations.

The results from the first level regression do not support a copper price exposure puzzle. Bartram and Bodnar (2007) describe the exchange rate exposure puzzle by stating that that in most studies on exchange rate exposures the share of significantly exposed companies is about twice the chosen level of statistical significance. The share of companies with significant copper price exposure exceeds two times the both the 5 and 10% levels.

Table 9. Comparison of copper exposure results estimated using MSCI world investible index and different data frequency for the period 2000 to 2005 and the regression model

Nr of companies

Data

frequency Time period

Nr (share) of companies with significant β2,i at 10% level Nr (share) of companies with significant β2,i at 5% level

β2,i mean R2 mean

30 daily 2000-2005 26 (86%) 25 (83%) 0.183 0.095

29 monthly 2000-2005 19 (73%) 15 (52%) 0.463 0.229

To make the current results comparable to the findings of other studies on specific industries, I take in mind the regression model characteristics of these studies. Tufano (1998) finds that 56% of the gold mining companies in his sample are positively and significantly exposed to gold prices as the gold exposures are estimated for a period of five years. Using a similar time period and daily data, the current research sample renders 83% of significant exposures at the 0.05 level. (Table 9) On the basis of monthly frequency data to estimate oil and gas exposures for the period 1999-2000, Jin and Jorion (2006) find that 87 % of the oil and gas companies have positive and significant gas exposures at 0.05 level and 29% of the same firms have oil exposures. With the same data frequency for the period 2000-2005 the research sample shows 52% of significantly exposed companies. (Table 9) The comparison with other commodities

, 0, 1, , 2, ,

(33)

exposure studies confirms the conclusion by Bartram (2005) that commodity price exposure is highly industry specific.

5.2. Second level regression (effect of Porter’s five forces on copper price exposure

)

Table 12 contains the regression coefficients and their t-statistics for testing the four research hypotheses. The White test for heteroscedasticity applied as described in Dougherty (2007, p.230) shows that the null hypothesis about homoscedasticity of the residuals is not rejected. The regression model for the heteroscedasticity test accounts for 12.7% of the variance in the squared residuals.

Table 10. The Effect of Porter’s five forces on the copper price exposure

The effect is examined with the use of a time-series cross-sectional database and a random-effects multiple regression for the following model . The four variations of the model include different combinations of the variables defined in Table 3. The t-statistics are presented in parentheses.

Regression coefficients Dependent variable: Copper price

exposure I II

Intercept 0.481 (3.288) ** 0.464 (3.467) ***

1. Bargaining power

interaction dummy fabricating - market

share at the output market -4.287 (-3.086) ** -4.337 (-3.645) *** interaction dummy mining - market share

at the output market -14.738 (-1.634) -13.068 (-2.232) * 2. Extent of vertical integration capacity mismatch -0.004 (-0.217) --- dummy 2 processes 0.0132 (0.202) 0.020 (4.492) *** dummy 3 processes -0.009 (-0.170) --- dummy 4 processes 0.194 (2.995) ** 0.200 (4.449) *** 3. Substitutes in the product portfolio ratio of AL and CU revenues 0.061 (1.924) 0.061 (1.940)

4. Control variables company size -0.013 (-0.850) -0.015 (-1.078) company indebtedness -0.549 (-4.338) *** -0.553 (-4.434) *** Model statistics Nr of observations 142 142 R2 _0.195 _0.194 Probability 0.001 0.000

*** Significant at the 0.001 level; ** Significant at the 0.01 level; * Significant at the 0.05 level

2,i 0 1Xi i

(34)

The results in Table 10, model variation I outline the regression coefficients for all the independent and the control variables. Only two of the independent variables get significant coefficients – the interaction fabricating dummy-market share at the output market (γ = -4.287 significant at the 0.01 level) and the four processes dummy (γ = 0.194 significant at the 0.01 level). Based on the number of significant exposure determinants in variation I and the significant correlations between the independent variables, I intend to test whether the inclusion of insignificant variables results in insignificant outcome for exposure determinants that are actually significant.

The capacity mismatch and the dummy variables about the number of production processes measure two aspects of vertical integration. The capacity mismatch is negatively correlated with the interaction mining dummy – share at the output market (Pearson correlation coefficient -0.421 significant at the 0.01 level, Table 7) Moreover, the dummy variable three processes is strongly negatively correlated with the dummies about two and four integrated processes (respectively Pearson correlation coefficients -0.425 and -0.415 significant at the 0.01 level, Table 7). Therefore in variation II of the regression model the vertical integration is measured only in terms of dummy variables about the number of integrated production processes. In variation II the interaction mining dummy – market share at the output market and the two process dummy render also significant regression coefficients. If the vertical integration is measured with the capacity mismatch and the three process dummy, the regression model gets significant at the 0.05 level. Both variations of the regression model are significant at the 0.001 level and explain almost 19.5% of the variance in the dependent variable. The following equation represents the variables with significant effect on copper price exposure as estimated in the second variant of the regression model:

0.464 4.337 *[ ] 13.068*[ ] 0.020 * 2 0.200 * 4 0.553* InteractionDummyMining MarketShareOutputMarket InteractionDummyFabricating MarketShareOutputMarket

Dummy processes Dummy proceeses Indebtedness

β

= − −

− −

+ + −

(5)

(35)

fabricating market seems to have a major determining impact on the market balance of power. The consumers of fabricated copper products consist of five different industries – transportation, electrical equipment and electronics, building and construction, industrial machinery and consumer products. (The World Copper Factbook 2007, p.7) Moreover Geman (2005, p.171) points out that the customers of the metal fabricated products produce at higher profit margins than the companies on the supply side in fabricating. Therefore the companies on the demand side seem more inclined to sacrifice a part of their profit margin and the suppliers can more easily pass through price increases.

The second research hypothesis about the effect of bargaining power on copper price exposures gets empirical support in the results of variation II of the regression model. The interaction dummy mining – market share at the output market has a negative effect on the copper price exposure, significant at the 0.05 level. (γ = -13.068) The bargaining power of suppliers is not the only possible explanation why the companies involved only in copper mining tend to have lower copper price exposures The dummy variable about mining as the end market of the value chain helps in selecting companies whose costs are not influenced by the copper prices at the commodity exchanges. As the companies involved only in mining cannot pass price decreases to their suppliers, they are highly motivated to make successful use of the remaining strategies for managing copper price exposures and have lower exposures.

(36)

integration failure is common and the major reasons behind it are the difficulties with the coordination of inventories and capacity and the high establishment costs.

Except the transaction cost approach employed in motivating Hypothesis 3, the current studies on vertical integration use also the resource-based view of the firm. A vertical integration decision by a company can be put into practice through external or internal growth – merger/acquisition or establishing own company. In both cases the relation between the existing and the newly added company involves transfer of knowledge, technology and assets. According to Broedner, Kinkel and Lay (2009) the companies use vertical integration to make effective use of their resources and most of all to create unique capabilities that will turn into sustainable competitive advantage and cost minimization is not the main goal any more. By company capabilities here Broedner, Kinkel and Lay (2009) mean complex patterns of coordination among people and between the human and the remaining resources of the companies. The idea that the cost minimization is not the main objective of vertically integrated companies in the sample is supported by the insignificant result of capacity mismatch as a metric variable of the extent of vertical integration.

The fourth research hypothesis about the substitutes in the product portfolio of the company receives no support in both model variations. Although not significant, the direction of the effect is positive as hypothesized. Haushalter (2000) finds a negative significant effect of the ratio of oil and gas revenues on the hedged proportion of the production for the sample of 100 US oil and gas companies. As the use of hedging instruments and the commodity price exposure are negatively correlated, the expected effect of the oil-gas revenues ratio on the exposure to oil prices is positive

Porter’s Five Forces and Their Effect on Copper Price Exposure of Companies in the Copper Industry

Porter’s Five Forces and Their Effect on Copper Price

Exposure of Companies in the Copper Industry

Gergana Todorova VALKOVA

Student number: 1834398

G.Valkova @student.rug.nl

Supervisor: Dr. Nanne Brunia

Co-accessor: Dr. Wim Westerman

Porter’s Five Forces and Their Effect on Copper Price

Exposure of Companies in the Copper Industry

Abstract

Table of contents

I. Introduction

II. Literature review

βπ

π

π

π

π

β

III. Methodology

IV. Data description

V. Results

)

β