Long-term iron ore price modeling: Marginal costs vs. incentive price

Alexander Pustovn, Alexander Malanichev1, Ilya Khobotilov1

14661 Rotunda Dr Dearborn, MI 48120, United States

a r t i c l e i n f o

Article history:Received 4 July 2013Accepted 10 September 2013

Jel classification:E37

Keywords:Marginal costsIncentive priceLong-term priceIron oreForecasting

a b s t r a c t

The paper studies and applies the approaches to forecast long-term (LT) real prices of iron ore. This priceis crucial for valuation of investments in Greenfield iron ore projects on the horizon of more than 5 years.The forecast is obtained by three different approaches which are usually used by investment bankanalysts: marginal costs approach and 2 approaches based on calculation of incentive price. The paperconcludes that there has been a structural shift on the iron ore market and LT iron ore prices will behigher by 20–30% than the average of industry forecasters suggest. This is related to the 2 key factorswhich were taken into account in this study—depletion of existing iron ore deposits and targeted returnon investments for new projects. In addition, escalated industry costs inflation is claimed to be the factorwhich will bolster nominal iron ore prices at high levels in the long-term. Using a Monte-Carlosimulation approach, confidence interval for future iron ore price was estimated.

& 2013 Elsevier Ltd. All rights reserved.

Introduction

Iron ore2 is the key input in pig iron production, which is thenused to manufacture steel. Iron ore as well as many othercommodities is noted for both persistent trends and rapid turn-arounds in terms of global prices. There was a long period ofdeclining prices in real USD terms from mid-1960s to early 2000s,which led to underinvestment in new iron ore supply projects(Fig. 1). Generally there was no interest for companies to invest ingreenfields on the market for which demand prospects andinvestment returns were questionable. So when 10 years agoChina's steel production boom has set to emerge, tight iron oresupply conditions on the back of robust demand growth allowedthe iron ore price to take off (Sukagawa, 2010). From 2003 to 2012price has increased 5-fold to $127/t on FOB Brazil basis.

This growth can be attributed to 4 key characteristics of theiron ore market:

� Robust iron ore demand driven by China. Global iron oredemand grew 6% yoy during the last 10 years while China's oneadded more than 15% yoy. Global iron ore consumption

exceeded 1900 million metric tonnes (mmt) in 2011 of whichChina consumed almost 1000 mmt [Worldsteel, 2012].

� Persisting supply constraints.3 There is generally not enoughexisting iron ore capacity to satisfy ever growing demand foriron ore, which encourages implementation of new projects.However, this new supply is not able to keep up with thedemand. Implementation of new projects is impeded by anumber of reasons such as infrastructure bottlenecks, skilledlabor shortage, growth in capital intensity, increased regulatoryrisk with rising taxes globally, relocation of business to regionswith immature business environment like Africa, lack of newequipment, etc. (De Angele, 2011).

� Steep cost curve, deposit depletion and operating costsinflation. Steep 4th quartile of the iron ore supply (cost) curveleads to high price sensitivity to even small changes in demand.As noted in many reports, China is the marginal iron oreproducer almost entirely occupying the highest cost 4th quar-tile of the iron ore cost curve and, therefore, determining ironore price (Malanichev and Pustov, 2011). This right part of thecost curve becomes even steeper with the time because China'siron ore production costs inflation, which grew more than 25%yoy historically, exceeds inflation of countries on the left side ofthe curve. This is amplified by rapid deposit depletion headingup to 3% per annum globally which is partially represented bydeclining % Fe content in iron ore [Hamilton, 2011].

� High level of industry consolidation with the BIG-3 players(Rio Tinto, BHP-B and Vale) accounting for close to 70% of theinternational iron ore trade (Zhu, 2012). This indicates that the

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/resourpol

Resources Policy

0301-4207/$ - see front matter & 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.resourpol.2013.09.003

n Corresponding author. Tel.: þ7 495 926 77 66; mob.: þ1 313 412 04 38;fax: þ7 499 150 8800.

E-mail addresses: alexander.pustov@severstalna.com (P. Alexander),aleksander.malanichev@severstal.com (M. Alexander),ilya.khobotilov@severstal.com (K. Ilya).

1 Tel.: þ7 495 926 77 66; fax: þ7 499 150 8800.2 Hereinafter “iron ore” stands for iron ore fines, 62% Fe. Iron ore is the key

input in pig iron production; the latter is used to produce steel. 3 Terms “production” and “supply” are used interchangeably.

Resources Policy 38 (2013) 558–567

iron ore market is oligopolistic with the BIG-3 being able toinfluence market prices and receive a higher price premiumcompared with a market of perfect competition. A goodexample of BIG-4's strong stance on the market is the 2010'sshift from the 40-year-old benchmark system of negotiatedannual contracts to quarterly contracts directly linked to theiron ore spot market [Wilson, 2012]. This move allowed theBIG-4 to extract higher profits from the growing spot prices.

These high prices make the iron ore market very attractive fornewcomers aspiring to bring new supply to fill in the demand gapand earn solid profits. But in order to justify capital investments innew projects future price assumptions are required. Given thatiron ore projects have a long lead time nature, with engineeringand construction taking up to 10 years (Correa and Grimaldi,2012), investments need to be evaluated on the long-term horizon.Therefore, 2022 year was taken as a proxy to long-term period.Usually long-term prices used in financial models are convertedinto real USD terms to neglect inflation. So the true question is,where the long-term price in real USD terms is headed—is it goingto grow further on the wave of Chinese development or fallsignificantly as it was seen in the past?

According to consensus forecast by the leading investmentbanks4, the iron ore price would ease to $88/t in the long-termfrom around $150/t in 2012 (Fig. 2). At first glance this looks logicalconsidering that the current price is exceeding the past high by3-times and currently proved iron ore reserves would suffice asmany as 40 years of consumption at today's rates. The historicalanalysis of commodity price cycles confirms that price peaks werealways not sustainable (Cashin et al., 1999). Moreover, the conceptof LT price itself assumes that the market would cool down afterthe boom period. China would decelerate its steel production andiron ore consumption growth. India would not become “a secondChina” as it does not need that much iron ore to be importedbecause it has enough domestic iron ore to produce steel. Iron oresupply tightness would ease with the entry of new projects, whichtotal volume in theory could account for almost 3 billion metrictonnes in the next 10 years, if projects are delivered within theannounced timeframes. This exceeds current global production byas much as 1.5 times.

However, there are several reasons to doubt that the consensusforecast of LT iron ore price properly reflects the new reality ofrapid deposit depletion, constrained supply growth and escalatedcost and capex inflation:

� It is not clear what assumptions stand behind the consensusforecast as there is no explicit relationship between supply–demand balance and iron ore price in the available models.

� There are no modern transparent models verified by scientificcommunity. Available models like the World Bank's one or themodel considering the effects of Carajas iron ore on global ironore prices (da Silva Neto, 1993) have been developed long timeago and do not take into account structural changes on the ironore market.

� Traditionally, available to the authors forecasts do notexplicitly take into account iron ore deposits depletion as wellas escalated industry operating and capital costs inflation.To convert prices from real to nominal terms they usually usejust a US consumer prices inflation or GDP deflator, which leadsto underestimation of forecasts.

� Based on the 10-year retrospective analysis, it is evident thatconsensus forecasts are usually lower than actual iron ore

prices. The reason for this is that usually market analystssimply assume LT price to equal the average historical priceover the cycle.

Thus, this research aims at achieving 3 following objectives,which would help clarify if prices could stay at those high levelsfor at least a decade to come:

� Overview and debate pros and cons of existing approaches toLT price forecasting.

� Explicitly state and discuss assumptions used to forecast pricesincluding deposits depletion as well as industry operating andcapital costs inflation.

� Show a possible range of forecast prices and evaluate its upsideand downside uncertainties.

Overview of approaches to LT price forecasting

The paper analyzes LT price forecasts produced via three keyapproaches which are usually used by commodity analysts frominvestment banks [e.g. Barclays, Merrill Lynch, Citi, Itau BBA,J.P. Morgan, 2011], analytical agencies [e.g. CRU, 2011] and indus-trial companies [e.g. Cochilco, 2008]:

(1) Marginal costs (MC) approach;(2) Global marginal incentive price (MIP) approach;(3) BIG-4 MIP approach.

Marginal costs (MC) approach generally assumes that LT priceequals marginal operating costs (Jones, 1986), i.e. highest produc-tion costs needed to bring the last piece of supply to the market.The 2 elements of data required in this approach are operatingcosts split by country and assumptions regarding future demandgrowth (Table 1).

The MC approach gives the lowest threshold of prices at a givendemand, because it assumes that a new project, being marginal,will never pay back the invested capital.

The second and third approaches use the concept of incentivepricing. This concept is based on the notion that producers willonly invest if they believe that prices will be high enough to coverall their operating and capital costs and provide target return oncapital over the life of the mine project (Berry and Cooper, 2011).Both approaches assume that LT price equals marginal incentiveprice. The key difference between global marginal incentive price(MIP) and BIG-4 MIP is that the former considers incentive pricesfor the whole pool of announced projects. While the BIG-4 MIP isbased only on incentive prices for projects owned by BIG-4, whichassumed to have the best knowledge and bargaining power on themarket [Wilson, 2012].

General assumptions for MC and global MIP approach

In the long-run a number of key assumptions should be used todetermine LT prices:

� As a proxy for the LT period global iron ore demand growth5

in 2012–2022 is taken: compound annual growth rate (CAGR)of 1.8%, which is a decrease from 7.0% over the past decade dueto an expected slowdown of steel consumption growth inChina. This will expand the global iron ore market by384 mmt to reach circa 2189 mmt (Fig. 3). In comparison

4 Consensus is an average of Citi, Macquarie, J.P. Morgan, UBS, Merrill Lynch,Morgan Stanley, Credit Suisse. 5 All S–D balances are calculated for 63% Fe content.

P. Alexander et al. / Resources Policy 38 (2013) 558–567 559

consensus projects a lower rate of 3% annual growth in thefuture suggesting that our forecast is conservative. Markets areconsidered to be in equilibrium, so demand equals supply.

� Global iron ore demand is considered inelastic relative toprice, i.e. iron ore consumption is only marginally dependenton its price (Gonzales and Kaminski, 2011). This is true becauseiron ore constitutes about 30% of rolled steel production costsand the latter contributes only as less as 5 to 20% of final costsof product like an automobile or a building (World steel infigures, 2011).

� Ex-China iron ore capacity would grow in 2012–2022 at circathe average level of annual capacity additions over the last5 years of roughly 100 mmt. This equates to a 1014 mmtincluding 350 and 189 mmt to satisfy demand growth on theinternational market and in exporting regions, respectively,155 mmt to cover capacity stoppage in China and 320 mmt tosubstitute depleted deposits. This forecast is less than expecta-tions of some industry analysts who forecast capacity additionsto exceed 1000 mmt even without projects of BHP-B, Vale andRio Tinto (De Angele, 2011) as well as consensus forecast at6–7% CAGR. However, this is unlikely, given that industry isfaced with severe project delays exceeding 3 years for majorprojects, which dramatically constrain supply as mentionedearlier (Tonks and Blanch, 2011). Moreover, many ex-BIG-4projects including those financed by Chinese steel companiesare unlikely to become producing at all given that most of themare economically unviable [Wilson, 2012].

� China's government follows the rational economic behavior,i.e. they will stop its domestic iron ore production if it is lesscost-efficient than imported ore. In our base scenario 60% or15 mmt of production is stopped. It is worth mentioning that inreality China's iron ore production is unlikely to be stopped oreven decreased dramatically under pressure from cheap oreavailable via imports. This is circa 1 billion metric tonnes (inrun-of-mine grade) industry which employs dozens of millions

of Chinese people and if a significant number of mines isstopped this could lead to serious political tensions and socialturmoil.

� Depletion of existing deposits is assumed at circa 30 mmtglobally each year according to McKinsey's estimates[McKinsey, 2013]. This means that in order to simply sustainiron ore production at the current levels, new projects need todeliver 320 mmt to the market in the next 10 years. The processof depletion is perfectly described by the following quotation:“Today's rich ore was yesterday's waste” (Antunes, 2012).

So, once the long-run assumptions regarding the iron oremarket are made and discussed, the next steps in projecting LTiron ore price should be as follows:

� Build the long-term supply curve on operating costs for MCapproach or “incentive” supply curve on operating costs andcapital investments of new iron ore projects for global MIPapproach.

� Find LT iron ore price at an intersection of demand and supplycurves.

Marginal costs based forecasts

Gauging the 1538 mmt (i.e. international supply and China'sdomestic supply in 2022) on the supply curve which is based onoperating costs, the LT price comes out at $85/t or circa one-halfof iron ore price in 2012 (Fig. 4). This is the marginal cost ofChina's high cost iron ore production, as in this scenario Chinawill continue to produce 110 mmt compared to circa 250 mmtin 2012.

As we would see in the next chapter, demand and supply, arethe key factors of LT price sensitivity and uncertainty. That iswhy Table 2 was built to test how LT price would behave underdifferent supply–demand scenarios. If we were to assume 1% perannum depletion rate, compared 3% historically, and 1% rate ofdemand growth, LT price would have fallen to $63/t. In fact, wecan already observe that many market participants, especiallyBIG-4, now project the future demand to deteriorate. But this isperhaps a strategy suggested by game theory. BIG-4 intends toalter market perceptions of future outlook by suddenlyacknowledging a deterioration of demand. This is aimed atmaking junior iron ore players appear vulnerable so that itwould be significantly more difficult for them to raise capitaland seek project approval. It should help BIG-4 effectivelyattune the future supply–demand balance in their favor (Priceand Garran, 2012). However, if demand accelerates to 3.6%,

Fig. 1. Global iron ore price dynamics in 1920–2012.Sources: SBB [2], prepared by the authors.

Fig. 2. Historical and projected global iron ore price.Sources: for calculation of consensus forecast: Macquarie, Credit Suisse, MerrillLynch, J.P.Morgan, Citi, Goldman Sachs as of Mar-May 2013.

Table 1Comparison of elements taken into account in 3 different approaches to long-termiron ore price forecasting.Source: prepared by the authors.

Element 1) MarginalCosts

2) Global MarginalIncentive Price

3) BIG-4 MarginalIncentive Price

Operating costs

Return onInvestment

–

Demand growth –

Oligopolisticcharacter of themarket

– –

P. Alexander et al. / Resources Policy 38 (2013) 558–567560

which half of the historical growth, and depletion rate stays at3%, iron ore price would jump to $156/t.

In order to estimate confidence interval of iron ore price weapplied Monte Carlo Methodology [Malvin, 2008]. Monte Carlomethods are a class of computational algorithms that rely onrepeated random sampling to compute their results. In our casethe variable parameter is the realization of iron ore projects:different combination of realized projects changes the cost curve(Fig. 4) and results in different LT price. The algorithm includes thefollowing steps (Table 3):

� Each iron ore project in the database has its Nominal capacity,Cost of production and Odds of its implementation. Odds wereestimated based on the range of parameters for each project (e.g.implementation stage, availability of financing, location, etc.);

� N�K array of random numbers in the 0–1 range is generated,where N – number of projects in the database, K – number ofvariations; 115�1000 in the case;

� Each of N project's odds is compared with a random number ineach of K variations: if a project's odds is higher than a randomnumber in a given variation, then the project assumed to beimplemented and vice versa;

� Each implemented project comes into a probable cost curvewith its full Nominal capacity. Therefore K probable cost curvesare obtained. Given the long-term demand estimation K prob-able LT prices are obtained, and LT price's probability distribu-tion is determined (Fig. 5).

Another major uncertainty factor which influences LT price andshould be taken into account when estimating confidence intervalis long-term iron ore demand. Three alternative iron ore demandgrowth scenarios are developed, based on various global economydevelopment and steel consumption/production assumptions.A probability for each scenario was estimated by authors basedexpert interviews.6 Applying long-term demand variations toprobable cost curves’ spread derived by Monte Carlo method,long-term price's probability distribution for each demand growthscenario is obtained. Multiplying long-term price's probability by

Fig. 3. Global iron ore supply split.Sources: worldsteel, analysis and forecast by the authors.

Fig. 4. International and China's domestic iron ore supply curve in 2022.Sources: Macquarie, Merrill Lynch, Goldman Sachs, analysis by the authors.

6 List of experts: Larry Hatheway, UBS, chief economist; Yaroslav Lissovolik,Deutsche bank, chief economist for CIS region; Charles Roxburgh, Director,McKinsey Global Institute; John Walker, Director, Oxford Economics; Willem Buiter,Chief economist, Citigroup; Jacob Nell, executive director, Morgan Stanley research;John Johnson, CEO, CRU.

P. Alexander et al. / Resources Policy 38 (2013) 558–567 561

the scenarios’ probabilities, weighted average long-term priceprobability distribution is calculated (Table 4).

Based on this analysis, LT price would range from $85 to $95/tin MC approach with 92% probability.

To recap on MC approach, on the strong side it is easilyexplainable and is related to operational decision-making regard-ing producing iron ore mines but it also has a number of short-comings. The most obvious one is that it does not take intoaccount the capital cost element which is quite important fordevelopment of new projects. This is resolved by the incentiveprice approaches presented later in the paper.

Incentive price concept

Incentive price is the price point that encourages the entry ofa new project by covering not only operating costs but alsodiscounted capital investments and brings target return on

invested capital. The calculation is based on the followingassumptions:

a. Net Present Value (NPV)¼0;b. Cash flows are a constant for each year of operation;c. Internal Rate of Return (IRR) is fixed at a certain %. In our case

this is 15%, which is standardly used to valuate projects by theleading investment banks [Barclays, Citi, 2011] and industrialcompanies [Cochilco, 2008];

d. Incentive price for any project shipping from certain exportports is calculated on CFR China basis using a net backcalculation: CFR China¼FOB export portþfreight from Exportport to China;

e. Life of mine¼25 years, after which the remaining capital isspent to shut down the mine;

f. Construction period¼6 years using major iron ore projects as abenchmark;

g. Corporate tax rate is assumed based on the regional location ofthe project.

Table 2Long-term iron ore price in MC approach—sensitivity to supply–demand changes, USD/t, real 2011 USD terms, CFR China.Source: prepared by the authors.

International iron ore demand and China's demand for domestic iron ore in 2022

CAGR in 2012–22 (%) mmt Depletion rate, % per annum

1.0% 2.0% 3.0% 4.0% 5.0%

�0.3 1295 $63 $66 $69 $72 $750.5 1417 $68 $72 $75 $83 $851.2 1539 $74 $82 $85 $92 $922.4 1751 $89 $92 $101 $111 $1363.6 1993 $105 $136 $156 na na

na—No available supply.

Table 3Example of Monte Carlo method use for iron ore projects implementation in multiple variations.

Projects Incentive price,USD/t FOB Brazil

Nominal capacity,mmt/y

Project'sodds (%)

Random numbers in the 0–1 range Randomized nominal capacity, mmt/y

Var 1 (%) Var 2 (%) Var 3 (%) … Var K (%) Var 1 Var 2 Var 3 … Var K

Project 1 $60 20.0 70 4 77 93 67 17 20.0 20.0 20.0Project 2 $80 10.0 20 57 34 83 79 2 10.0Project 3 $100 50.0 50 56 64 93 7 5 50.0 50.0…Project N $200 15.0 15 27 86 66 45 77

Fig. 5. International and China's domestic iron ore probable supply curves in 2022, and long-term price's probability distribution.Source: prepared by the authors.

P. Alexander et al. / Resources Policy 38 (2013) 558–567562

h. Mathematically first assumption can be presented in thefollowing formula:

NPV¼ ∑n

i ¼ 0

FCF

ð1þ IRRÞi¼ ∑

n

i ¼ 0

ðNOPATþDepr�CAPEXÞð1þ IRRÞi

¼ 0 ð1Þ

where

FCF Free cash flow;NOPAT Net operating profit after tax;

NOPAT¼ ðrevenue�royalty�cashcost� deprÞ � ð1�taxÞ ð2Þ

Revenue¼ V � IP ð3Þ

V annual production volume, mmt;IP Incentive Price, raziBUSD/t, CFR China;tax corporate tax rate, %;n period consisting of lead time (D) and life of mine

(LOM), years;CAPEX capital expenditure, USD;Depr depreciation (CAPEX/LOM), USD;

Let's also introduce a coefficient ri:

ri ¼1

ð1þ IRRÞið4Þ

Given formulas (2)–(4), IP formula can be derived from the NPVequation (1):

IP ¼Operating CostsþEffect CAPEX Charge�Effect Dept Tax Shield;

ð5Þwhere

Effect CAPEX Charge¼ CAPEXV

� 1D� 1

1�tax� ∑D

i ¼ 0ri∑n

i ¼ Drið6Þ

– effective CAPEX charge;

Effect Depr Tax Shield¼ CAPEXV

U1

LOMU

tax1�tax

U ∑n

i ¼ Dri ð7Þ

– effective depreciation tax shield.

Effectively, this calculates an iron ore price which will allow theproject to have a certain IRR. In the example calculation (Table 5)the incentive price equals $124/t.

It should be mentioned that calculation of incentive prices fordifferent iron ore projects also includes normalization to a singlebase of iron content, 62% Fe in our case. Though for simplicityreasons this was not shown in formula (5).

The key advantage of incentive price is essentially its disad-vantage at the same time. Investors would receive the expectedprofits only in the case if future supply–demand is not worse thanexpected. When this is not the case, new projects would coveronly their operating costs without returning capital to investors. Inorder to investigate the possible impact of this, sensitivity anduncertainty analysis was conducted.

Sensitivity and uncertainty analysis

Sensitivity and uncertainty analysis is important for determiningkey factors which are most relevant for price forecasting. Sensitivity is

calculated using the following equation (Saltelli et al., 2008):

εi ¼ΔIPi

IP� FiΔFi

ð8Þ

where

εi sensitivity (change) of normalized price to change infactor i, %;

IP incentive price, USD/t;ΔIPi incentive price change given a factor i change;Fi factor i value, which influences IP (demand, Fe content,

etc.);ΔFi 1% change of factor i value (mmt, USD/t or %).

The analysis of key factors suggests that incentive price is mostsensitive to demand growth and % Fe content in iron (Fig. 6). The latteremphasizes the high influence of deposit depletion on the prices, asdeclining % Fe grades is one of the indications of this process.

Apart from sensitivity, uncertainty analysis which takes account offactor variations is also interesting to conduct. Uncertainty analysis isbased on the following formula (Saltelli et al., 2008):

Ui ¼ εi �siFi

ð9Þ

where

Ui uncertainty of price change given standard deviation offactor i, %;

εi sensitivity (change) of normalized price to change infactor i ,%;

Fi factor i value (mmt, USD/t or %);si standard deviation of factor growth which is assessed

based on historical data (for demand and freight rates) orcurrent data contained in the project database (for allother factors);

si=Fi variation of factor i, %.

According to the analysis, IRR, OPEX and CAPEX are importantin terms of uncertainty impact on price (Fig. 7). This is mostly dueto the fact that these factors have large variation.

As to royalty a 30% rate was tested and apparently there is onlya modest 17% iron ore price uncertainty with regard to royalty% hike, given that the price is not highly sensitive to royalty changes.However, in July 2012 a new 30% mining tax was introduced inAustralia [The Age, 2012]. The LT price might experience growth ofroughly 20%, if the new tax will not be ceased.

So to recap, LT price forecasts could differ significantly based ondifferent values of factors. However, the key factor for bothsensitivity and uncertainty is the global demand growth and,hence, supply growth. In order to remove price uncertaintyregarding these factors, it makes sense to build different scenariosof supply and demand based on estimation of probabilities,allowing gauging confidence intervals of the forecasted price. Thiswas analyzed for MC approach in Table 2 and will be investigatedlater in the paper for Global MIP one.

Global marginal incentive price approach

The future “incentive” supply curve was built using incentiveprices for new projects and operating costs for existing ones (Fig. 8).

This approach is based on an assumption that new projects’delivery requires the forecasted LT price to be equal or higher thanthe incentive price, which requires two assumptions. Firstly,market participants are able to correctly forecast LT prices.Secondly, new projects coming to the market need to cover not

P. Alexander et al. / Resources Policy 38 (2013) 558–567 563

only operating costs but also invested capital. Hence, whendeciding whether to start a project, companies would expect LTprice to be not less than the project's incentive price.

Because new projects with less incentive prices are considered tocome in the market first and foremost, there is a possibility that LTprice in this approach would be determined by marginal costs andnot by incentive price. In theory this could not happen, because ifthere is demand on the market, new low-cost projects will substitutethe existing ones. But in reality there are a lot of hurdles delayingimplementation of projects (Hynes and Jansen, 2011).

Moving on to the specifics, the 2022 iron ore demand isprojected at around 1538 mmt, therefore, the LT iron ore price

would equal $113/t. This is an incentive price of Strike's Apurimanproject in Peru.

Considering LT price sensitivity to different supply–demandscenarios, we come up with $85/t assuming 0% return on newprojects and 3% rate of demand growth (Table 6). This is almostidentical to $88/t, which consensus expects, probably suggestingthat they either do not take into account deposit depletion orreturn on investments in their forecasts.

Applying Monte Carlo analysis described above to iron oreprojects with their incentive prices, probability distribution oflong-term iron ore prices in Global MIP approach is obtained Fig. 9.

Taking into account various iron ore demand growth scenarioswith their probabilities, weighted average LT price probabilitydistribution in Global MIP approach is calculated.

Based on this analysis, LT price would range from $115 to $125/tin Global MIP approach with 76% probability Table 7.

So, while the Global MIP approach is easily explainable andtakes account of capital investments, it still has most of theshortcomings of MC approach:

� It requires the long-term demand growth assumption, which ishard to predict correctly.

� It requires future “incentive” supply curve, which is built onfragmented data regarding project operational costs and capitalinvestments.

� It is highly sensitive to small changes in demand as the 4thquartile of the supply curve is very steep.

� It assumes the iron ore market to be in perfect competition,while it really is an oligopoly (Hui and Xi-huai, 2010).

The next approach, based on BIG-4 Marginal Incentive Pricedoes consider the oligopolistic character of the iron ore marketand does not require demand and supply projections, as BIG-4 areassumed to implicitly provide this information in doing theirinvestment decisions.

BIG-4 marginal incentive price approach

BIG-4 marginal incentive price (MIP) approach assumes thatiron ore market is oligopoly, where BIG-4 is the most informed andinfluential market players controlling 70% of international trade

Table 4Long-term iron ore price probability distribution.Source: prepared by the authors.

InternationalþChina domestic demand Price range, USD/t 62% Fe CFRChina (7$5)

2012–2022CAGR, %

Volume in2022, mmt

Probability(%)

$65(%)

$75(%)

$85(%)

$95(%)

$105(%)

�0.3 1295 2 5 37 51 7 00.5 1417 22 0 17 66 18 01.2 1539 60 0 6 74 21 02.4 1751 16 0 0 34 67 0Total price range probability, % 0 8 65 27 0

Table 5Incentive price calculation for Vale's Carajas Serra Sul project in Brazil, USD/t unlessstated otherwise, real 2011 USD terms.

# Variable Value

1 OPEX $342 Royalty $33 Operating Costs $374 CAPEX, mn.USD $200004 Capacity, mn t 905 CAPEX, USD/t $2226 Corporate tax rate 30%7 Life of mine, years 258 Effective tax shield on depreciation �$59 IRR 15%10 Construction Period (years) 611 Effective CAPEX over mine life $7412 Incentive price, FOB Brazil $10913 Incentive price, CFR China $124

Fig. 6. Sensitivity analysis of iron ore incentive price.Source: prepared by the authors.

Fig. 7. Uncertainty analysis of iron ore incentive price.Source: prepared by the authors.

P. Alexander et al. / Resources Policy 38 (2013) 558–567564

market and are expected to hold this influence in the long-term[Wilson, 2012]. Therefore, the incentive price of the BIG-4'smarginal project7 could be used as a benchmark for LT price.

It also assumes that it is in the interest of the BIG-4 tounderestimate CAPEX of new projects in order to increase entrybarriers for other companies and get favorable funding (Price andGarran, 2012). This means that the approach estimates the lowerboundary of LT price.

Based on the results of analysis presented in Fig. 10, the long-termprice derived via this approach is US$124/t or the incentive price ofVale's Carajas Serra Sul project. This project has the following details:

� Total CAPEX¼$20 billion� Capacity¼90 mmt� CAPEX per tonne¼$222� OPEX per tonne¼$34� Royalty per tonne¼$3

BIG-4 MIP approach effectively eliminates all of the short-comings of the previous approaches as it assumes that the future

Fig. 8. “Incentive” supply curve based on combined existing iron ore supply cost curve and incentive price curves for international iron ore market in 2022.Sources: Macquarie, company reports, analysis by the authors.

Table 6Long-term iron ore price in Global MIP approach—sensitivity to supply–demand changes, USD/t, real 2011 USD terms, CFR China.Source: prepared by the authors.

International iron ore demand and China's demand for domestic iron ore in 2022

CAGR in 2012–22 (%) mmt IRR, %

0% 10% 15% 20% 25%

-0.3 1295 $72 $89 $101 $117 $1330.5 1417 $77 $92 $107 $127 $1421.2 1539 $85 $96 $113 $137 $1512.4 1751 $101 $111 $130 $150 $1793.6 1993 $156 $152 $151 $182 $237

Fig. 9. Probability distribution of long-term iron ore prices in Global MIP approach.Source: prepared by the authors.

7 Marginal project—project with the highest capital investments and operatingcosts per tonne of capacity.

P. Alexander et al. / Resources Policy 38 (2013) 558–567 565

supply–demand balance has been already incorporated intoincentive price of BIG-4’s marginal project. Whereas BIG-4 willutilize its market power to alter the market conditions in a waythat will deliver expected returns on invested capital.

However, it is worth mentioning one of the risks associatedwith this approach. It is well known, that the commodity ownerhas two options, either to sell the product or to postpone its saleand keep underground inventories (Cynthia Lin and Wagner,2007; Hotelling, 1931; Kronenberg, 2008). Therefore, Vale'smarginal project could be an option which would be developedonly if the market situation is favorable. This is quite possiblegiven that Vale has invested only US$3 billion or less than 20% oftotal capex (Hanano, 2013). If this is the case, LT price would bedowngraded to $115/t which is an incentive price of BHP-B's RGP-5 expansion.

Comparison of long-term iron ore price forecasts

Answering the question which was raised at the beginning ofthe paper, the LT iron ore price in real 2011 USD terms woulddecline from the current level of circa $150/t in any of thepresented scenarios, however, with different attitude (Fig. 11).One can see that consensus is 20–30% lower than our incentiveprice estimates suggesting that forecasts included in the con-sensus either do not include deposit depletion or do not takeinto account the required return on investments. This is verylikely to be true because usually analysts do not explicitlymention this.

Our forecasts obtained by Global MIP and BIG-4 MIP are identicalequaling $113 and 124/t, respectively. But BIG-4 MIP approach seemsto be the most straightforward compared to the other ones.

These LT price forecasts would not hold true if demand collapses,e.g. China experiences hard landing, and such turnarounds have been

seen in the history (Cashin et al., 1999). To make it more concise, if inthe next 10 years China experiences a 10% reduction of steelproduction compared to 2012 level, LT price would sit on thelow marginal costs level (roughly $63/t). But from the current statea probability of such an event looks insignificant (Deverell andGarvey, 2012).

Presented forecasts are in real 2011 USD terms and so theyexclude USD inflation and industry operating costs escalation (wagegrowth, local currencies appreciation, etc.). However, in many casesprices in nominal USD terms are required to valuate iron ore projects.In order to convert LT prices from real to nominal USD terms a 6%annual deflator is recommended, which is 2.5–3 times higher thanUS CPI inflation in the next decade. The deflator consists of halvedhistorical average inflation of CAPEX and OPEX plus US CPI inflation.Industry deflator is halved based on the assumption that iron oredemand growth would moderate in the future.

LT price in nominal USD terms in 2022, with the industry inflationfactored in, would range from $150 to 220/t. This could exceed thepeak price of $174/t in nominal USD terms reached in 2011.

Conclusions

Three approaches for forecasting of long-term real price of ironore were studied and applied which give almost identical forecasts:

� Marginal costs (LT price¼$85/t), which is mostly useful inanticipating future lows of price during periods of cyclicalmetals and mining market weakness because it does notconsider that new projects need to return invested capital.

� Global marginal incentive price (LT price¼$113/t), which iswidely used in the industry and based on the notion that newiron ore projects need to cover not only operating costs but alsocapital investments.

Table 7Global MIP long-term iron ore price probability distribution.Source: prepared by the authors.

Fig. 10. Incentive prices for marginal iron ore projects of BIG-4.Source: company data, prepared by the authors.

Fig. 11. Comparison of LT iron ore price forecasts.

P. Alexander et al. / Resources Policy 38 (2013) 558–567566

� BIG-4 marginal incentive price (LT price¼$124/t), which is themost straightforward approach as it implicitly uses marketassumptions incorporated in investment decisions of BIG-4.

Based on Monte-Carlo, confidence interval for future iron oreprice ranges from $85 to $95/t with 92% probability in MCapproach and from $115 to 125/t with 76% probability in GlobalMIP approach.

It is claimed that the future LT iron ore price in real 2011 USDterms will fall from the 2012’s level of circa $130/t, however, not asmuch as consensus suggests. Apparently, this is due to the factpresented approaches take into account deposit depletion asopposed to the approaches of most of the other analysts whichdo not.

Using a 6% industry deflator, LT price could escalate to $150–220/t by 2022 compared to 2011’s peak of $174/t. So probably theactual peak of nominal iron ore prices is yet to come.

In the longer-term after 10 years iron ore prices could very wellbe supported at a high level if India and Africa follow a projectedpace of economic development.

Acknowledgments

The authors appreciate helpful contribution and advises ofColin Hamilton of Macquarie Capital Ltd.

References

Antunes, R., 2012. Iron ore: Exploring Hard Rocks and Logistics. Flow EquityResearch, 14 Mar: 1-9.

Berry, G., Cooper, S., 2011. Metal Magnifier: Suspicious Mind. Barclays Capital EquityResearch, 24 Aug: 1-65.

Brooks, L., 2011. Long-term outlook for iron ore. In: CRU Presentation for EU Ironore Conference, 10 Oct.

Bussitil, M., 2011. West Africa: The New Iron Ore Province. J.P.Morgan AustraliaEquity Research, 09 Dec: 1-85.

Cashin, P., McDermott, J., Scott, A., 1999. Booms and Slumps in World CommodityPrices. IMF Working Paper No. 99/155. Available at: ⟨http://www.imf.org/external/pubs/ft/wp/1999/wp99155.pdf⟩.

Correa, L., Grimaldi, P., 2012. China notes: iron ore—positive on the cycle and near-term pricing. Barclays Capital Equity Research, 3–4. (5 Mar).

Cynthia Lin, C.Y., Wagner, G., 2007. Steady-state growth in a Hotelling model ofresource extraction. Journal of Environmental Economics and Management54 (1), 68–83.

da Silva Neto, Alfredo Lopes, 1993. The international effects of mining projects. Thecase of Carajas iron ore. Resources Policy 1993, 124–130. (June).

De Angele, R.R., 2011. Iron Ore: Fundamentals in favor of suppliers, raising priceforecasts across the curve. J.P.Morgan Latin America Equity Research: pp. 1–13.

Deverell, R., Garvey, M., 2012. Have We Reached Peak Steel Demand in China? WeThink Not. Credit Suisse Fixed Income Research, 08 Mar.

Gonzales, I.H., Kaminski, J., 2011. The Iron and Steel Industry: A Global MarketPerspective. Gospodarka Surowcami Mineralnymi, 1–27.

Hanano, Fawzi, 2013. As market surplus looms, which projects will make the cut?Goldman Sachs, 1–119. (19 Mar).

Heimlich, E., 2008. Long-term copper price projection. Cochiclo. Available at:⟨http://www.cochilco.cl/english/pdf/Long-Term_Copper_Price_Projection.pdf⟩.

Hotelling, H., 1931. The economics of exhaustible resources. The Journal of PoliticalEconomy 39 (2), 137–175.

Hui, H., Xi-huai, Y., 2010. A game theory analysis of international iron oremanufacturers strategic alliance. In: Control and Decision Conference—CCDC’09: pp. 3535–3539.

Hynes, D., Jansen, H., 2011. Iron Ore: 2014 and Beyond. Citi Commodities Strategy5 Apr: 21-22.

Jeffrey, D.Wilson, 2012. Chinese resource security policies and the restructuring ofthe Asia-Pacific iron ore market. Resources Policy 37, 331–339.

Jones, Arlene, 1986. Prospects for an iron ore cartel. Resources Policy 1986, 103–115.(June).

Kalos, Malvin H., Whitlock, Paula A., 2008. Monte Carlo Methods. Wiley-VCH, pp.1–215.

Kronenberg, T., 2008. Should we worry about the failure of the Hotelling rule?Journal of Economic Surveys 22 (4), 774–793.

Lennon, J., Hamilton, C., 2011. Vale looks to the future and highlights the challenges.Macquarie Commodities Research 30, 1–5. (Nov).

Malanichev, A., Pustov, A., 2011. Econometric modelling of the global iron oremarket. Foreign Economic Bulletin 8, 55–65.

Mining Boom or Bust, 2013. McKinsey for Mining Indaba 04 Feb. pp. 1–22.Price, T., Garran, J., 2012. Old Numbers, New Bearish Spin. UBS Investment Research,

21 Mar: 1-6.Saltelli, A., Ratto, M., Andres, T., Campolongo, F., Cariboni, J., Gatelli, D., Saisana, M.,

Tarantola, S, 2008. Global Sensitivity Analysis. John Wiley & Sons, The Primer.Sukagawa, Paul, 2010. Is iron ore priced as a commodity? Past and current

practices. Resources Policy 35, 54–63.Tonks, A., Blanch, F., 2011. Iron Ore: Prices up, Volume Down. Merrill Lynch Bulk

Strategist 07 Jun: 3-6.We can't reveal mining tax receipts TheAge.Com.Au. Available at: ⟨http://news.

theage.com.au/breaking-news-national/we-cant-reveal-mining-tax-receipts-wong-20130115-2cqkg.html⟩.

World steel in figures, 2011. Worldsteel Association. Available at: ⟨http://www.worldsteel.org/dms/internetDocumentList/bookshop/WSIF_2011/document/World%20Steel%20in%20Figures%202011.pdf⟩.

Zhu, Zhirui, 2012. Identifying Supply and Demand Elasticities of Iron Ore. Thesis atDuke University. pp. 1–57.

P. Alexander et al. / Resources Policy 38 (2013) 558–567 567