You are on page 1of 9

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/226815638

Forecast of global steel prices

Article  in  Studies on Russian Economic Development · May 2011


DOI: 10.1134/S1075700711030105

CITATIONS READS

11 3,729

2 authors, including:

Alexander Malanichev
Rosneft
15 PUBLICATIONS   110 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Decline curve analysis View project

All content following this page was uploaded by Alexander Malanichev on 14 May 2019.

The user has requested enhancement of the downloaded file.


ISSN 10757007, Studies on Russian Economic Development, 2011, Vol. 22, No. 3, pp. 304–311. © Pleiades Publishing, Ltd., 2011.
Original Russian Text © A.G. Malanichev, P.V. Vorob’ev, 2011.

RUBRIKA
RUBRIKA

Forecast of Global Steel Prices


A. G. Malanichev and P. V. Vorob’ev
Severstal JSC
Received

Abstract—The article presents a regression model for forcasting global average annual steel prices. Instead of
price factors, we used the coefficient of load capacity of world steel plants and prices for iron ore and coking
coal. On the basis of these factors, steel prices were predicted for 2010–2012.
DOI: 10.1134/S1075700711030105

Economic forecasts rarely come true with a high a situation when experts frequently substitute forecasts
degree of accuracy, however, this does not lead to the by wishful thinking, i.e., instead of forecasting what is
eradication of an interest in such forecasts. We can see going to be, they express their anticipations. In addi
an analogy with weather forecasts, which are frequently tion, a considerable amount of time is spent on the
incorrect, but they do not express such a strong devia coordination of different trends (the growth rate of steel
tion from the real weather as other forecast methods. consumption, steel prices, raw material prices, etc.).
Economic forecasts restrict the future ambiguity and Formal methods allow for the elimination of the above
thus promote more confident longterm development mentioned subjective judgement and additional labor
strategies and investments. If one is to choose between a costs through the identification of regularities in the
forecast and its absence, the former is explicitly prefer object to be forecast through the construction of eco
ential because it allows for the reduction of losses from nomic models.
probable wrong decisions, taking advantage of opening In the process of price forecasting (in virtually every
opportunities or escaping potential threats. industry), we can distinguish the following approaches
Longterm price forecasts have a special signifi to model construction:
cance. Longterm investment decision making in any (1) decomposition of price time series: detection of
branch, including ferrous metallurgy, is based on the the trend, autocorrelation, seasonal and random com
understanding of price trends during the implementa ponents;
tion and payoff phases of a project. In the ferrous met (2) factorial analysis based on a regression model:
allurgy industry, the time slot can exceed 10 years. It is allocation of essential price factors, quantitative assess
vital that the trends of final product prices be compara ment of their impact on the regression, exogenous fac
ble with the expected cost dynamics. It allows for more tors, and price forecast construction.
reliable forecasts of producers’ margins and, conse Frequently, the relationship between the dependent
quently. of investment payback periods. and independent variables in a regression analysis
The object of the analysis in this article are long reminds a “black box”, i.e., mechanisms that determine
term price tendencies in the global rolled iron market. the presence of a relationship among variables are not
The purpose was to propose a forecast procedure for always obvious to analysts. In order to thoroughly
world steel prices that would include compatible trends understand the actual processes, we need to develop
of steel prices and costs. more sophisticated models that can be called structural.
As a result, we have developed a mediumterm fore Structural models imitate reality at an abstraction
cast algorithm for the average world steel price, which is level, which not only allows for observing relationships
based on a regression between metalroll prices, global between variables, but also answering the question why
steel capacities and the dynamics of steel production these relationships exist. Typically, such structural mod
costs. The derived regression model provides a good els in the form of mathematical relationships or systems
description of the historical dynamics of world steel of theoretical principles are the initial stage of an analy
prices since 1980, including the drop in prices during sis. After this, we can formulate the hypotheses and
the crisis of 2009. The simplicity of the model allows for regression model.
use in the calculation of price targets for strategic busi Metallurgical market modeling usually occurs on
ness plans in metallurgical companies. the basis of the first two approaches. A brief overview of
Approaches to price forecasting in the steel industry. the work carried out in this area follows.
The lack of popular intuitive prediction methods among A time series analysis is based on the future reitera
practitioners in the metallurgical industry [1] results in tion of the past, and the most important information for

304
FORECAST OF GLOBAL STEEL PRICES 305

Price, USD/t
Demand Dt

Marginal producer
p*
costs
Supply
Maximum capacity

Qt World capacity, mln t


Utilization capacity Idle capacity

Fig. 1. Determination of the equilibrium rolled steel price in the world market.

forecasting is already in historical price series. There have grown as a result of multiple price increases for
fore, this method is not widely used for forecasting into metallurgical raw materials, which happen due to infra
the very distant future, as structural changes can take structural constraints imposed by suppliers, especially
place, and new price growth factors or noncyclic varia Australia. The consolidation of raw materials suppliers
tions in old factors can emerge. Nevertheless, this in the three largest global companies (Rio Tinto, BHP
method is limited in respect of forecasting periods Billiton, and CVRD) also played an important role in
(1–6 months). An example is the work of A. Mel the price growth for raw materials.
nichenko [2]. Steel production costs are widely used in practice as
The selection of significant factors affecting prices a price factor (see, e.g., [1, 6]). An increase in steel
and assessment of their quantitative impact are the base prices is equal to the increase in steel production costs
of a factor analysis. A forecast is constructed by specify (campanile, coking coal, scrap, etc.). This approach is
ing the most probable factor dynamics in price changes intuitively understandable, because it actually repeats
and calculating forecast prices on the base of a regres one of pricing methods (costs +) but does not give any
sion equation. idea about the margin dynamics of steel producers,
For rolled iron, literature considers the following depending on economic cycles.
price factors: The model presented in this paper takes into
(1) export prices used for predicting domestic mar account both factors of steel price formation (world
ket prices [3]; steel capacity determined by market demand and sup
(2) rolled iron supply to the market [3]; ply and production costs).
(4) production volumes in consuming industries [3, Equilibrium model of demand and supply in the rolled
4]; iron market. The pricing model in the world market of
(5) macroeconomic variables, such as the amount of rolled steel is based on the hypothesis that producers’
money [3]; marginal expanses (production prices) at current mar
ket prices for a given amount of production are equal to
(6) world capacity [5]; consumption [7]. Marginal producers have the largest
(7) production costs [1, 6]. expanses of all producers, and their aggregate output
The work of O. Kononova [5] is an example of a provides for the global steel demand. All the world steel
simultaneous application of time series and factor anal production capacity can be ranked in an increasing
yses. This work proposes to exploit the productive order in respect of production costs. As a result, we will
capacity utilization as a primary factor for steel prices. obtain a worldwide capacity distribution curve in
At the same time, autocorrelation in the price dynamics respect of steel production costs. At the same time, it
was taken into account with the help of the ARIMA represents the supply curve in the rolled steel market,
model. because it shows the minimum price required by manu
In spite of the high accuracy of retrospective data, facturers to produce a given amount of rolled steel.
there were some discrepancies between the forecast and In each period, the market has certain demand Dt
price dynamics of metallurgical raw materials, i.e., that must be equal to steel production volumes Qt in an
including costs. In the past few years, world steel prices equilibrium state.

STUDIES ON RUSSIAN ECONOMIC DEVELOPMENT Vol. 22 No. 3 2011


306 MALANICHEV, VOROB’EV

As Fig. 1 shows, the equilibrium price P* is deter Contract prices for raw materials (coking coal and
mined by the intersection of the capacity utilization fines) are a good approximation of the cost dynamics
(demand) and steel production cost (supply) curves. for marginal steel producers. The components of the
This intersection point also represents a marginal pro Australia–Japan freight are not considered separately,
ducer’s costs for whom it is still profitable to produce at since they correlate well with raw material prices. For
such a low price. the same reason, the model does not include such an
The equilibrium price can also vary because of shifts important component of metallurgical raw materials as
in demand (changes in steel consumption) and supply iron scrap.
(changes in production capacity and steel production The raw materials cost index, which is measured at
costs) curves. the current U.S. dollar exchange rate and, therefore,
The demand curve can shift to the left or right, includes an inflationary trend, takes into account dollar
depending on the dynamics of the global economy and inflation.
the major steelconsuming industries—construction, Thus, its equation is as follows:
engineering, and pipe industry. P t = αCU t + βRMC t + γt + ω, (2)
The supply curve can change its shape:
(1) stretch horizontally as a result of an increase or where t is the time trend variable (as well as the time
decrease in steel production capacities (Capacitiest); index in the case of other variables), which was 0 in
2009 and increases or decreases by 1 in each subse
(2) shift vertically due to changes in prices of major quent year; α, β, γ, and ω are regression coefficients,
raw materials for ferrous metallurgy—ore, coking coal, which are estimated on the basis of historical data
scrap, etc.; (1980–2009).
(3) stretch vertically as a result of nonuniform The average global steel price Pt is calculated as an
changes in producer costs over different segments of the average of four hotrolled coil prices under the terms of
cost curve and changes in commodity prices due to the FOB in the United States, Europe, China, and exports
lack of raw material sources (the right section of the cost from Russia (sources are the analytical agencies
curve) or their presence (the left section). As a rule, the 1 2
latter are verticallyintegrated companies, which are SBB and CRU .
secured by own raw materials. Therefore, the right side The raw material cost factor is evaluated as a
of the cost curve becomes steeper quicker than the one weighted sum of coking coal and iron ore prices:
when market prices for raw materials increase. As the
described effect is quite difficult to evaluate, it is not RMC t = 0.6 × Coking coal price t
(3)
considered within the framework of the model. + 1.6 × Iron ore price t .
In order to reconcile the shifting of the supply and
demand curves, it is advisable to turn to a dimensionless The weights are determined on the basis of expert
parameter—capacity utilization CUt. In an equilibrium assessments of raw material consumption per t of hot
3
state, at time t, it is simply the following relationship: rolled sheets (WSD , 2010). The price of Australian
Dt Qt exports is chosen as a representative price for hard cok
CU t = 
 = 
. (1) ing coal, whereas the price of Brazilian exports is cho
Capacities t Capacities t sen as a representative price of fines (in all instances, on
Thereby, the two main factors of the equilibrium the terms of FOB). Data sources are CRU and WSD.
steel price are the capacity utilization rate and steel pro The capacity coefficient is estimated on the basis of
duction costs for marginal producers. data of the World Steel Association [8].
Regression model of rolled steel prices. In order to Figure 2 shows the initial data used for the assess
test the hypothesis that there are two major factors of ment of the regression model.
steel price dynamics, a regression model was built. It is Estimated formula (2) based on annual data from
based on the following assumptions. 1980 to 2009 is as follows:
There is a linear relationship between the price of P t = 13.45CU t + 3RMC t – 5.2t – 972. (4)
rolled iron Pt, the load coefficient of the world steel pro
duction capacity CUt, and raw material costs RMCt. All of the model coefficients are significant. The
Other factors affecting the price (development of determination coefficient R2 is equal to 90%; i.e., the
steelmaking technologies, transfer of steel production model cannot only explain 10% of changes in average
to developing countries, etc.) are recorded by means of world steel prices. The calculated values are close to
time trends. actual prices, i.e., this allows for the realization of pre
dicted estimates by 67%.
Marginal steel producers are located in the Asian
region (Japan and China), which have the greatest 1 SBB—Steel Business Briefing Ltd.
weight in world steel consumption and purchases raw 2 CRU—Commodities Research Unit.
3 WSD—World Steel Dynamics.
materials in the global free market.

STUDIES ON RUSSIAN ECONOMIC DEVELOPMENT Vol. 22 No. 3 2011


FORECAST OF GLOBAL STEEL PRICES 307

USD/t % USD/t
1000 90 1000
900 900
85
800 800
80 700
700
600 75 600
500 70 500
400 65 400
300 300
60
200 200
100 55 100
0 50 0

2003
2005
2007
2009
1993
1995
1983
1985

1997
1999
1987
1989

2001
1991
1981

2003
2005
2007
2009
1983
1985
1987
1989

1993
1995

2001
1997
1999
1981

1991
Year Year
Fig. 2. Baseline data for the model: raw material costs
(fines and coking coal); global average prices of hot Fig. 3. Average world hotrolled prices: 䉫 actual price com
rolled coil; ⎯ capacity utilization (right scale). parison; 䉭 estimates based on regression.

Let us provide an interpretation for the coefficient in uted to the decrease in world steel prices. Also, in 2004,
the model. A 1% growth in capacity utilization leads to the regression underestimated the actual increase in
an increase in rolled stock prices by 13.45 US dollars per steel prices. However, in 2004, speculative factors served
t. Respectively, steel price increase by 100 US dollars their purpose, which resulted in a further price correc
when capacity utilization increases by 7.5%. tion in 2005, when the gap between the actual and esti
mated costs substantially reduced.
An increase in raw material costs by 1 US dollar
leads to an increase in rolled stock prices by 3 US dol Analysis of the model sensitivity. This is a standard
lars, out of which 1 US dollar is directly related to the construction stage, which allows for the evaluation of
increased prices of coking coal and fines. The remain the price sensitivity to changes in the market environ
ing 2 US dollars are defined by rising prices for other ment (through changes in the exogenous parameters of
cost elements (scrap, ferroalloys, freight, energy, and the model). Establishing the range of possible changes
labor), which tend to vary in the same direction as the in the exogenous factors, it is possible to determine the
prices of coking coal and fines. liability of business risk to fluctuations in market activ
Every year, the time trend lowers prices by 5.2 US ity. On the base of the regression model, the sensitivity
dollars; this is due to regional industrial developments of steel prices to changes in the raw materials market (by
with low production costs and industrial technological the means of raw material costs) and demand (by the
advancements. means of capacity utilization) was analyzed (Table 1).
Figure 3 compares actual and calculated rolled stock If the world steelmaking capacities are loaded by
prices based on the regression. In general, the calcu 66% (i.e., there are about 500 mln t of excess steelmak
lated values accurately approximate to the actual prices. ing capacities), steel prices are sensitive to cost changes.
A discrepancy is observed in the dynamics during In this case, a 25% increase in raw material costs
1980–1985, when the strengthening US dollar contrib increases steel price from 520 to 670 USD/t, i.e., by

Table 1. Sensitivity analysis of prospective rolled steel prices to changes in capacity utilization and raw material costs (rel
ative to 2009)

Capacity Raw material costs (increase rate relative to 2009), USD/t


utilization, % –50% –25% 0% 25% 50% 100% 125%
668 110 261 411 561 711 1011 1161
664 164 315 465 615 765 1065 1215
66 218 365 520 670 819 1119 1269
66 + 5 286 436 586 736 886 1186 1336
66 + 10 353 504 654 804 954 1254 1404
66 + 15 421 571 721 871 1021 1321 1471
66 + 20 488 639 789 939 1089 1389 1539

STUDIES ON RUSSIAN ECONOMIC DEVELOPMENT Vol. 22 No. 3 2011


308 MALANICHEV, VOROB’EV

%
$ 450
$ 400
$ 350
$ 300
$ 250
$ 200
$ 150
$ 100
$ 50
$0
1981 1984 1987 1990 1993 1996 1999 2002 2005 2008
Year
Fig. 4. Relative contributions of the capacity utilization ( ) and raw material cost factor ( ) to steel price growth.

World capacity Rolled stock


Coal and iron ore Calculation of the raw utilization price prediction on
price predictions material cost factor prediction the basis of
regression (4).

Fig. 5. Algorithm of mediumterm price forcasting of rolled steel stocks.

28%. An increase in capacity utilization by 5% leads to Based on these, a consensus forecast is prepared for the
an increase in prices from 520 to 586 USD/t (13%). exogenous factors in the regression model (Table 2) and
Capacity utilization and the raw materials cost index after this, the price of hotrolled products is predicted
are equally significant in explaining the historical (Table 3).
dynamics of steel prices (Fig. 4). According to the derived prediction, in 2012, prices
However, we should note that prior to 2002, capacity for rolled stock will amount to 752 USD/t or increase by
utilization was more significant than the cost dynamics. 241 US dollars in comparison to 2009 (511 USD/t). At
This can be explained by the fact that commodity prices the same time, raw material costs will increase by
were relatively stable. 51 USD/t. According to the regression model, the fac
After 2002, the factor of raw material costs had a sig tors of raw material costs affect prices by a factor of 3,
nificant impact due to the rising raw material costs i.e., a growth of 51 US dollars leads to an increase in
owing to the increased imports of raw materials into steel prices by 153 USD/t. Exactly in this proportion
China. The 2009 world price reduction in the same pro prices increase in the consensus prediction (from 510
portion can be explained by a decrease in the utilization USD/t in 2009 to 660 USD/t in 2012).
of steel capacities and raw material prices. Thereby, the consensus prediction appears to be
Projected global prices for rolled iron in 2010–2013. based on the assumed cost dynamics of steel produc
Figure 5 shows a general prediction scheme based on tion. However, besides costs, steel prices will grow as a
the proposed regression model. result of a 7% increase in the steel load capacity (from
The development of a price forecast based on the 66% in 2009 to 73% in 2012). Even this more than con
proposed regression model requires an independent servative forecast of loading capacity provides an addi
prediction factor: a capacity utilization coefficient and tional increase in prices by 94.5 USD/t (7% × 13.5).
raw materials cost index (price dynamics of coking coal Consequently, by 2012, prices will increase by 153 USD/t
and fines). due to costs and by 95 USD/t due to increased utilization
The capacity utilization coefficient is calculated as a capacities (a general increase of 248 USD/t). As there is
ratio of world steel production volumes to the global a temporary decreasing trend, the final prediction of the
steelmaking capacity. Global investment banks and spe steel price increment constitutes 241 USD/t.
cialized metallurgical agencies regularly update their We can conclude that the consensus prediction of
predictions for steelmaking capacity, world steel pro external experts does not take into consideration the
duction volumes, and prices for coal and iron ore. factor of growth in the steelmaking capacity and is only

STUDIES ON RUSSIAN ECONOMIC DEVELOPMENT Vol. 22 No. 3 2011


FORECAST OF GLOBAL STEEL PRICES 309

Table 2. Consensus predictions for coking coal, iron ore and hot rolled steel prices

2009 2010 2011 2012


Producer Date Terms Region (assess
ment) rediction

Coking coal, USD/t


Troika Dialog 14.01.2010 FOB Australia 129 180 190 185
Citi 13.01.2010 FOB Australia 172 182 200 155
Macquarie 15.12.2009 FOB Australia 129 180 190 160
Goldman Sachs 11.01.2010 FOB Australia 129 180 – –
DB 18.01.2010 FOB Australia 129 175 190 190
Macquarie 15.12.2009 FOB Australia 129 180 190 160
UBS 03.02.2010 FOB Australia 129 200 180 160
Consensus prediction: hot x x x –54 36 4 –10
rolled sheet, increase rate, %
Fines, USD/t
WSD 23.12.2010 FOB Brazil 62 70 61 51
UBS 03.02.2010 FOB Australia 61 85 89 85
DB 18.01.2010 FOB Australia 60 81 73 73
Goldman Sachs 15.01.2010 FOB Australia 60 81 81 77
Merrill Lynch, USD/t fines 13.01.2010 FOB Australia 97 146 153 153
Citi, USD/t fines 13.01.2010 FOB Australia 109 108 112 112
CRU 01.01.2010 FOB Australia 61 112 – –
CRU 01.01.2010 FOB Brazil 56 104 – –
Macquarie 15.12.2009 FOB Australia 97 126 132 126
Consensus prediction: fines, x x x –30 29 –1 –4
increase rate, %
Hot roll, USD/t
UBS 03.02.2010 FOB Midwestern 552 606 606 716
United States
MS 29.01.2010 FOB United States 574 690 726
CRU 27.01.2010 FOB Midwestern 686 811 823 775
United States
DB 18.01.2010 FOB Europe 569 634 662 634
ML 14.01.2010 FOB Europe 497 660 730 790
AME 07.01.2010 FOB Latin America 468 468 520 570
Macquarie 15.12.2009 FOB Latin America 524 639 643 635
Renessance Capital 30.11.2009 FOB Russia 495 545 566 598
CityGroup 11.11.2009 FOB United States 535 638 700
Troika Dialog 15.10.2009 FOB United States 462 594 618 638
GS 08.10.2009 FOB United States 486 550 575 620
Consensus prediction: fines, x x x –44 18 5 4
increase rate, %

STUDIES ON RUSSIAN ECONOMIC DEVELOPMENT Vol. 22 No. 3 2011


310 MALANICHEV, VOROB’EV

Table 3. Global rolled steel price predictions for 2010–2012

2010 2011 2012


Attribute 2004 2005 2006 2007 2008 2009 (assess
ment) Prediction

World steelmaking capacity 86 84 86 85 78 66 69 70 73


coefficient, %
Steel Production, mln t 1069 1141 1243 1346 1336 1199 1320 1408 1494
% 89 95 104 112 111 100 110 117 125
Steelmaking capacity, mln t 1246 1356 1453 1583 1713 1816 1917 1997 2036
Raw material costs
Steel costs, USD/t 79 132 137 133 262 201 268 273 252
% 39 65 68 66 130 100 133 135 125
Fines costs (Brazil, FOB), 21 36 42 47 71 62 80 80 76
USD/t
% 34 57 68 75 114 100 129 129 123
Coking coal (Australia, 75 125 116 98 250 172 234 243 218
FOB), USD/t
% 44 73 68 57 145 100 136 142 127
Time trend –5 –4 –3 –2 –1 0 1 2 3
Average world price of hot
rolled sheet, USD/t
fact 601 540 589 596 866 511 – – –
consensus prediction esti – – – – 866 511 605 635 660
mation
on the base of regression 446 576 605 581 866 519 749 781 752
% 86 111 117 112 167 100 145 151 145
Note: Prices are nominal and average: the base year for interest accounting is 2009. The data source of steelmaking capacity and world steel
production is from World Steel Association and data sources of historical raw materials and steel prices is CRU.

based on rising raw material prices in the industry. The main advantage of the proposed model is that it
Therefore, the use of regression models provides a more allows for receiving rolled stock price predictions,
accurate prediction of prices. which agree with world economy predictions (demand
for steel), the expected commissioning of metallurgical
CONCLUSIONS facilities, and the price dynamics of metallurgical raw
materials. Another equally important advantage is the
The introduced regression model explains the simplicity of using the derived regression equation for
dynamics of average world hotrolled steel prices by two
exogenous factors: the world steelmaking capacity coef calculating economic forecasts and the strategic devel
ficient and raw material production costs of crude steel. opment plan of a major steel company in Excel.
The historical data analysis showed that during assess
ing the possible effect on steel price, the relative load
capacity and costs are not constant in time. During ACKNOWLEDGMENTS
1980–2000, the load capacity played an important part.
However, after 2000, the growth in crude steel prices We thank our colleagues T. Verasto, M. Hudalov,
relatively increased the influence of costs on the A. Pustov, and O. Kononova for valuable comments and
dynamics of steel prices. suggestions.

STUDIES ON RUSSIAN ECONOMIC DEVELOPMENT Vol. 22 No. 3 2011


FORECAST OF GLOBAL STEEL PRICES 311

REFERENCES 5. O. Kononova, “LongTerm Price Predictions in Metal


lurgy. A TenYear Price Prediction for HotRolled
1. WSD. The Chinese “Head Fake”. Chinese Ball Carrier to
Products,” National Metallurgy (2007).
Agile to Be Stopped, (Steel Scorecard, no. 4, 2010).
2. A. Melnichenko, “Metal Production Price Predic 6. V. Vlasjuk, Steel Industry after a Year of Crisis: The Way
tion,” Intellectual Resources (2005). Passed and the Outlook for Recovery (UPE Co.
Research&Consulting, 2009).
3. M. Denisenko, “Features of Price Predictions for Fer
rous Metallurgical Companies,” National Metallurgy 7. “Where Will Steel Prices Go in 2010?” UBS.
(2003). QSeries®: Global Steel (February 2, 2010).
4. A. Malanichev, “Regression Analysis of Nickel Prices,” 8. Global Steel Capacity Development, WSA. ECON
National Metallurgy (2005). Autumn Meeting, (September 17, 2009).

Malanichev Alexander: Ph.D., Head of Strategic Marketing


Vorobjov Pavel: PhD, senior analyst of Strategic Marketing

STUDIES ON RUSSIAN ECONOMIC DEVELOPMENT Vol. 22 No. 3 2011

View publication stats

You might also like