the chinese steel industryâ€™s transformation : structural change, performance and demand on...

The Chinese Steel Industry’s TransformationStructural Change, Performance and Demand on Resources

Edited by

Ligang Song

Associate Professor, Crawford School of Public Policy, Australian National University

Haimin Liu

Vice President, China Steel Industry Development Research Institute, Beijing

Edward ElgarCheltenham, UK • Northampton, MA, USA

M3021 - SONG 978184844658 PRINT.indd iiiM3021 - SONG 978184844658 PRINT.indd iii 23/11/2012 14:5123/11/2012 14:51

© Ligang Song and Haimin Liu 2012

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical or photocopying, recording, or otherwise without the prior permission of the publisher.

Published byEdward Elgar Publishing LimitedThe Lypiatts15 Lansdown RoadCheltenhamGlos GL50 2JAUK

Edward Elgar Publishing, Inc.William Pratt House9 Dewey CourtNorthamptonMassachusetts 01060USA

A catalogue record for this bookis available from the British Library

Library of Congress Control Number: 2012939094

ISBN 978 1 84844 658 8

Typeset by Servis Filmsetting Ltd, Stockport, CheshirePrinted and bound by MPG Books Group, UK

M3021 - SONG 978184844658 PRINT.indd ivM3021 - SONG 978184844658 PRINT.indd iv 23/11/2012 14:5123/11/2012 14:51

03

v

Contents

List of contributors vi

Foreword vii

Preface ix

1 Steel industry development and transformation in China: an

overview 1

Ligang Song and Haimin Liu

2 Metal intensity in comparative historical perspective: China,

North Asia and the United States 17

Huw McKay

3 Economic growth, regional disparities and core steel demand in

China 45

Jane Golley, Yu Sheng and Yuchun Zheng

4 China’s iron and steel industry performance: total factor

productivity and its determinants 69

Yu Sheng and Ligang Song

5 The technical efficiency of China’s large and medium iron and

steel enterprises: a firm- level analysis 89


6 The backward and forward linkages of the iron and steel

industry in China and their implications 106


7 China’s shift from being a net importer to a net exporter of steel

and its implications 129

Haimin Liu and Ligang Song

8 China’s iron ore import demand and its determinants: a time-

series analysis 145


9 Restructuring China’s steel industry and the implications for

energy use and the environment 162

Guoqing Dai and Ligang Song

Glossary 177

Index 179

M3021 - SONG 978184844658 PRINT.indd vM3021 - SONG 978184844658 PRINT.indd v 23/11/2012 14:5123/11/2012 14:51

vi

Contributors

Guoqing Dai Institute of Development Studies, Shoudu (Capital) Steel

Corporation, Beijing.

Jane Golley Australian Centre on China in the World, ANU College of

Asia and the Pacific, Australian National University, Canberra.

Haimin Liu China Steel Industry Development Research Institute,

Beijing.

Huw McKay Westpac and Australian National University, Sydney and

Canberra.

Yu Sheng Crawford School of Public Policy, ANU College of Asia and

the Pacific, Australian National University, Canberra.

Ligang Song Crawford School of Public Policy, ANU College of Asia

and the Pacific, Australian National University, Canberra.

Yuchun Zheng China Steel Industry Development Research Institute,

Beijing.

M3021 - SONG 978184844658 PRINT.indd viM3021 - SONG 978184844658 PRINT.indd vi 23/11/2012 14:5123/11/2012 14:51

vii

Foreword

Chinese economic reform and opening to the international economy since

the late 1970s have changed the country and the world. The developments

in the steel industry in the reform period are central to those changes, illu-

minative of them, and of immense significance in themselves. This book

throws new light on these historic changes for Chinese and foreign readers

alike.

Chinese civilization was the first to use many of the qualities of iron on

a significant scale. We learn in Chapter 1 that under the Northern Song

dynasty a thousand years ago, China was producing as much iron as

Europe on the eve of the industrial revolution in 1700. In steel- making as

in many things, China lost its head start in the second millennium. China

was not producing much more iron under the late Qing at the turn of the

twentieth century than it had been at the end of the first millennium, by

which time the domestic industry was tiny by modern standards.

The steel industry was an integral part of the industrialization of the

North Atlantic countries and later Japan as modern economic growth

took place from the late eighteenth through the nineteenth and early

twentieth centuries. China was not part of these transformational devel-

opments in the history of humanity until the second half of the twentieth

century. Even then, it endured a long detour under central planning as the

Communist Party established its rule from 1949 through the first three

decades after the revolution – the steel and heavy industry were favoured

by the authorities but still failed to prosper.

In the steel industry as in many parts of the Chinese economy, market-

oriented reform and integrating Chinese production into international

markets spurred productivity growth and the expansion of production.

Chinese steel production rose from 32 million tonnes at the commence-

ment of the reform era in 1978 to 128 million tonnes in 2000, and reached

630 million tonnes in 2010. The immense expansion during the reform

period was accompanied by much higher productivity, higher quality of

output, and much closer calibration of product quality to the requirements

of the market.

These developments in steel were important to Chinese economic

success in the reform era. They were also transformational for the world.

M3021 - SONG 978184844658 PRINT.indd viiM3021 - SONG 978184844658 PRINT.indd vii 23/11/2012 14:5123/11/2012 14:51

viii The Chinese steel industry’s transformation

The opening of Chinese production to higher- quality and more cost-

effective international supplies of the main steel- making raw materials put

immense pressure on global markets for iron ore and metallurgical coal.

This book tells the story of how all this happened. It is a solid case study

of an industry changing on a scale and at a pace that has no precedent

in global economic history. It will be a useful reference for those seeking

to understand the Chinese experience of economic reform, the impact of

Chinese economic growth on the global economy, and the future trajec-

tory of economic change in China. It will have useful points of reference

for those who specialize in industrial economics, resource economics,

and the economics of the transition out of central planning and inward-

looking policies. It will be of interest to people in the mining industry

who are seeking to understand the immense expansion in opportunities in

their own industry in recent times and especially in the early twenty- first

century. Finally, it should attract the attention of people who are simply

fascinated by the remarkable story of the world’s most populous country’s

belated and subsequent participation in modern economic growth.

Ross Garnaut

University of Melbourne, May 2012

M3021 - SONG 978184844658 PRINT.indd viiiM3021 - SONG 978184844658 PRINT.indd viii 23/11/2012 14:5123/11/2012 14:51

ix

Preface

This book is a product of an Australian Research Council (ARC) Linkage

Project (LP0775133) which has been conducted by the team from the

China Economy Program in the Crawford School of Public Policy at

the Australian National University in cooperation with Rio Tinto and

the China Steel Industry Development Research Institute attached to the

China Iron and Steel Association in Beijing.

When we proposed the study to the ARC for funding the project in 2006,

the world economy was experiencing an unprecedented demand shock to

the commodity markets resulting from the rapid growth of the Chinese

economy. As one of the pillar industries, the steel industry plays a key part

through its increasing demand on resources in driving the current resource

boom. The book provides a central reference work on the Chinese steel

industry. Included are both macroeconomic studies of developments in

Chinese resource demand with particular reference to the ferrous metals

complex, and microeconomic studies that utilize the comprehensive firm-

level data to evince new knowledge of both firm and industry performance

with respect to their productivity, technical efficiency, or and industrial

linkages. The book also discusses trade in steel products and the impact of

the restructuring of the industry on the environment.

In completing this work, we have received the support and assistance

from various people and institutions. We would like to thank first our

team members on the project from both Australian National University

and the Chinese steel industry. We gratefully acknowledge the financial

support for the project from the ARC and Rio Tinto. We thank the

China Steel Industry Development Research Institute for providing

some data which were used in carrying out some of the quantitative

analyses in the book. We also acknowledge the arrangement made by

the China Iron and Steel Association for us to visit Shoudu (Capital)

Steel Corporation in Beijing for conducting the firm- level interviews

and seeing the production processes in steel- making. In the course

of completing the project, we ran two workshops in Beijing and one

in Canberra at which the preliminary results and draft chapters were

presented and discussed respectively. We are very grateful to all the

participants in the workshops from both Australia and China for their

M3021 - SONG 978184844658 PRINT.indd ixM3021 - SONG 978184844658 PRINT.indd ix 23/11/2012 14:5123/11/2012 14:51

x The Chinese steel industry’s transformation

contributions to the discussion, which helped us to improve and com-

plete the project.

Finally, we would like to thank our publishers, Edward Elgar, for their

interest in publishing this work. Ms Bijun Wang, a visiting PhD student at

Crawford School from the China Centre for Economic Research at Peking

University, provided assistance with respect to formatting and referencing

the manuscript. We thank Bijun for her help in finalizing the manuscript.

Thanks also go to Mr Luke Meehan for providing assistance in editing

the introductory chapter and Dr Nicola Chandler for copy- editing the

manuscript.


Canberra and Beijing, May 2012

M3021 - SONG 978184844658 PRINT.indd xM3021 - SONG 978184844658 PRINT.indd x 23/11/2012 14:5123/11/2012 14:51

1

1. Steel industry development and transformation in China: an overview


THE DEVELOPMENT OF CHINA’S STEEL INDUSTRY1

The steel industry epitomises traditional industrialization. The major

economies of the United Kingdom, France, Germany, Japan, Korea

and the United States experienced stages of development where the steel

industry played a pivotal role in transforming their economies. The role of

the steel industry in this development is more than symbolic; the technol-

ogy and ready availability of the steel products enabled further economic

growth and development. Industries essential for industrialization and

modernization, such as machinery and building infrastructure, were able

to grow and expand.

China has a long history of iron and steel production. Hartwell (1962,

1966, 1967, cited by Findlay and O’Rourke, 2007) described the remark-

able expansion in Chinese iron and steel production during the Northern

Song dynasty (the period 960–1126 ce): ‘The scale of total production,

and of the levels of output and employment in individual plants, was far

in excess of anything attained by England in the eighteenth century, at the

time of the Industrial Revolution.’ Hartwell estimated that iron produc-

tion in China in 1078 was of the order of 150 000 tonnes annually:

The entire production of iron and steel in Europe in 1700 was not much above this, if at all. The growth rate of Chinese iron and steel production was no less remarkable, increasing 12- fold in the two centuries from 850 to 1050. (Findlay and O’Rourke, 2007, p. 65)

Iron produced during this time was used primarily for agricultural and

military purposes. A thousand years ago China was the largest iron pro-

ducer in the world, but for historical and institutional reasons the iron and

steel industries were not fully developed until centuries later.

M3021 - SONG 978184844658 PRINT.indd 1M3021 - SONG 978184844658 PRINT.indd 1 23/11/2012 14:5123/11/2012 14:51

2 The Chinese steel industry’s transformation

The development of China’s modern steel industry can be traced back

to the establishment of Hanyang Iron Works in 1890.2 In the following

58 years to 1948, China’s total accumulated pig iron output reached 22

million tonnes, and crude steel nearly 7 million tonnes. The highest indi-

vidual year was 1943, with iron production reaching 1.3 million tonnes

and steel 0.9 million tonnes. During this period, the steel industry was

located mainly in the Anshan area of North- East China, producing more

than 90 per cent of the country’s total steel output.

The wars which wracked the country for much of the 1940s almost

ruined the steel industry. When the People’s Republic of China (PRC) was

founded in 1949, the national total production of pig iron was only 250 000

tonnes. In the same year, the country’s production of steel was 158 000

tonnes, accounting for 0.2 per cent of the world’s total steel production

and ranking twenty- sixth in the world.

Yet production recovered quickly and by the end of 1952 the country

had restored and expanded 34 blast furnaces and 26 open hearths. The

national total production of iron, steel and rolled steel in 1952 was 1.9,

1.4 and 1.1 million tonnes, respectively, topping all previous records.

Meanwhile, the regional distribution of steel production showed no sig-

nificant changes, with 70 per cent being produced in the northeast, 23 per

cent in the east and north, and 7 per cent in the hinterland.

In the 30 years following the founding of the PRC, the steel industry was

regarded as a pivotal link for industrialization. With the help of the former

Soviet Union, a generally complete steel industry system was formed with

‘three big, five middle and 18 small’ steel enterprises,3 but this burgeoning

steel industry development faced further setbacks with the implementation

of the ‘Great Leap Forward’ and later the ‘Cultural Revolution’.

The highly centralized planned economic system hampered the develop-

ment of productive forces in the steel industry, albeit after having played

a major role in restoring production in the 1950s. Consequently, the

industry saw very slow technological progress. In 1978 China’s total steel

production was only 32 million tonnes, less than three weeks of current

output levels. The per capita steel production was merely 33 kg, a fifth of

the world average levels. The industry’s technology, equipment, product

variety and quality, as well as technical and economic indicators, all lagged

far behind developed countries. For example, when the world average

ratio of open- hearth steel- making to total steel- making fell below 20 per

cent in the late 1970s, China’s ratio still stood at 35.5 per cent. When the

ratio of continuous casting was more than 50 per cent in Japan and 30 per

cent in Europe, China’s was merely 3.5 per cent. As a result of obsolete

technologies, out of total production, the energy consumption per tonne

of steel was as high as 2.52 tonnes of standard coal, with the yield of crude


Overview of the Chinese steel industry 3

steel in rolling finished steel around 74 per cent.4 Furthermore, 28 per cent

of steel consumption relied on imports in 1978, costing foreign exchange

earnings.

The reform and opening- up policy of 1978 brought China into a new era

of growth and development. The development of the steel industry since

then can be divided broadly into three stages.

The first stage was the early period of reform and opening up, running

from 1978 to 1992. This stage is characterized as a gradual transition from

a highly centralized planned economy towards a preliminarily established

socialist market economy. Experiments on enterprise autonomy, profit

contracts and managerial responsibility systems were carried out in the

steel industry. Shoudu (Capital) Steel Corporation, the first batch of large

state enterprises experimenting with extended decision- making powers,

implemented the managerial responsibility system of contracting in 1981.

The new system brought firm and worker initiatives into play. As a result

the firm’s steel output and economic performance improved quickly.

Afterwards the contracted responsibility system spread step by step across

the industry. By the end of 1992, 103 out of 110 key steel enterprises had

implemented managerial responsibility system reforms.

During this reform stage, China changed from a rigid system of state-

fixed prices and centralized purchase and sales to allowing steel enterprises

to purchase raw materials in the market. It also allowed them to sell a

certain proportion of planned production, and all the excess steel prod-

ucts, through their own channels at market prices, which were usually

higher than planned prices. The country gradually lowered the ratio of

mandatory planned rolled steel, reaching 20 per cent in 1992. These meas-

ures boosted incentives for production in the industry.

These steel enterprises were allowed to use retained profits for their

expansion, bonuses and employee welfare payments. The industry’s

retained profits in 1992 reached 5.8 billion yuan, accounting for 56 per

cent of total profits. Of retained profits, 3.8 billion yuan was used for

enterprise development, providing 26 per cent of funds sourced from both

the government and enterprises for upgrades and renovation. The average

annual incomes for workers in the steel industry increased from less than

500 yuan in 1978 to around 3800 yuan in 1992.

Financing for investment in the industry was transformed from relying

heavily on state allocations before 1978 to relying on the enterprise itself

by self- raising, bank loans and foreign capital. At the same time steel

enterprises were permitted to make independent decisions and undertake

technical innovations. These reforms adjusted the power–responsibility–

favour relations between the state and enterprises. This made it clear that

the enterprises were the principal point of interest.



The steel industry also worked towards opening up. During the 14 years

from 1978 to 1992, more than 700 advanced technologies were introduced

and US$6 billion in foreign capital was utilized. In particular, two modern

large steel enterprises, Baoshan Iron and Steel Corporation (launched in

1978 and put into operation in 1985) and Tianjin Seamless Steel Tube

Corporation (launched in 1989 and put into operation in 1996), were

established. Meanwhile, many old steel plants were rebuilt and restruc-

tured. These notable changes to the technology structure of the country’s

steel industry saw the gap between it and world- class practices narrow.

This initial stage (1978–92) saw significant achievements in outputs. By

1992 there was a 1.6- fold increase in steel production; the domestic market

share had increased by 17 per cent, the ratio of open- hearth steel- making

to total steel- making was reduced to 11 per cent, the ratio of continuous

casting to the total rose to 30 per cent, and the total production energy

consumption per tonne of steel output fell to 1.6 tonnes of standard coal

or by 62 per cent.

Despite greater autonomy granted to enterprises under the contracted

responsibility system, China’s steel enterprises were still subordinate to

the government. Further, varying contractual conditions together with the

dual- track steel price system caused a disparity among steel enterprises in

terms of performance. This disparity induced some firms to bargain with

the government, distorting the market’s role in resource allocation.

The second stage was the early period of establishing a socialist market

economy from 1993 to 2000. In this stage, the main focus of China’s

reform was the setting up and improvement of market systems. Key to

this was establishing a complete modern enterprise system – separating the

roles of government as the owner and manager of state- owned enterprises

(SOEs), and making the enterprises the true market entities responsible for

their own profits and losses.

As for the steel industry, mandatory plans for production and sales were

abolished in 1993, and the dual- track steel price system ended. Thereafter,

steel enterprises made their own decisions on production and sales based

on market demand. The steel market developed rapidly in all parts of

China. With the development of the securities markets, transforming into

a joint- stock company and listing on the stock markets became the new

financing channel for a Chinese steel enterprise. By the end of 2000 there

were 27 steel enterprises listed in the domestic and/or international secu-

rities market. This raised significant investment funds for development,

and more importantly improved companies’ corporate governance and

management skills.

At the same time, the steel industry not only continued to utilize foreign

capital to upgrade obsolete technology but also utilized overseas resources



to make up for the domestic scarcity of raw materials. Total imports of

iron ore reached 70 million tonnes in 2000, increasing nearly eightfold

compared with 1978. Some enterprises began to buy or set up jointly

owned iron ore production bases in Peru and Australia.

During this period the steel industry faced many challenges, including

continuously declining steel prices, chain debts and the periodic return of

overcapacity. It also went through a difficult macroeconomic environ-

ment, with overheating just before the Asian financial crisis in 1997 and

then a fall in output in the aftermath. Nevertheless, the steel enterprises

streamlined their businesses, readjusted their product mix and carried

out technical innovations around energy savings and cost reductions. As

a result the industry’s technological bases and ability to adapt to market

changes improved greatly.

Along with the steel enterprises’ own efforts the Chinese government

offered them supporting policies, such as debt- to- equity swaps and dis-

counts for technological transformation. These policies helped China

become the world’s largest steel- producing country in 1996, with total

output surpassing 100 million tonnes. Its steel production in 2000 reached

128 million tonnes, an increase of 59 per cent from 1992.

This stage saw the fastest structural adjustment of the steel industry.

By the end of 2000, open- hearth steel- making was almost eliminated, five

years earlier than planned; the ratio of continuous casting reached 87 per

cent, surpassing the 75 per cent target and catching up with world aver-

ages; and the total energy consumption per tonne of steel output fell to 885

kg of standard coal, a decrease of 56 per cent from 1992.

The third stage has been the deepening of reform and fast economic

growth period since 2001. With the new century, the Chinese iron and steel

industries experienced significant and influential external developments.

Following China’s entry into the World Trade Organization (WTO) in

2001, market laws and regulations were geared towards reaching inter-

national standards, integrating the steel industry further into the world

market. China’s manufacturing share increased from about 5 per cent in

the mid 1990s to over 17 per cent of the world’s total manufacturing in

2009. Over the reform period, the urbanization ratio rose to 46 per cent

in 2010, rising from only 19 per cent back in 1978, transferring nearly 300

million people from rural to urban areas.5 This large- scale urbanization

boosted the investments in housing and infrastructure.6 All these devel-

opments led to the rapidly increasing demand for steel from domestic

sources. For example, steel consumption increased by 16 per cent per

annum from 2000 to 2010. In meeting this rising demand, the industry’s

total investment increased from 36.7 billion yuan in 2000 to 453.1 billion

yuan in 2010, with an annual growth rate reaching 28.5 per cent over this



period. Steel production rose as a result. According to the figures from the

statistical yearbooks, in 2010 the ferrous metal industry accounted for 4.6

per cent of the total industrial employment, 8.3 per cent of the total indus-

trial value added, 25 per cent of total industrial energy consumption and

between 10 and 16 per cent of the total emissions of the main pollutants

from the industry sector.

Further trade liberalization has led to the sharp reduction of import

duty as well as the complete abolition of quantitative import restrictions,

which has exposed steel enterprises to the fierce competition of the inter-

national market. China’s rapid economic growth led to rapidly increasing

demand for steel from domestic sources. The increased competition from

the market entry of those non- state firms has forced the large and medium

state- owned steel firms to deepen the corporate reform, to include share-

holding and the separation of government functions from management.

To further separate government functions from enterprise management,

the Bureau of Metallurgical Industry at both state and local level was dis-

solved. Instead, the China Iron and Steel Association,7 a self- regulatory

organization of the steel enterprises, acted as a bridge between enterprises

and government.

Steel enterprise reform proceeded towards developing a more diver-

sified ownership structure. By the end of 2010 more than 50 steel

enterprises were listed on stock markets and 50 per cent of large and

medium- sized steel enterprises, in terms of operating revenue, were trans-

formed into joint- stock companies. Private steel enterprises also grew

rapidly. Non- state enterprises accounted for about 45 per cent of the

total output of the steel industry in 2010. Reorganization and mergers

and acquisitions (M&As) have also been part of the process of industrial

agglomeration.

The steel industry is accelerating its pace of globalization. The China

Iron and Steel Association and the largest steel enterprises became

members of the World Steel Association (WSA) at the end of 2004. They

have taken part in worldwide dialogue and negotiations, and adopted

common actions as a response to resource, environmental and market

changes. The rapid expansion of steel production has forced the industry

to utilize overseas resources on an unprecedented level. Imported iron ore

now accounts for two- thirds of the total consumption in the steel industry.

For example, to produce 567 million tonnes of steel in 2009, China’s steel

industry consumed 850 million tonnes of iron ore, of which 602 million

tonnes were imported in that year, raising its import dependence ratio for

iron ore to 74 per cent. The share of China’s consumption of iron ore in

world total iron ore consumption increased from 20 per cent in 2000 to

56 per cent in 2009. Many steel enterprises are also undertaking outward



direct investment in the mining sectors in order to secure stable and long-

term resource supplies (Song et al., 2011).

ACHIEVEMENTS IN THE REFORM PERIOD

Any shortage of steel in China may now be consigned to history. Since

the reform and opening up of 1978, and especially since 2000, China’s

steel production capacity has expanded rapidly. The industry underwent a

period of extraordinary growth in both total sales and total profits which

increased at an average annual rate of 32 and 44 per cent respectively over

the period 2001–07.8 The end of 2010 saw China’s total steel production

reach 630 million tonnes, 18 times the output in 1978. The crude steel pro-

duction grew at an annual growth rate of 17.2 per cent after 2001. China’s

share of global steel production increased from 4.4 per cent in 1978 to 15

per cent in 2000 and to 45 per cent in 2010, a share which has been unpre-

cedented in the entire history of industrialization.9

In the past, China relied on imported steel to fill the supply shortfall.

Gross imported billet and rolled steel in the period from 1978 to 2004

amounted to 478 million tonnes. After deducting exports, net imports

were 352 million tonnes, accounting for 12.6 per cent of China’s total con-

sumption of crude steel. Increasing exports and decreasing imports of steel

products found China realizing a rough balance in 2005, becoming a net

exporter of steel products in 2006. Such an historic change implies China’s

steel industry is capable of meeting the needs of the country’s economic

development. It also suggests that the international competitiveness of

Chinese steel products has improved immensely.

Iron and steel production quality and variety have increased dramati-

cally. Currently China’s self- sufficiency rate in most steel products exceeds

100 per cent. Only some high- value- added products, such as cold- rolled

ordinary steel board (strip) and electric steel, are net imported. Most steel

products used in industry – such as machinery, automobiles, shipbuild-

ing, home appliances, oil, electricity and railways – are home- made. The

product qualities are sufficient to meet the basic needs of those industries.

Some varieties have even reached internationally advanced levels. China’s

steel exports have gradually shifted from producing long products to pro-

ducing higher- value- added sheets and pipe products.

The industry has also achieved enhanced standards in terms of technol-

ogy and equipment, and an increased localization rate. The accumulated

fixed- asset investments of the steel industry, which were a mere 60 billion

yuan in the first 30 years from 1949, reached 2.6 trillion yuan from 1978 to

2010. In addition to the establishment of world- advanced steel enterprises



– such as Baoshan Iron and Steel Corporation and Tianjin Seamless

Steel Tube Corporation, and some private steel enterprises – most of

those investments went to the upgrading of outdated equipment and the

restructuring of old steel enterprises. From 1978 to 2010 the number of

large blast furnaces over 1000 m3 in volume grew from 10 to 260, of which

28 were over 3000 m3; the ratio of continuous casting grew from 3.5 per

cent to 98 per cent, which is above the world average. The modern steel

industry is encouraged to rely more on autonomous innovation rather

than depend solely on introduced techniques and equipment. In 2010

small and medium metallurgical equipment is domestically produced,

while the localization rate of large metallurgical equipment is over 90

per cent.

The industry also experienced a remarkable rate of technological

progress, resulting in improved technical and economic indicators. Many

indicators of domestic productivity are outstripping those of developed

countries. For example, since 1978 the overall ratio of rolling steel being

produced has increased to over 95 per cent from 75 per cent; total pro-

duction energy consumption per tonne of steel has fallen from 2.5 tonnes

of standard coal to 605kg of standard coal; freshwater consumption per

tonne of steel has fallen to 4 tonnes; and labour productivity per tonne per

person- year has increased from 33 tonnes to 400 tonnes.10

NEW CHALLENGES AND READJUSTMENT

The market- oriented industry, corporate reform and opening- up policy

have been the decisive factors in the development of China’s steel indus-

try. Enterprises were released from the rigid centralized planning system,

boosting competitiveness (enhanced in large part by the low cost of

labour) and allowing the development of profit- making incentives, leading

to enhanced performance. The establishment and development of the

market system enabled and urged steel enterprises to face the challenges

of market competition, which again improved their productivity and effic-

iency. China’s rapid economic growth provided a huge demand for steel

products, which gave impetus to the rapid growth and expansion of the

industry.

Despite these achievements, China’s steel industry still faces many chal-

lenges which demand deepened reform and consolidation. The state his-

torically has dominated the steel industry. The transformation of SOEs in

the past turned many steel enterprises into market players. However, they

are still constrained by the traditional state- dominant system in orienting

development strategies, making investment decisions, conducting M&As,



restructuring, appointing senior managers and employing workers. As a

result the industry’s overall economic performance remains behind devel-

oped countries, by some margin. Private steel enterprises, although more

flexible, require further improvement in implementing modern technol-

ogies, following codes of conduct and upgrading management skills

according to market principles.

Market competition is the catalyst for improving the overall quality

of the steel industry, but the way competition has worked in it has been

complicated by the cyclical fluctuations of the macroeconomy. In times

of prosperity, steel enterprises have tended to assess the market prospects

overoptimistically and expand production blindly. This has resulted in

large amounts of overinvestment and backward production capacity

being utilized. In times of weak demand, disorderly competition by cutting

prices has occurred, and the industry has sometimes relied upon govern-

ment intervention to alter the supply–demand balance. These behaviours

and fluctuations have added to structural adjustment costs, slowed down

technological progress and wasted social resources.

The domestic market is still segmented and the degree of industrial

concentration is quite low. In 2000, the share of steel output by the top

ten firms and the top four in total output were 49 and 32 per cent, respec-

tively. The years to 2006 saw a falling ratio of industry concentration, to

35 per cent for the top ten and 19 per cent for the top four, owing to the

large number of small firms entering the market seeking to meet the rising

domestic demand for steel. The benefits of industrial consolidation in

responding to the problems associated with the use of materials, energy

and the environment thus led to the ratio of industry concentration rising

again, in 2010 increasing to 49 per cent for the top ten and 28 per cent for

the top four (the latter is still below the level of 2000). Despite the progress

made, the industry concentration ratio is far below that of developed

countries, which ranges between 70 to 80 per cent for the top four or five.11

The rapid increase in demand for steel products and the rising prof-

itability of the industry stimulated the entry of many non- state small

firms, usually supported by local governments for the purposes of

increasing local employment and taxation. These small firms tend to

use backward production capacities and technologies, adding further

difficulties to restructuring the industry. This is the root cause of the

problems associated with capital misallocation, low quality standards,

duplication of construction effort and blind expansion of production

capacity, as well as structural overcapacity. These problems are intrinsi-

cally related to issues of wasteful investment, inefficiency in material use

(including energy, water and electricity) and environmental problems.

Such industrial segmentation also hampers the technological progress



as smaller firms lack the resources for research and innovation. The

industry needs further structural reform to address these problems at

the microeconomic and industrial levels, and the government needs to

do its part by strengthening the existing regulatory system with respect

to market entry and the environment, and reforming its relationship

with enterprises.

The industry faces the pressure of rising costs of production resulting

from the high prices of energy, water and iron ore in addition to the rising

costs of labour and transport on which the industry heavily depends.12

These rising costs of production have further squeezed the profit margin

for the industry. When the industry passes on the price rises to the con-

sumers, it affects future demand for steel. To cope with this, the strategy

for the industry needs to be shifted from an emphasis on pure expansion

of scale to a focus on optimization of the structure of production includ-

ing the product structure through industrial upgrading and technological

change. The industry is also compelled to reduce the costs of produc-

tion, increase productivity and international competitiveness through, for

example, an increase in industrial research and development (R&D) and

improved corporate management. The introduction of advanced foreign

technologies, equipment, capital and resources has also helped the indus-

try to realize a leapfrogging developmental path.

An offsetting factor which helps the industry to reduce resource intensi-

ties, including primarily the use of iron ore in producing steel in the future,

is that there will be an increasing proportion of steel demand which is met

by scrap. China is still at the phase of industrialization where the accumu-

lated stock of steel is not sufficiently large for more scrap to be recovered

and used in steel- making. In 2008, the proportion of electric furnaces

using scrap for making steel was only 9 per cent of total steel production

in China, which was far below the world average level of 31 per cent. In the

same year, the proportion in the United States was 58 per cent while the

proportion in the European Union (15 countries) was above 40 per cent

(Yang, 2010).13

China paid an excessive environmental price for the rapid develop-

ment of its industries, including the steel industry, with an environmental

ramification well beyond its border. China became the largest global

carbon emitter in 2007,14 and yet the country is still in the middle phase

of industrialization (according to the current level of per capita income)

with the growth and expansion of the manufacturing sector (especially

heavy industries) generating more emissions. China needs, and has an

obligation to achieve, emission reduction targets as part of the global

effort in confronting the challenge of climate change. The government

needs to be clear about the scale, pattern and pace of growth, which will



meet China’s future demand for steel while ensuring that the industry’s

development is conducive to environmental protection. At the moment,

the government’s macrocontrol policies and regulatory measures curb the

development of large enterprises, but leave the small ones and low- level

projects unaffected. This leads to a high proportion of backward and

low- level production capacities being utilized in the industry. China has

to rely on exporting to absorb the surplus of steel after meeting domestic

demand.

The share of steel exports in total world steel production has experienced

both rising and falling trends in recent decades. In 1975 the share was 23

per cent, then rose to a peak of 40 per cent in 2000. It fell to 34 per cent in

2008 and further to 26 per cent in 2009 (World Steel Association, 2010).15

In contrast to this trend, China has been a net exporter of steel since

2005. In 2008, Chinese net exports were 40.7 million tonnes of steel, and

ranked number one in the world, followed by those of Japan (32.4 million

tonnes), Ukraine (26 million tonnes) and Russia (23 million tonnes). In

the same year, the United States was the world’s largest net importer of

steel (12.7 million tonnes) followed by the European Union (27 coun-

tries) with 11.4 million tonnes, United Arab Emirates (10 million tonnes),

Thailand (9.4 million tonnes) and South Korea (8.8 million tonnes).16

Exporting steel products to world markets helps ease the problem of

industrial overcapacity. However, an increase in exports of steel has made

industrial restructuring (including ownership reform, industrial concen-

tration and technological progress) a less urgent task. It has also made

the tasks of reducing the resource and pollution intensities of the industry

more difficult. Furthermore, China’s exports of steel are causing trade

frictions with others, especially those to developed countries such as the

United States and the European Union. The government has adopted

various measures such as the imposition of export taxes and the reduction

of export tax rebates for certain products in order to limit the increase in

exports of steel. However, the industry’s low cost and other advantages

will continue to run their course despite the fact that the government

intends to see the role of the steel industry as essentially to meet domestic

demand. The challenge therefore is how the Chinese government could

bring steel production back in line with the changes in domestic demand

without relying too much on exports.17

China will continue to be the largest steel producer in the world for

the time being, driven largely by the ongoing process of urbanization,

industrialization and her integration with the global economy. China’s

level of per capita income needs to be tripled from the current level before

the peak level of metal intensity is attained, something which is forecast

to happen around 2024. By then, China’s total steel output will be in the



vicinity of 1 billion tonnes (McKay et al., 2010). This prospect of China’s

future metal intensity and the magnitude of its output raises an important

question as to how the world supplies of key resources including energy

and minerals, as well as the environment, will accommodate the continual

growth in China. As Garnaut has said (2012), ‘one only has to identify

the possibility of China absorbing more resource- based products than

the currently developed world to raise some fundamental questions about

“limit to growth”’. The steel industry can do its part in overcoming this

limit to growth in the process of China’s modernization as the industry is

scale- , capital- , resource- and pollution- intensive. In fact, the industry will

be compelled to do so because in recent years the Chinese government has

promulgated a number of key laws and regulations with respect to energy

use and the environment such as the ‘Environmental Protection Law’, the

‘Law for Prevention of Air Pollution’, the ‘Law for Prevention of Water

Pollution’, the ‘Law for Prevention of Solid Waste Pollution’ and the ‘Law

for Energy Saving’. Given the current level of the industry development, it

is a challenging task for the industry to comply fully with the requirements

of these laws.18

The world economy has entered a period of development requiring huge

adjustment and rebalancing. Resource scarcity, demographic change,

climate change and global imbalances are global shared concerns. The

Chinese government is responding to these changes by transforming the

model of its growth and development (Song, 2010). Accordingly, the

requirements for the steel industry have also changed, as is reflected in

a lower level of resource intensity, the higher variety and quality of steel

products, and an increasing environmental constraint. These changes call

for optimizing the industrial structure, enhancing technological progress,

improving corporate management, and, most fundamentally and cru-

cially, deepening the structural reform of the steel industry, including its

ownership and concentration.

STRUCTURE OF THE BOOK

The aim of this volume is to provide a central reference work on the

Chinese steel industry. The chapters fall loosely into three groups. The

first group, comprising Chapters 2 and 3, are macroeconomic studies of

developments in Chinese resource demand with particular reference to

the ferrous metals complex. Chapter 2, by Huw McKay, utilizes an inter-

national comparative framework with a strong historical bent. McKay

argues that while China’s experience with metal intensity currently resem-

bles that of Korea, this is a temporary phenomenon. China’s eventual



path is expected to borrow from both the US and Japan, but will retain

Sino- specific characteristics.

Chapter 3, by Jane Golley, Yu Sheng and Yuchun Zheng, is an attempt

to apply the logic of provincial convergence to the metal intensity field.

The authors showcase a novel approach to the estimation of underly-

ing steel consumption by province. The inferences emanating from this

subnational approach make a fascinating counterpoint to the discussion

in Chapter 2. Readers who come to the study of China with a belief that

its industrialization path is sui generis will find much to commend in the

provincial approach adopted here.

The second group, comprising Chapters 4, 5 and 6, are microeconomic

studies utilizing granular data to evince new knowledge of both firm and

industry behaviour. All three are co- authored by Yu Sheng and Ligang

Song. Based upon the unique findings presented, a number of policy

recommendations are put forward in this cluster of chapters.

Chapters 4 and 5 should be considered as a pair. Utilizing firm- level

data, the authors investigate productivity outcomes of all steel firms in

the structurally significant period of 2000–03, and efficiency outcomes of

state- owned enterprises (SOEs) in the period 1999–2005. This era was an

immensely important time for the industry. Coming out of the turmoil of

the 1990s, with the twin shocks of the 1992–94 boom–bust cycle and the

Asian financial crisis, and then being subjected to a further disruption in

the form of the ‘tech wreck’ recession, the industry was also confronted

with an imperative requirement for major structural adjustment and a dra-

matic transformation of ownership. The situation was clearly extremely

fluid. Understanding the industry at this time is crucial to making sense of

developments later in the decade.

Sheng and Song show that these various stresses encouraged a number

of firms to change their behaviours resulting in both level and aggregate

gains in both productivity and efficiency. In addition to rigorously docu-

menting these trends, the authors add to our knowledge by splitting their

sample between large and medium firms and their smaller counterparts, as

well as discussing the nature of firm ownership. It turns out that the deter-

minants of productivity are very different when the size of the firm is con-

sidered, a finding that brings with it powerful implications for industrial

policy both inside China and in other developing and/or transition econo-

mies. Along the way, Sheng and Song are able to make some methodologi-

cal improvements to the techniques utilized in previous literature, and are

thus able to correct the prior tendency to understate the contribution of

capital to output – which gives profound implications for the analysis of

returns to scale in the industry.

Chapter 6 studies the backward and forward linkages of the steel



industry, utilizing industry- wide census data. The results indicate strongly

that steel is vital to the very fabric of the Chinese economy. After esti-

mating productivity spillovers from steel to other industries, the authors

conclude that upstream and downstream industries have seen opposite

effects of the increase in steel industry efficiency. Downstream firms

have been shown to improve their productivity as a response to the steel

industry, but upstream firms have suffered. The authors conclude that the

continual increase in import penetration in upstream sectors accounts for

this result.

The third group, comprising Chapters 7, 8 and 9, offer three fresh prac-

tical perspectives on the industry. Chapter 7, by Haimin Liu and Ligang

Song, details the nature of China’s international trade in ferrous metals

and points out that the net export status achieved by the industry in the

lead- up to the financial crisis is neither sustainable nor desirable. Liu and

Song highlight the difficulties for the Chinese government in bringing steel

production back into line with domestic demand, and suggest the ways

forward to align the balance between demand and supply of steel products

without relying excessively on exports.

Chapter 8, by Yu Sheng and Ligang Song, focuses on the determinants

of the iron ore trade. The authors consider time- series data from 1960,

capturing both the autarkic and more open eras of Chinese industrializ-

ation. Their conclusion – that domestic demand for ferrous metals is the

principal determinant of China’s burgeoning imports needs – should

not be controversial, given China’s long- running net import position.

Additionally, the study highlights that the relatively low quality of China’s

own iron ore reserves, coupled with its very strong demand and lack of

scrap resources, leads to a position where import demand is inelastic to

price. That result may embolden iron ore negotiators who sit on the supply

side of discussions.

Chapter 9 by Guoqing Dai and Ligang Song argues that while the

steel industry has already achieved a great deal in terms of reducing its

environmental footprint, greater efforts are required in moving forward.

At the national level the steel industry is a very prominent consumer

of energy and a large emitter of pollutants and waste water. Therefore,

progress in improving the steel industry’s own environmental perform-

ance through enhanced technological progress, economies of scale

and corporate management will contribute strongly to the aggregate

outcome. Put another way, if the steel industry is unable to improve

its performance, it will be difficult for the country as whole to meet its

aspirational goals.



NOTES

1. The data presented in this chapter are taken mainly from the Chinese statistical yearbooks and China steel industry statistics supplemented by a special report on the steel industry development in China prepared by China Iron and Steel Association in 2008.

2. Hanyang Iron Works was established in 1890 and went into operation in 1894. It was the first integrated iron and steel works in modern China and was also one of the largest in Asia, with an annual output of 60 000 tonnes of steel.

3. Three big: Anshan, Wuhan and Baotou Iron and Steel Company; five middle: Taiyuan, Chongqing, Beijing Shijingshan, Maanshan and Xiangtan steelworks; 18 small: Handan, Jinan, Linfen, Xinyu, Nanjing, Liuzhou, Guangzhou, Sanming, Hefei, Jiangyou, Wulumuqi, Hangzhou, Echeng, Lianyuan, Anyang, Lanzhou, Guiyang and Tonghua steelworks.

4. The ratio increased to 94 per cent in 2010. 5. The urbanization ratio is defined as the ratio of urban population to total population. 6. According to the data from China Iron and Steel Association (CISA), the housing

sector consumed more than 50 per cent of steel produced in recent years. 7. The CISA is a national steel industry organization. The members consist mainly of steel

production enterprises, which account for 80 per cent of the national total steel output. Some trading firms, equipment manufacturers, construction firms as well as consulting companies are also members of the CISA.

8. The profit rate from sales grew by an average of 9.1 per cent per annum over the same period.

9. For a historical comparison, the United Kingdom was the largest steel producer in the world before the 1890s. In 1885, the UK’s steel output accounted for about 30 per cent of the world total steel output That top position was then taken by the United States from 1886 to 1971, and then the former Soviet Union from 1971 to the late 1980s, and Japan for only a brief period in the early 1990s (Yang, 2010).

10. Chapters 4 and 5 in this volume detail the causes of these improvements in performance. 11. For example, Japan’s top five firms produce 79 per cent of the total steel output;

Korea’s top two firms produce 80 per cent of its total output (Yang, 2010).12. World iron ore prices (the long- term contract prices) rose by 8.9 per cent in 2003, 18.6

per cent in 2004, 71.5 per cent in 2005, 19 per cent in 2006, and 9.5 per cent in 2007. In 2008, the prices rose by 65 per cent for Brazilian ore and 79.8 per cent for Australian (CISA report, 2008).

13. The world average proportions of electric furnaces in steel- making were gradually increasing over time, rising from 14 per cent in 1970 to 22 per cent in 1980, then further to 28 per cent in 1990 and to more than 30 per cent in 2006 (CISA report, 2008).

14. An estimate by the World Steel Association shows that China’s steel industry was ranked number one in terms of its carbon emissions among all the steel industries in the world in 2007. China’s emission share accounted for about 51 per cent of the total emis-sions emitted by world steel industries in 2007 followed by the European Union (12 per cent), Japan (8 per cent), Russia (7 per cent), the United States (5 per cent) and others (17 per cent) (CISA report, 2008).

15. The quick fall in the share of exports of steel in total production in 2009 over the pre-vious year may be due largely to the impact of the global financial crisis (GFC).

16. World Steel Association (2010).17. See Chapter 5 for a detailed discussion of this issue.18. The International Iron and Steel Industry Association (IISI), at a meeting held in

Berlin, Germany, in October 2007, published the statistics on its members’ CO2 emissions. IISI’s 180 members have agreed on the plan for reducing CO2 emissions. According to the data, only 20 per cent of the steel production in China could meet the requirements set by IISI in 2006 (CISA, 2008).



REFERENCES

China Iron and Steel Association (2008), ‘On the path of restructuring the Chinese steel industry’, Beijing, July.

Findlay, R. and K.H. O’Rourke (2007), Power and Plenty: Trade, War, and the World Economy in the Second Millennium, Princeton and Oxford: Princeton University Press.

Garnaut, R. (2012), ‘Australia’s China resources boom’, Australian Journal of Agricultural and Resource Economics, 56 (2), 222–43.

Hartwell, R. (1962), ‘A revolution in the Chinese iron and coal industries during the Northern Sung, 960–1126 ad’, Journal of Asian Studies, 21 (2), 153–62.

Hartwell, R. (1966), ‘Markets, technology, and the structure of enterprise in the development of the eleventh- century Chinese iron and steel industry’, Journal of Economic History, 26 (1), 29–58.

Hartwell, R. (1967), ‘A cycle of economic change in imperial China: coal and iron in northeast China, 750–1350’, Journal of the Economic and Social History of the Orient, 10 (7), 102–59.

McKay, H., Y. Sheng and L. Song (2010), ‘China’s metal intensity in comparative perspective’, in R. Garnaut, J. Golley and L. Song (eds), China: The Next Twenty Years of Reform and Development, Canberra: Australian National University E- Press, and Washington, DC: Brookings Institution Press, pp. 73–98.

Song, L. (2010), ‘China’s rapid growth and development: an historical and interna-tional context’, paper prepared for the 34th PAFTAD Conference on China in the World Economy, Peking University, Beijing, 7–9 December.

Song, L., J. Yang and Y. Zhang (2011), ‘State- owned enterprises’ outward invest-ment and the structural reform in China’, China and World Economy, 19 (4), 38–53.

World Steel Association (2010), World Steel in Figures 2010, Brussels: World Steel Association.

Yang, L. (2010), Studies on the Sustainability of China’s Steel Industry under the Constraints of Iron Ore Resources, Beijing: Metallurgical Industry Press.


17

2. Metal intensity in comparative historical perspective: China, North Asia and the United States

Huw McKay

INTRODUCTION

The aim of this chapter is to shed light on China’s future path of steel

and base metal intensity by referencing the experience of relevant peers

through their point of entry into the global strategic transition (Snooks,

1998) – which is the vehicle by which industrialization has been dissemi-

nated round the world – and beyond.

Will China follow a path like Korea, which has stayed in the metal-

intensive sweet spot for a sustained period of time; or will it touch only

briefly on the middle- income sweet spot of metal intensity en route to the

current resting place of the European economies and their offshoots? Will

it eventually sit just on the more metal- intensive side of the high- income

cohort, in a place similar to that where Japan resides?

These questions go to the very roots of Chinese long- run economic

strategy and performance. The immense scale of China’s megastate means

that its strategic choices will generate substantial externalities that will

require assertive responses from others. A better understanding of the

path of metal intensity through time in a broader range of countries would

be a great help to those tackling the immense task of meeting and respond-

ing to China’s long- run metal demands.

A broad conclusion of the analysis is that China is unlikely to follow the

Korean path once it moves deeper into middle- income status. The superfic-

ially attractive correlation between the Korean and Chinese paths, based

on limited time series, is perceived sceptically from the medium- term point

of view. The final path is likely to borrow from certain aspects of the expe-

rience of the United States and Japan, but the Chinese path will be distinc-

tive. The United States is an apposite comparison as an economy built

on a continental scale with a low population agglomeration ratio, while

Japan is relevant due to the explosive but finite gains in world export share



it enjoyed as its technological strategy unfolded, and the constraints that

high population densities place on urban lifestyles. Together, these econ-

omies define some reasonable parameters for considering the Chinese case.

While the United States and Japan are clearly relevant to a medium-

term assessment, evidence on metal intensity from these jurisdictions

should not be deterministic for our purposes. The sketch of Chinese metal

intensity assembled here assumes a strong increase until the period of 2015

to 2020, then a flattish peak emerging from 2020 to 2025, before an orderly

decline develops in the second half of the 2020s. These dates are informed

by Sino- specific analysis with pragmatic, selective input from historical

case studies.

The national relationships between metal intensity and standard macro-

economic variables are complex and idiosyncratic. The firm implication

from them is that each nation’s industrialization process and its relation-

ship to metal intensity is sui generis and should be treated as such. Any

attempt to generalize across this field should be carried out with extreme

caution. While the investigation of intranational (provincial) data on

metal intensity has much to recommend it, at this stage the results derived

should be regarded as tentative rather than conclusive.

METAL INTENSITY IN THE UNITED STATES: THE CASE FOR THE ‘KUZNETS CURVE FOR STEEL’

The longest national time series available in this field is for steel use per

capita in the United States. Here we analyse the US experience in the

context of the Kuznets formulation.

A Kuznets relationship is represented by a second- order polynomial,

with income per capita and its square term on the right- hand side of the

equation and the relevant development metric on the left. To test for this

relationship in the metal field, we regress the Hodrick–Prescott filter of

annual steel use per capita in the United States from 1929 to 2002 against

the natural logs of the aforementioned right- hand- side variables. For the

relationship to be robust the estimated coefficients need to be opposite in

sign and statistically significant.

The empirical evidence in favour of a Kuznets relationship in the field

of long- run steel intensity in the United States is strong. Not only are the

coefficients correctly signed and significant at the 1 per cent level, but also

the adjusted fit of the model is surprisingly high. We conduct the same test

with the raw data and get broadly similar results.

The results for all tests are presented in Appendix Tables 2.A1 and 2.A2

at the end of this chapter and here in Figures 2.1(a) and 2.1(b). For the


Metal intensity in comparative historical perspective 19

twentieth century United States at least, a Kuznets curve for steel (KCS)

seems to exist.

The existence of a KCS is consistent with the inferences of the synthe-

sis view of metal intensity put forward in the early 1990s. Prior to this

time, the literature on metal usage was divided into two distinct schools.

They were the consumer preference school that pioneered intensity of

0.1

0.2

0.3

0.4

0.5

0.1

0.2

0.3

0.4

0.5

1929

t/capitat/capita

1939 1949 1959 1969 1979 1989 1999

1929 1939 1949 1959 1969 1979 1989 19990.0

0.1

0.2

0.3

0.4

0.5

0.6

0.0

0.1

0.2

0.3

0.4

0.5

0.6t/capitat/capita

Fitted estimate, +/–1 std errorUS steel intensity, unadjusted annual data

Fitted estimate, +/–1 std errorUS steel intensity, annual HP filter

b

a

Sources: Steel data from US Geological Survey, various years. Population data from US Historical Abstract, various years; author’s calculations.

Figure 2.1 Kuznets curves of US steel intensity



use (IU) analysis, and the leapfrogging school. The consumer preference

school argued that IU (defined as the volume of metal consumed per unit

of output) increased in low income economies over time as demand for

durable goods created derived demand for metals (International Iron and

Steel Institute, 1972; Malenbaum, 1973, 1975). In this view of the world,

as economies transition towards developed status, the consumption basket

shifts progressively towards services such as health, education and recrea-

tion, at the expense of the eventually saturated metal- intensive durables

goods market. Thus the development of consumer preferences with rising

incomes creates an inverted U- shaped IU curve with a definable turning

point.

The leapfrogging school argued that the ability of low- income econ-

omies to skip whole generations of technologies gave a downward bias

to IU over time (Hwang and Tilton, 1990). Essentially, the leapfrogging

school argued that a low- income economy’s ability to import technology

could transplant it to the same point on the hypothesized IU schedule

as an advanced economy; or alternatively, they were able to navigate to

lower IU schedules relative to those that previous generations of industrial

countries had inhabited at equivalent income levels. The implication was

that a low- income economy was just as likely to see a decline in its IU as it

moved towards middle- income status, rather than see the rise assumed by

the consumer preferences school.

A synthesis was achieved by the work of Lohani and Tilton (1993).

They argued that there was partial truth in the teachings of both schools

that could be reconciled in a single theory by a relatively simple empirical

test. Building on the implications of Hwang and Tilton (1990), Lohani

and Tilton studied changes in the IU of a cross- section of low- income

economies between 1977 and 1988 to test both the extant theories and

the viability of a synthesis view. Their hypothesis was that IU in the low-

income economies was related linearly to both income per capita (change

in purchasing power and consumption patterns) and time (change in the

technological frontier). If the extreme version of the leapfrogging school

was correct then the coefficients derived from their cross- sectional regres-

sion should have been zero for income per capita, and negatively signed

for the time trend. If the extreme version of the consumer preferences

school was correct then the coefficients should have been zero for the time

trend and positively signed for income per capita. The synthesis would

see a positively signed coefficient for income per capita and a negatively

signed coefficient for the time trend.

The result was that the synthesis view carried the day. More specifically,

the authors indicated that a real income growth rate of approximately 1

per cent per annum was required to keep IU stable against the underlying



gravity of the leapfrogging phenomenon. Therefore, low- income countries

achieving strong rates of economic growth will see rising IU, but those

that are stagnating will see IU fall due to technological change, or others

factors that are captured by the time trend. The synthesis view is thus a

‘proto- KCS’.

A CAUTIONARY TALE FROM THE ARCHIVES

The finding that steel intensity in the United States during the twen-

tieth century has followed an upside- down U- shape indicates that the

Kuznets framework may be applicable to the entire metal intensity field.

Unfortunately, not all relevant countries offer time series as long as those

of the United States. That leaves many scholars to rely on cross- sections,

as Lohani and Tilton did in establishing the synthesis view. The cross-

sectional data offers corroborating, if tentative, evidence for the national

KCS of the United States. It is very tempting to use these apparent rela-

tionships to define a generalized path of metal intensity through the indus-

trialization process.

While tempting, the validity of such a methodology is highly debatable.

Kuznets’s original observation of the upside- down U in income per capita

and income distribution space (Kuznets, 1955), which became known as

the Kuznets curve, is actually an egregious example of cross- sectional

bias. Much as with our data on metal intensity, Kuznets had a patchy time

series of US income distribution (plus the United Kingdom and Germany/

Prussia/Saxony) and a cross- section of information from a few countries

at a spread of lower income levels.1 These economies provided the hump

in his hypothesized curve, ‘corroborating’ the patchy time- series evidence.

Subsequent experience of East Asian trajectories following their entry

into modern economic growth in the quarter- century following the Second

World War, where inequality was reduced between the low- and middle-

income stages of development, has shown that the Latin American and

South Asian paths observed by Kuznets are idiosyncratic rather than

general. Indeed, the Latin trajectory was an outgrowth of poor planning

decisions that ignored comparative advantage in favour of import substi-

tution (Lin, 2008). The dual impediments of caste and colonial overlord-

ship, which are redistributive strategies rather than surplus- enhancing

ones, seem sufficient to comprehend the South Asian case. These two

models encouraged the supernormal growth of a rent- seeking elite, with

predictable outcomes for income distribution. Therefore, the original

Kuznets curve is a cautionary tale for scholars of development looking to

cross- sectional data for predictive relationships.



The hump in a steel intensity cross- section is provided by two medium-

sized middle- income North Asian economies – Korea and Taiwan prov-

ince. They are both relatively new entrants to industrialization, with their

engagement occurring within the last half- century. Are these economies

typical or atypical? This judgement could validate or invalidate the cross-

section as representative, as we do not have readily available alternatives

to substitute into the middle- income space. Most development economists

would choose the latter taxonomy (atypical) if they were judging the case

globally and the former (typical) if assessing the case regionally. This is a

vexed issue we will revisit in different contexts in this chapter.

To be fair to Kuznets, who was the consummate empirical economist of

his generation, he was

acutely conscious of the meagreness of the reliable information presented. The paper is perhaps 5 per cent empirical information and 95 per cent speculation, some of it possibly tainted by wishful thinking. The excuse for building an elaborate structure on such a shaky foundation is a deep interest in the subject and a wish to share it. The formal and no less genuine excuse is that the subject is central to much of economic analysis and thinking; that our knowledge of it is inadequate; that a more cogent view of the whole field may help channel our interests and work in intellectually profitable directions; that speculation is an effective way of presenting a broad view of the field; and that as long as it is recognized as a collection of hunches calling for further investigation rather than a set of fully tested conclusions, little harm and much good may result. (Kuznets 1955, p. 26)

It is easy to agree and to empathize with many of the sentiments

expressed in this disclaimer. Indeed, Kuznets might have been putting the

case for a wider research agenda on metal intensity in the current day,

such is the overlap between his case and ours. However, the final assertion

is somewhat problematic when the task at hand is a practical forecasting

project that might guide real- world decision- making.

A mixed example of the practical application of the Kuznets framework

is the innovative extension of the hypothesis into the environmental field

(Grossman and Kruger, 1995). This has been a lucrative and knowledge-

enhancing application. This branch includes a burgeoning literature on the

applicability of the ‘environmental Kuznets curve’ (Cai and Du, 2008; Bao et

al., 2008; Bao and Peng, 2006). Yet the debate has also highlighted the limita-

tions of the framework as a generalizable forecasting system, with national

deviations from the central model common and only certain measures of pol-

lution behaving in accordance with theory. That has not stopped some schol-

ars from adopting the framework as a crude argument in favour of a passive

approach to negative environmental externalities. It is extremely important

that the caveats presented here are recognized and understood.



AN ASIDE ON AUTOMOBILE PENETRATION: WHY TIME SERIES MATTER

While cross- sectional metal intensity data of recent vintage is available for

a wide variety of countries at various states of engagement with the strate-

gic transition, using these data alone has the potential to be highly mislead-

ing. A focus on comparative time- series data is crucial to define the path

from the genesis of industrialization to peak metal intensity to a mature

state. Cross- sections can help with this task, but they can easily mislead

rather than guide. The difficulty is that while a cross- section may be able

to define states of nature at the initiation and maturity of the process, the

path between these two states may be hidden, and the peak identified erro-

neously, if the sample is imperfect.

An instructive example of the potential problems associated with

cross- sectional data comes from the field of automobile penetration.

Contemporary data on the number of passenger cars per 1000 persons for

a wide variety of countries are readily available. In addition, there are time

series for the United States (from 1929) and Japan (from 1960). Ignoring

the two time series for a moment, the cross- sectional data implies that a

simple linear association exists between income per capita and automobile

penetration, with a clustering of observations in the lower left and upper

right corners of the chart space (Figure 2.2a).

The time series tell a far richer story (Figure 2.2(b)). They indicate that

the path between low and high automobile penetration can be dramati-

cally different. Once again, we find that a reliance on relationships inferred

solely from a cross- section would lead to damaging forecasting errors.

Japan is highly urbanized, densely populated and without a domestic oil

resource. The United States is moderately urbanized, reasonably sparsely

populated and it controls a great deal of oil. The Japanese were quite

rational to follow the path they did and the Americans likewise.

Here again, we are struck by the diversity of national experiences, rather

than their similarity. The ability to ascribe any economy as ‘typical’ seems

very limited. Once again, we find little guidance on the transition between

stages, with scant evidence from middle- income locales. Forecasters operat-

ing in the automobile penetration field have realized this, and are projecting

a non- linear, concave path for China (International Monetary Fund, 2005).

While the ultimate peak level of automobile penetration is certainly contest-

able given the environmental and congestion issues that come with height-

ened automobile use, these issues will be no more pronounced in Chinese

cities than in the ever denser populations of Hong Kong, Singapore, Korea

and Japan. Indeed, the levels of automobile penetration in Hong Kong and

Singapore might be seen as a lower boundary for the Chinese case.



There is a very practical reason why we should be interested in the time

series on automobile penetration besides the aforementioned analogical

and methodological considerations. The evidence for a Kuznetsian rela-

tionship in the historical steel intensity of the United States is comfort-

ably sufficient when couched in the simplest of forms. However, when

Sources: Japanese Statistician, World Bank, IMF, US Historical Abstract.

0

10

20

30

40

0

Passenger cars per 1000 people

GDP per capita

China

Italy

Australia

Spain

UK

Korea

Hong Kong

Singapore

Germany

Sweden

Canada

Belgium

France

Argentina

Russia

Malaysia

Mexico

South Africa

Thailand

BrazilIndonesia

US

Japan

50 100 150 200 250 300 350 400 450 500 550

a

0

10

20

30

40

Passenger vehicles per 1000 people

GDP per capita

Japan from 1960

USA from 1929

0 50 100 150 200 250 300 350 400 450 500 550

b

Sources: Japanese Statistician; US Historical Abstract.

Figure 2.2 Auto penetration and income – a linear relationship?



data on automobile penetration are included in the test (in original form

and as a squared term) as additional information on the stage of devel-

opment, the model improves on all metrics by a non- trivial margin. Most

simply, the adjusted R2 (the ‘fit’) rises from 0.63 to 0.71 and the standard

error declines. (For estimation output, refer to Table 2.A3 in the appen-

dix.) Thus we can conclude that automobile penetration is an important

measure for assessing both the general stage of economic development and

the likely metal intensity of the economy.

These conclusions add to the complexity of the original task. The pres-

ence of a Kuznets relationship in the long- run steel intensity of the United

States offered some hope that a generalizable relationship between metal

demand and development level could be defined. The presence of a rela-

tionship between metal intensity and automobile penetration raises hopes

of an indirect test of the homogeneity of national paths, but as two major

countries have followed such strikingly disparate paths regarding auto-

mobile penetration, we cannot credibly infer or back- cast Japan’s metal

intensity path using the United States example.

OBSERVATIONS ON THE NORTH ASIAN PEER GROUP

The backbone of the comparative analysis of metal intensity within North

Asia is shown in Table 2.1. China, Japan and Korea are its subject. Per

capita consumption of steel, aluminium and copper (sourced from the

International Monetary Fund (IMF); units are kilograms) during their

respective take- off phases are benchmarked against macroeconomic vari-

ables that a priori are expected to have a relationship with metal intensity.

The macroeconomic variables chosen for the table are: gross domestic

product (GDP) per capita (under both purchasing power parity (PPP) and

market exchange rate weights); industrial value added (IVA) as a percent-

age of GDP; urban population as a share of total population; exports

of goods and services as a percentage of GDP; and trade (exports plus

imports) as a percentage of GDP; merchandise exports; percentage of

world exports; gross fixed capital formation (GFCF) as a percentage of

GDP; and gross savings as a percentage of GDP.

Keeping all of the caveats from the preceding sections of the chapter in

mind, the following points have emerged from an investigation of the data

available. They are all relevant to constructing an educated, if tentative,

guess about China’s future trajectory.

IVA tends to peak as a share of GDP between income levels of $US10 000

and 15 000 per capita in PPP terms, and then decline. That accords with



Table 2.1 Relevant metrics for metal intensity analysis (kg)

Years

from

take- off

GDP ppp

index

Steel use

index

Copper

use index

Alum. use

index

IVA

index

Savings

index

China

1978

0 100 100 100 100 100 100

5 135 89 100 99 93 94

10 223 109 134 100 92 98

15 322 202 225 191 97 113

20 504 200 305 329 96 110

25 729 408 647 679 95 116

Years

from

take- off

GDP ppp

US$/

capita

Steel use

kg/capita

Copper

use

kg/capita

Alum. use

kg/capita

IVA

%GDP

Savings

%GDP

China

1978

0 679 44 0.4 0.6 48.2 37.6

5 919 39 0.4 0.6 44.6 35.4

10 1516 48 0.5 0.6 44.1 36.8

15 2187 90 0.8 1.1 46.6 42.4

20 3423 89 1.1 1.9 46.2 41.4

25 4951 181 2.4 4.0 46.0 43.4

Years

from

take- off

GDP ppp

index

Steel use

index

Copper

use index

Alum. use

index

IVA

index

Savings

index

Japan

1960

0 100 n.a. 100 100 n.a. 100

5 149 n.a. 134 189 n.a. 100

10 243 n.a. 243 545 n.a. 120

15 282 100 227 650 100 98

20 333 99.0 304 869 99 93

25 374 96.6 311 869 96 94

30 461 95.1 392 1211 95 100

35 489 82.7 346 1153 83 87

Years

from

take- off

GDP ppp

US$/

capita

Steel use

kg/capita

Copper

use

kg/capita

Alum. use

kg/capita

IVA

%GDP

Savings

%GDP

Japan

1960

0 5115 n.a. 3.3 1.6 n.a. 34.2

5 7614 n.a. 4.4 3.0 n.a. 34.1

10 12 435 n.a. 7.9 8.8 n.a. 41.1



GFCF

index

Exports

index

Goods

trade

index

World

export

share

index

Urban

pop

index

Pop per

km2

index

Memo

GDP mkt

exch US$/

cap

100 100 100 100 100 100 100

98 164 134 101 116 107 138

106 257 234 160 137 115 226

127 351 310 200 159 123 325

114 306 222 322 182 130 501

133 445 362 432 206 135 733

GFCF

%GDP

Exports

%GDP

Goods

trade

%GDP

World

export

share %

Urban pop

%

Pop

per km2

persons

Memo

GDP mkt

exch mkt $/

cap

29.6 6.6 14.3 0.8 18.7 103 165

29.0 10.9 19.2 0.8 21.6 110 228

31.5 17.1 33.5 1.7 25.6 118 373

37.7 23.3 44.4 2.4 29.8 126 536

33.8 20.3 31.8 3.3 34.0 133 827

39.4 29.6 51.9 5.8 38.6 138 1209

GFCF

index

Exports

index

Goods

trade

index

World

export

share

index

Urban

pop

index

Pop per

km2

index

Memo

GDP mkt

exch US$/

cap

100 100 100 100 100 100 100

97 98 98 143 110 105 149

119 101 101 195 123 110 244

100 120 120 203 132 118 284

98 128 128 204 138 124 337

86 134 134 290 141 128 379

99 98 98 258 146 131 469

85 86 86 268 150 133 498

GFCF

%GDP

Exports

%GDP

Goods

trade

%GDP

World

export

share %

Urban pop

%

Pop

per km2

persons

Memo

GDP mkt

exch mkt $/

cap

33.5 10.7 21.0 3.2 43.1 258 7099

32.5 10.5 19.6 4.6 47.4 270 10 566

39.8 10.8 20.4 6.3 53.2 285 17 298



Table 2.1 (continued)

Years

from

take- off

GDP ppp

US$/

capita

Steel use

kg/capita

Copper

use

kg/capita

Alum. use

kg/capita

IVA

%GDP

Savings

%GDP

Japan

1960

15 14 424 41.6 7.4 10.5 41.6 33.4

20 17 058 41.2 9.9 14.0 41.2 31.9

25 19 129 39.8 10.2 14.0 39.8 32.2

30 23 599 39.5 12.8 19.6 39.5 34.1

35 25 019 34.4 11.3 18.6 34.4 29.8

Years

from

take- off

GDP ppp

index

Steel use

index

Copper

use index

Alum. use

index

IVA

index

Savings

index

Korea

1974

0 100 100 100 100 100 100

5 130 173.4 221 173 125 119

10 179 282.6 511 349 134 152

15 263 556.8 760 841 142 181

20 365 957.1 1203 1464 143 181

25 435 974.9 1844 1710 139 169

Years

from

take- off

GDP ppp

US$/

capita

Steel use

kg/capita

Copper

use kg/

capita

Alum. use

kg/capita

IVA

%GDP

Savings

%GDP

Korea

1974

0 3722 84 1.0 1.0 29.3 20.2

5 4848 146 2.2 1.8 36.6 23.9

10 6649 237 5.1 3.6 39.1 30.6

15 9792 468 7.6 8.6 41.6 36.4

20 13 597 787 12.0 15.0 41.9 36.6

25 16 172 819 18.3 17.5 40.7 34.2



the observed peak in catch- up growth rates (deepening industrial engage-

ment) of approximately $US13 000 in latecoming East Asian countries

(Garnaut et al., 2008; Perkins and Rawski, 2007). That said, China is an

outlier in the sample, with a high IVA share prior to its entry into modern

economic growth, a relic of the self- sufficiency ethic that underpinned the

pre- 1978 anti- strategic economy, and the price distortions that went along

with the industrial bias of the command system.

GFCF

%GDP

Exports

%GDP

Goods

trade

%GDP

World

export

share %

Urban pop

%

Pop

per km2

persons

Memo

GDP mkt

exch mkt $/

cap

33.4 12.8 22.8 6.5 56.8 305 20 135

32.8 13.7 25.8 6.5 59.6 319 23 917

28.7 14.4 22.8 9.3 60.6 331 26 940

33.1 10.5 17.3 8.3 63.1 339 33 280

28.4 9.2 14.9 8.6 64.6 344 35 322

GFCF

index

Exports

index

Goods

trade

index

World

export

share

index

Urban

pop

index

Pop per

km2

index

Memo

GDP mkt

exch US$/

cap

100 100 100 100 100 100 100

120 114 110 152 118 108 129

107 119 112 276 135 116 176

138 88 90 325 154 122 266

139 93 88 421 163 128 368

116 126 114 464 166 133 437

GFCF

%GDP

Exports

%GDP

Goods

trade

%GDP

World

export

share %

Urban pop

%

Pop

per km2

persons

Memo

GDP mkt

exch mkt $/

cap

26.9 30.0 56.9 0.6 48 357 2489

32.2 34.3 62.4 0.9 56.7 386 3221

28.8 35.6 63.6 1.6 64.9 413 4386

37.1 26.3 51.1 1.9 73.8 434 6615

37.3 27.9 50.3 2.4 78.2 457 9159

31.1 37.8 65.0 2.7 79.6 476 10 884

Source: Author’s calculations using data sourced from the International Monetary Fund (IMF).



In contrast to the historical record on IVA, urbanization tends to keep

increasing well past middle- income levels. It also tends to plateau rather

than decline. That is almost certainly due to the fact that IVA loses ground

to services activity, which thrives on agglomeration, rather than land-

intensive endeavours, which do not. In short, while a nation can develop

to a post- industrial state, under the technological paradigm it does not

develop to a post- urban state. Rather, it concentrates its population

further. History is littered with evidence that the optimal city size increases

with the institutional complexity that lags the strategic demands generated

by technological change (Snooks, 1997).

The degree of outward orientation and metal intensity are positively

related. The most metal- intensive economy in the sample, Korea, is a

middle- income, export- oriented manufacturer. It is also highly urbanized

and has extremely high population density. Neither a plateau in its urban-

ization rate and IVA share, nor a decline in its investment and savings

rates, has handicapped its ability to raise metal intensity beyond the usual

turning point between $10 000 and $15 000 per head. Its ability to continue

gaining global market share beyond these points has enabled metal inten-

sity to continue rising. Brazil, a low- income economy with a weak outward

orientation, is the counterpoint.

China’s metal intensity was relatively insensitive to the very early stages

of industrialization and gains in export share. In the current decade,

though, metal intensity has become substantially more sensitive to devel-

opment, catching and surpassing the comparable Japanese and then

Korean rates. The low sensitivity seen in the early years is consistent with

the conventional light- to- heavy industrial path pursued by many new

latecomers. In Japan and Korea metal intensity grew extremely rapidly in

the 5–15 years from the take- off period, while the Chinese experience was

less dramatic. China’s surge comes in the 15–25- year era. This difference

may be partially attributable to the inadequacy of our time series, which

prevents the use of true global strategic transition (GST) entry points for

Japan and Korea.2

Copper and aluminium use have been more sensitive to the Asian indus-

trialization process than has steel use. This is perhaps due to the rapidly

increasing degree of openness that has been a feature of Asian industrial-

ization, and to the relative demands of the traded and non- traded sectors

vis- à- vis ferrous and non- ferrous metals. Ferrous metals have a lower

value- to- weight ratio than the non- ferrous complex and are therefore less

likely to be traded. Further, as an economy ascends the value chain, high-

value- added durable goods will displace heavy industrial products in the

output mix, raising the demand for non- ferrous vis- à- vis ferrous metals.

While rural–urban migration underpins demand for housing, infrastruc-



ture and steel, the demands of outward- oriented manufacturing appear to

be the stronger force at higher levels of income.

Bringing these observations together, we can say that an economy might

be fairly said to be in a metal intensity sweet spot if the following is true:

it is actively engaged in industrialization, with strong strategic leadership

sponsoring technological progress; it is approaching the GDP per capita

level consistent with a peak in industry share; it is in the midst of an urban-

izing trend; it is moving up the manufacturing value chain; it is making

global market share gains and tapping into external economies of scale;

it is saving enough to rapidly build the capital stock without a binding

external financing constraint; it is reaching an age of fixed and rotating

capital stock where depreciation costs are non- trivial; it has a willingness

to increase population density and provide the requisite infrastructure to

do so; and it is moving to extend access to the key pillars of social and

economic infrastructure to all its citizens.

FOREIGN DIRECT INVESTMENT AND LONG- RUN CONSTRAINTS GLOBAL ON MARKET SHARE

China’s high degree of openness to foreign direct investment (FDI) con-

trasts with Japan but gels with Korea. China’s rapid gains in world export

market share have been primarily a function of the activities of foreign-

funded firms, particularly in the post- WTO- accession era (Figure 2.3).

Trade fragmentation, with China serving the role of ‘assembler of first

resort’, has amplified both the trend of foreign participation and the

growth of export market share (Athukorala and Yamashita, 2008). These

observations open up another difficult forecasting problem. Korea has

continued to expand its export market share almost uninterrupted and has

achieved extremely high rates of metal intensity.

China’s path of metal intensity to date is closest to the Korean experi-

ence. However, were inward FDI to stabilize or decline (or were another

region to assume ‘assembler of first resort’ status)3 then export market

share could presumably do likewise, at least until an alternative strategy

could be adopted (McKay, 2008). That would prevent China from follow-

ing the Korean path. Indigenous Korean firms have established themselves

as globally competitive innovators in a number of sectors such as electron-

ics and shipbuilding. Indigenous Chinese firms have yet to do so, and are

probably at least a decade away from establishing strong brand awareness

among non- Chinese consumers. This perspective on the Chinese develop-

ment path cautions against excessive reliance on the Korean example.

China’s ability to continue expanding its world export share will surely



be constrained by its enormous absolute scale and the resistance of its large

trading partners. The strong penetration of Chinese- made imports is already

a major political issue in the United States. The bilateral trade position

is a rallying point for both sides of US politics and the exchange rate is a

lightning rod.4 The historical example of the angst created by the bilateral

balance between the US and Japan, and the USD/JPY exchange rate, is the

obvious precedent. Over the coming decade, the expected appreciation of

China’s nominal exchange rate, and a rise in the relative price level as admin-

istrative distortions are progressively removed (Huang and Tao, 2010) and

productivity catch- up continues, should appreciate the real exchange rate

and reduce the current level of cost competitiveness enjoyed by the export

sector.5 At some point, China will be unable to seriously expect to expand

its exports at a faster pace than aggregate world demand. Korea’s small size

offers it the luxury of continuing to follow a strategy led by export manufac-

turing at income levels far higher than a larger economy – let alone a meg-

astate like China (Snooks, 1997) – would find obstacles placed in its path.

China will clearly be constrained in its choices to some extent by the stra-

tegic activities of its competitors. The attitudes of governments to Chinese

0

2

4

6

8

10

12

6

% of world exports

% of world GDP

2006

2001

Indigenous firms only

All firms

2010

8 10 12 14

Note: China’s share of world merchandise exports, with and without foreign- funded firms.

Sources: Underlying data from IMF and CEIC, author’s calculations.

Figure 2.3 China’s adjusted world export share



purchases of mineral resources in their domains will eventually become as

important as access for Chinese exports is in the current phase. This area

has already become contentious, with a Chinese principal prevented from

purchasing an energy asset in the United States, and the presence of great

angst within Australia regarding direct inroads into the resource sector by

Chinese investors (Drysdale and Findlay, 2008).

DOES INTRANATIONAL CONVERGENCE HOLD THE KEY?

A provincial approach to the Chinese data on metal intensity has the

potential to eventually remedy the shortcomings of the national data

sets.6 The provincial data also tentatively support a Kuznets- style metal-

intensity path. It is eminently reasonable to assume that there is a high

level of relevant information content in the paths already traced by

China’s wealthier provinces that are most actively engaged with the indus-

trialization and openness strategy.

This field of endeavour is particularly promising, but it does rely to a

degree on the arguable assumption that China’s wealthiest provinces have

already defined a structural peak in metal intensities from which they

are declining. In other words, they are beyond the ‘turning point’7 of an

explicit, generalizable Chinese KCS. It may be that observed reductions in

the metal intensity of the Beijing and Shanghai economies are the begin-

ning of a transition to a more cyclically informed usage, and therefore may

represent local rather than absolute maxima. After all, income levels in

Beijing and Shanghai are still only a quarter of present US levels, implying

that they are hardly at a level of productivity consistent with membership

of the strategic core. The US experience of metal intensity makes this

point about local versus global maxima in a number of historic contexts.

High- amplitude cyclicality was the norm throughout the twentieth century

(Figure 2.1(a)).

It might also be argued that Shanghai and Beijing are atypical observa-

tions. If these top- tier cities are treated as outliers, then it creates problems

for the cross- sectional analysis, because without them the Kuznets rela-

tion is far from clear. If these cities are excluded, then it would be prudent

to wait for a broader selection of provinces to mature in their steel usage

before attempting to define a definitive turning point for an economy as

diverse as China’s.

The proclivity of high- income countries to maintain an elevated if

somewhat reduced metal consumption pattern well beyond the peak in

industrialization implies that stock effects become a progressively more



significant determinant of metal demand. The analysis presented here

could therefore be augmented by data on relative capital stock, or levels of

flow investment which take account of the ongoing costs of depreciating

and replacing fixed and rotating capital. Further, the metal intensity of

fixed investment may rise as higher levels of technology are attained and

households demand more sophisticated products. House sizes increase

with income (raising ferrous input), they become more ‘wired’ (raising

non- ferrous input) and structurally sound (again raising ferrous input).

Deepening the social infrastructure stock brings initial demand for wiring

and piping and then ongoing maintenance requirements. It therefore

seems obvious that models of metal intensity that focus solely on flow

variables will be incomplete. Each of these arguments is relevant to the

provincial analysis of Chinese metal intensity, as well as the study of metal

intensity across countries.

WHITHER CHINA?

It remains to make an assessment of what the Chinese path of metal inten-

sity will look like. The reasoning of this chapter indicates that while China

may look like Korea at the time of writing, it is unlikely to do so at higher

income levels. The main constraint in this regard will be China’s inability

to follow an export- led strategy in the same way that Korea has done.

China’s continental landmass represents a major contrast with the

compact nature of either the Japanese archipelago or the Korean penin-

sula. The United States is a far better comparison in this regard. The

United States’s effort to build an internal megamarket with a continental

transport system linking a multitude of metropolitan nodes tallies well

with China’s existing plans to expand interprovincial commerce via the

construction of an ambitious national road and rail network.

The density of population in China is far higher than in the United

States at any point in its history. Therefore, while China will have a

national intercity transport network similar to that of the United States,

intracity transit systems will presumably develop very differently. China

is therefore less likely to develop cities characterized by suburban sprawl

that rely upon automobiles. Rather, it will borrow from Japanese and

European models of mass public transit. One study (McKinsey Global

Institute, 2008) estimates that China will construct 170 mass transit

systems by 2025, servicing the majority of a projected 221 urban agglom-

erations with populations in excess of 1 million people. This projected

style of urbanization reinforces the assessment that China’s future path of

automobile penetration will not resemble that of the United States.



China’s ongoing urbanization drive is proceeding on a scale that does

not have a precedent. The westward expansion of the population of the

United States (Snooks, 1997, chapter 12), or the mass Siberian migration

of European Russians (ibid., pp. 442–6), is a tempting but not terribly

productive analogy. These two historical examples were based around

physical resource acquisition and exploitation. In China’s case, labour is

moving to expedite the industrialization strategy, not to exploit untapped

physical resources. The combination of immense scale and a distinct stra-

tegic underpinning argue that this particular factor must be considered on

its own merit without the benefit of firm historical guidance.

The aforementioned factors – external saturation, internal integration,

mass urbanization and urban style (alongside an anticipated path for

income per capita) – must guide our medium- term judgement. The United

States is a reasonable comparison in some cases, Japan in others. At times,

though, China’s uniqueness shines like a beacon.

Some milestones and relevant trajectories can be defined to give some

idea of the timing of the peak in Chinese metal intensity. China should

reach the $13 000 per capita income level around 2015, if it continues to

expand at the rates projected by growth accountants, and the demog-

raphers at the United Nations have done their sums correctly (Perkins

and Rawski, 2007; Wang, 2007; Garnaut et al., 2008; He et al., 2007;

United Nations, 2007). Given that its distance from the strategic leader-

ship will still be substantial at this point (around 20 per cent of US GDP

per capita), a relative level well below that at which other East Asian

countries began to experience decelerating growth (Garnaut et al., 2008),

we might reasonably consider 2015 as the earliest possible time that per

capita growth would begin to decline. Indeed, we note that the ‘turning

point’ in the general KCS represented by cross- sectional data, for what

it is worth, is around US$24 000 per capita. In the KCS estimated for the

United States, the turning point was around US$17 000 per capita. China

cannot be reasonably expected to reach either point any sooner than 2023

or 2019, respectively.

Further to those points, the five- year period 2015–20 is the turning

point for China’s demographic profile as projected by the United Nations

population division (United Nations, 2007).8 The momentum of urbani-

zation will also have calmed by this time, with 72 million rural to urban

migrants anticipated in the five years to 2020, down 10 per cent from the

estimated peak rate of the period 2005–10. By the five years to 2030, the

rate of urbanization will have declined by 38 per cent from peak rates, with

the urban share of the population surpassing 60 per cent, equivalent to the

global average.

The saturation point for the global economy regarding China’s exports is



another core factor to consider. This was done in the previous section with

specific reference to differentiating China and Korea. Here we consider

the Japanese example. Japan’s export share grew in trend terms through

most of the Bretton Woods era, paused during the mid 1970s as energy

exporters reigned, and then continued its ascent all the way up to the Plaza

Accord of 1985. At this point Japanese GDP per capita had reached three-

quarters of the United States’s level (it would peak around 85 per cent in

1991); it had been the world’s second- largest economy for 16 years; and

the real exchange rate has more than doubled in value since the yen was

de- pegged from the US dollar. Further nominal and real yen appreciation

beyond 1985 resulted in a loss of competitiveness and precipitated a trend

decline in Japan’s export share in the second half of the 1980s. Japan also

had nowhere else to go from a sectoral perspective. By this stage Japanese

firms defined the technological frontier in many sectors. It became clear

that Japan could no longer seriously expect to sustain export growth faster

than the rate of growth in aggregate world trade. As its cost competitive-

ness eroded, increasingly it was unable to do even that. Japan’s externally

focused development strategy had been exhausted (McKay, 2008).

If these broad economic (as opposed to political) themes and relativities

need to be replicated in the Chinese instance before we reach the satura-

tion point for world export market share, then we are some way from

reaching such a peak. China will only reach two- fifths of US GDP per

capita levels by 2030, a level not far removed from the Japanese position

in the mid 1960s. Japan expanded its export share for two decades beyond

this landmark. Further, China’s real exchange rate has only just embarked

upon an appreciating trajectory. The economy has only recently exited a

protracted disinflationary period. Modest flexibility was introduced into

the nominal exchange rate regime in July 2005 (Golley and Tyers, 2007;

McKay, 2007). It seems unlikely that the level of the real exchange rate

will present a major hurdle for Chinese export market share gains for a

significant period of time.

China’s world export market share is approaching the levels at which

Japan’s share peaked, but while China is still increasing its share of world

output extremely rapidly, that growth accommodates a higher natural

share of global export trade. China is likely to become the largest economy

in the world, if not the richest. It will become far larger in relative terms

than Japan was when it achieved a 9 per cent share of world exports. As we

argued above, the growth in China’s export share will submit to gravity at

a GDP per capita level well below that currently prevailing in Korea, but

it will not face any time soon the economic hurdles that constrained Japan.

If we recognize that the politics of the situation are relevant, then it may be

that China will find that rapid global market share gains cease well ahead



of the schedule facing a ‘normal’ economy. Even a very pessimistic reading

of the situation would put that point deep into the 2010s, or possibly early

in the 2020s.

The United States increased its share of world merchandise exports

from around 3 per cent in 1800 to 9.8 per cent in 1860, 13.2 per cent by

1900 and 16 per cent in 19609 (Lipsey, 2000, p. 688, table 15.1). Clearly,

gains in world export share can continue for a very long time before the

global economy becomes saturated with the output of a nation that is itself

increasing its share of global output.

As discussed in this chapter, Chinese automobile penetration is antici-

pated to follow a path that is moderate by either Japanese or American

standards. The International Monetary Fund’s projections for automobile

ownership per thousand persons incorporate a 250 per cent expansion

between 2010 and 2020, and a 180 per cent rise between 2020 and 2030

(International Monetary Fund, 2005: 182). At 267 cars per 1000 people,

China would be less than half of the way to the saturation point observed

elsewhere. That implies that further automobile penetration beyond that

point would contribute to keeping Chinese metal intensity at a relatively

high level beyond the turning point.

The preceding discussion sketches a very broad range within which

various factors relevant to China’s intensity of steel use might peak: as

early as 2015 and as late as 2030. It seems fairly safe to trim this distribu-

tion at the near end. The idea of resource- intensive high growth decelerat-

ing as soon as 2015, at such low levels of relative income, relative capital

stock, in the midst of the urbanization drive and prior to the peak in

export market share, does not seem plausible. However, we might reason-

ably see the period 2015–20 as the likely moment when China conclusively

veers off the Korean trajectory and begins to define a more distinctive

individual path with a flatter gradient of increase.

If we allow China to follow Korea up to 2015, when we assume Chinese

income per capita has reached $13 000, then that would imply steel con-

sumption of 700 kg per capita, or 910 million tonnes. That is equivalent to

80 per cent of global steel output in 2005. A demand profile like that will

obviously put extreme pressure on global supply potential. It may encour-

age substitution decisions, and it may crowd out smaller consumers. In

sum, if China was to follow this path, other countries may have to review

their own strategies to cope, whether they are net importers or net export-

ers of resources.

The supply response to the strategic demands of the Chinese development

process will be critical to the final outcome. A strategic price signal, in the

form of the spectacular commodity price rises observed since 2003, is starkly

evident, so the incentive to invest is currently very large. The dominant



players in supplying the market, the major global diversified resources com-

panies, are trumpeting their ability to meet huge projected Chinese demands

(Albanese, 2008; Kloppers, 2008). Taking that claim on board, while

simultaneously recognizing the unpredictable nature of the exploration

and discovery process and the oligopolistic nature of the industry, one sees

no reason to presume that excess supply of raw mate rials will emerge in a

sustainable fashion in ferrous metals markets. That will keep prices elevated

relative to historical norms and continue to drive the strategic demand for

future investment. Presuming the resource endowment exists then, we can

assume that China’s choices will not be curtailed by finite supply.

The period 2015–20 represents a transition phase in the projection, with

the structural drivers becoming more erratic and beginning to blend more

frequently with cyclical factors to determine metal demand. The precise

peak in Chinese steel usage per capita would thus occur somewhere within a

few years of 2020, with the distribution of likely outcomes skewed towards

the later dates in this range. Also, the use of ferrous scrap inputs to the steel

production process will be rising through this period, albeit from a very low

base, implying that China’s call on iron ore will be lower per unit of finished

steel. China’s low usage of scrap vis- à- vis steel producers in developed

countries, combined with the rapid increase in China’s share of finished

steel output, has raised the global ratio between steel output and iron ore

input. That has pushed iron ore prices higher than they would have been

otherwise. While this point is not integral to the central projection, it is a

matter of great importance for the iron- ore- producing community.

Importantly, it must be acknowledged that demand would become sig-

nificantly more volatile around the turning point. The experience of the

United States highlights that once cyclical factors become a material deter-

minant of per capita steel usage, the path of annual observations describes

a violent saw- tooth pattern. The gradient of the underlying trend is likely

to flatten appreciably once China moves assertively away from the Korean

course. It may even flatten absolutely. This seems more likely than a trend

of swift erosion in the intensity of steel usage on the far side of the turning

point. Given that the broad range initially defined stretched all the way

out to 2030, and the huge uncertainties at play, a hedging forecast allows

for a flat trend for a time following the tentatively defined peak.

CONCLUSIONS

Attempting to sketch China’s future path of metal intensity is not possible

without first assessing its overall strategic attitudes towards a multitude of

issues, both domestic and international.



The data presented in this chapter offer a broad range of perspectives

on the change in metal intensity at various distances from take- off. Some

tentative generalizations have emerged, as have some important findings

for individual countries. However, there are shortcomings in using either

the general- to- specific or the specific- to- general approach in a forecasting

framework.

As for China’s future path of metal intensity, it seems unlikely that it

will continue to resemble Korea’s beyond 2015 or so. The post- 2015 trajec-

tory will feature certain aspects that are reminiscent of the paths taken by

the United States and Japan, but the final outcome will not fully resem-

ble either. As a tentative and preliminary judgement, and assuming that

resource limits are not reached, the peak in Chinese steel usage per capita

is likely to occur at a point not too distant from 2020: possibly sooner,

but more likely later. The peak level of steel intensity will be well below

Korean levels of consumption, but close enough to Japan’s peak rate. It

is suggested that this peak will be sustained through the early and middle

2020s, before declining towards the end of the decade. Among all the

moving parts, the principal factor in deciding upon this long, flattish peak

is the current ‘under- urbanization’ of China relative to its development

level.

NOTES

1. Kuznets used data from India (1949/50), Ceylon (now Sri Lanka, 1950) and Puerto Rico (1948).

2. While an entry point for Korea in the early to middle 1960s is uncontroversial, Japan is a different case. Japan’s economic exploits prior to the Second World War are often discounted by non- specialists, but its industrial complex was strong enough to furnish a military force capable of defeating a European power (Russia) in a land and naval war in 1904/05. A late 1800s entry point, two generations after the first tier of European industrializers, at 37 per cent of western- European GDP/capita levels (Maddison, 2003), would be quite reasonable. This is a debate for another day, but it underscores once again that the available data fall well short of the worthy task at hand.

3. India, with its formidable labour supply profile, is the most commonly cited candidate for this role. A more likely outcome is that a group of competing low- income, labour- surplus economies will eventually displace China. In addition to India, long- run Asian candidates include Pakistan, Bangladesh, Indonesia, Vietnam and possibly emerging strategic states in North Korea and Myanmar. In 2050, the combined population of these countries excluding India is expected to be 1.1 billion (United Nations, 2007).

4. For multiple perspectives on China’s exchange arrangements, see McKay (2007), Fan (2006), Goldstein (2004), Eichengreen (2004), Frankel (2004), Prasad et al. (2005) and McKinnon (2006).

5. See Golley and Tyers (2007) for an examination of the dynamics of the real exchange rate. They argue that the depreciating impact of high domestic savings has more than offset a vector of appreciating factors, leading to the somewhat surprising depreciation of the real rate over the last decade.



6. A provincial approach to environmental problems, utilizing a non- linear framework, is also bearing fruit (Cai and Du, 2008; Bao et al., 2008; Bao and Peng, 2006).

7. This is not to be confused with a Lewisian turning point (Garnaut and Song, 2006a, 2006b).

8. For an analysis of the demographic profile, and an application to the long- run growth trajectory, see Golley and Tyers (2006).

9. The figure for 1960 is the author’s calculation from World Bank data accessed via sub-scription to the World Development Indicators database. All other data on historical United States exports are taken from the in- text reference.

REFERENCES

Albanese, T. (2008), ‘Winning strategies for global champions’, speech to the Australia–Israel Chamber of Commerce, Sydney, 16 June, accessed at www.riotinto.com/media/speeches_7862.asp.

Athukorala, P.C. and N. Yamashita (2008), ‘Global production sharing and US–China trade relations’, in L. Song and W.T. Woo (eds), China’s Dilemma: Economic Growth, the Environment and Climate Change, Canberra: Asia Pacific Press, and Washington, DC: Brookings Institution Press, pp. 59–89.

Bao, Q., and S. Peng (2006), ‘Economic growth and environmental pollu-tion: a panel data analysis’, in R. Garnaut and L. Song (eds), The Turning Point in Chinese Economic Development, Canberra: Asia Pacific Press, pp. 294–313.

Bao, Q., Y. Chen and L. Song (2008), ‘The environmental consequences of foreign direct investment in China’, in L. Song and W.T. Woo (eds), China’s Dilemma: Economic Growth, the Environment and Climate Change, Canberra: Asia Pacific Press, and Washington, DC: Brookings Institution Press, pp. 243–64.

Cai, F. and Du, Y. (2008), ‘The political economy of emissions reduction in China’, in L. Song and W.T. Woo (eds), China’s Dilemma: Economic Growth, the Environment and Climate Change, Canberra: Asia Pacific Press, and Washington, DC: Brookings Institution Press, pp. 226–42.

Drysdale, P. and C. Findlay (2008), ‘Chinese foreign direct investment in Australia: policy issues for the resource sector’, paper presented to Crawford School Public Seminar, Australian National University, 4 September.

Fan, G. (2006), ‘Global imbalance, China and the international currency system’, in R. Garnaut and L. Song (eds), The Turning Point in Chinese Economic Development, Canberra: Asia Pacific Press, pp. 87–102.

Frankel, J. (2004), ‘On the renminbi: the choice between adjustment under a fixed exchange rate and adjustment under a flexible rate’, paper presented to High- Level Seminar on Foreign Exchange System, Dalian, China, May.

Garnaut, R. and L. Song (eds) (2006a), The Turning Point in Chinese Economic Development, Canberra: Asia Pacific Press.

Garnaut, R. and L. Song (2006b), ‘China’s resources demand at the turning point’, in R. Garnaut and L. Song (eds), The Turning Point in Chinese Economic Development, Canberra: Asia Pacific Press, pp. 276–93.

Garnaut, R. and L. Song (eds) (2007), China: Linking Markets for Growth, Canberra: Asia Pacific Press.

Garnaut, R., S. Howes, F. Jotzo and P. Sheehan (2008), ‘Emissions in the Platinum Age: the implications of rapid development for climate change mitiga-



tion’, Garnaut Review Working Paper revised draft 2 May 2008, available at: www.garnautreview.org.au.

Goldstein, M. (2004), ‘Adjusting China’s exchange rate policies’, a discussion paper of the Institute for International Economics presented at the IMF seminar on China’s foreign exchange rate system, Dalian, China, 26–27 May.

Golley, J. and R. Tyers (2006), ‘China’s growth to 2030: demographic change and the labour supply constraint’, in R. Garnaut and L. Song (eds), The Turning Point in Chinese Economic Development, Canberra: Asia Pacific Press, pp. 203–26.

Golley, J. and R. Tyers (2007), ‘China’s real exchange rate’, in R. Garnaut and L. Song (eds), China: Linking Markets for Growth, Canberra: Asia Pacific Press, pp. 316–43.

Grossman, G. and A. Krueger (1995), ‘Economic growth and environment’, Quarterly Journal of Economics, 110 (2), 353–77.

He, J., S. Li and S. Polaski (2007), ‘China’s economic prospects 2006–2020’, Carnegie Papers, no. 83, April.

Huang, Y. and K. Tao (2010), ‘Causes and remedies of China’s external imbal-ances’, China Center for Economic Research Working Paper no. E2010002, 25 February, Peking University.

Hwang, K.H. and J.E. Tilton (1990), ‘Leapfrogging, consumer preferences, international trade and the intensity of metal use in less developed countries’, Resources Policy, 16 (3), 210–24.

International Iron and Steel Institute (1972), Projection 85: World Steel Demand, Brussels: International Iron and Steel Institute.

International Monetary Fund (2005), ‘Will the oil market continue to be tight?’, World Economic Outlook, April, Washington, DC: International Monetary Fund, chapter 4.

Kloppers, M. (2008), ‘Resourcing the future’, investor presentation, Sydney, NSW, 12 June, accessed at www.bhpbilliton.com/bb/investorsMedia/invest-mentPresentations.jsp.

Kuznets, S. (1955), ‘Economic growth and income inequality’, American Economic Review, 45 (1), 1–28.

Lohani, P.R. and J.E. Tilton (1993), ‘A cross- section analysis of metal intensity of use in the less developed countries’, Resources Policy, 19 (2), 145–54.

Lin, J.Y. (2008), ‘Rebalancing equity and efficiency for sustained growth’, in L. Song and W.T. Woo (eds), China’s Dilemma: Economic Growth, the Environment and Climate Change, Canberra: Asia Pacific Press, and Washington, DC: Brookings Institution Press, pp. 90–109.

Lipsey, R.E. (2000), ‘US foreign trade and the balance of payments 1800 to 1913’, in S.L. Engerman and R.E. Gellman (eds), Cambridge Economic History of the United States, vol. II: The Long Nineteenth Century, Cambridge: University Press.

Maddison, A. (2003), The World Economy: Historical Statistics, Paris: OECD.Malenbaum, W. (1973), Material Requirements in the United States and Abroad in

the Year 2000, Philadelphia, PA: University of Pennsylvania Press.Malenbaum, W. (1975), ‘Law of demand for minerals’, in Proceedings of the

Council of Economics, 104th Annual Meeting of the American Institute of Mining, Metallurgical and Petroleum Engineers, Dallas, TX, pp. 145–55.

McKay, H. (2007), ‘Reforming China’s exchange arrangements: monetary



and financial sovereignty, sequencing and the foreign exchange market’, in R. Garnaut and L. Song (eds), China: Linking Markets for Growth, Canberra: Asia Pacific Press, pp. 290–315.

McKay, H. (2008), ‘Asian industrialization: a strategic analysis with a memoran-dum on the Australian response’, Global Dynamic Systems Centre working paper WP004, accessed at ht t p : / / e c o n r s s s . a n u . e d u . a u / G D S C p a p e r s . h t m .

McKinnon, R. (2006), ‘Why China should keep its exchange rate pegged to the dollar: an historical perspective from Japan’, Stanford working paper, October, accessed at ww w . s t a n f o r d . e d u / ~ m c k i n n o n / p a p e r s / I n t e r n a t i o n a l p e r c ent20Finance%20China%20peg.pdf.

McKinsey Global Institute (2008), ‘Preparing for China’s urban billion’, March, accessed via subscription at www.mckinsey.com.

Perkins, D.H. and T.G. Rawski (2007), ‘Forecasting China’s economic growth to 2025’, accessed at http://post.economics.harvard.edu/faculty/perkins/papers/Chapter20.pdf.

Prasad, E., T. Rumbaugh and Q. Wang (2005), ‘Putting the cart before the horse? Capital account liberalization and exchange rate flexibility in China’, IMF policy discussion paper PDP/05/01, January.

Preobrazhensky, E.A. (1967), The New Economics, eds B. Pearce and A. Nove, Oxford: Oxford University Press.

Snooks, G.D. (1997), The Ephemeral Civilisation: Exploding the Myth of Social Evolution, London: Routledge.

Snooks, G.D. (1999), Global Transition: A General Theory of Economic Development, London: Macmillan.

Song, L. and W.T. Woo (2008), China’s Dilemma: Economic Growth, The Environment and Climate Change, Canberra: Asia Pacific Press, and Washington, DC: Brookings Institution Press.

United Nations (2007), World Urbanization Prospects: The 2007 Revision Population Database, accessed at ht t p : / / e s a . u n . o r g / u n u p / i n d e x . a s p .

Wang, X. (2007), ‘Pattern and sustainability of China’s economic growth towards 2020’, paper presented at the ACESA 2007 conference on ‘China’s Conformity to the WTO: Progress and Challenges’, Australian National University, Canberra, 13–14 July.



APPENDIX

Table 2.A1 Estimated output of filtered steel use per head

Variable Coefficient Standard

error

t-Statistic Probability

Y 0.6938 0.0464 14.9452 0.000 0

Y 2 −0.0398 0.0028 −14.4674 0.000 0

C −2.5770 0.1922 −13.4103 0.000 0

Dependent variable: Hodrick–Prescott filter of US steel use per capita (tonnes).Sample: 1929 to 2002.Observations: 74.Method: OLS.

R-squared 0.787 355 Mean dependent variable 0.374 73Adjusted R-squared 0.781 365 S.D. dependent variable 0.079 08S.E. of regression 0.036 976 Akaike information criterion −3.717 379Sum-squared residual 0.097 075 Schwarz criterion −3.623 971Log likelihood 140.543 F-statistic 131.444 8Durbin–Watson statistic 0.093 358 P (F-statistic) 0.000 0Y is US$ GDP per capita; Y 2 is its squared term; C is a constant.Y is entered as a natural logarithm.

Source: Author’s estimates.

Table 2.A2 Estimated output of unfiltered steel use per head


error


Y 0.7551 0.0740 10.2053 0.0000

Y 2 −0.0434 0.0044 −9.8894 0.0000

C −2.8347 0.3063 −9.2545 0.0000

Dependent variable: unadjusted data of US steel use per capita (tonnes).Sample: 1929 to 2002.Observations: 74.Method: OLS.

R-squared 0.6308 Mean dependent variable 0.3747Adjusted R-squared 0.6204 S.D. dependent variable 0.0957S.E. of regression 0.0589 Akaike information criterion −2.7849Sum-squared residual 0.2466 Schwarz criterion −2.6915Log likelihood 106.04 F-statistic 60.663Durbin–Watson statistic 0.9411 P (F-statistic) 0.0000Y is US$ GDP per capita; Y 2 is its squared term; C is a constant.Y is entered as a natural logarithm.




Table 2.A3 Estimated output of unfiltered steel use per head, including

automobile penetration


error


Y 0.6458 0.1239 5.2126 0.0000

Y 2 −0.0354 0.0067 −5.2648 0.0000

AU 3.3692 0.8586 3.9241 0.0002

AU 2 −0.3001 0.0731 −4.1031 0.0001

C −11.8872 2.1825 −5.4466 0.0000

Dependent variable: unadjusted data of US steel use per capita (tonnes).Sample: 1929 to 2002.Observations: 74.Method: OLS.R-squared 0.7287 Mean dependent variable 0.3747Adjusted R-squared 0.7130 S.D. dependent variable 0.0957S.E. of regression 0.0513 Akaike information criterion −3.0389Sum-squared residual 0.1813 Schwarz criterion −2.8832Log likelihood 117.44 F-statistic 46.3339Durbin–Watson statistic 1.2623 P (F-statistic) 0.0000Y is US$ GDP per capita; Y 2 is its squared term; C is a constant.AU is automobiles per 1000 persons; AU 2 is its squared term.Y and AU are entered as natural logarithms.



45

3. Economic growth, regional disparities and core steel demand in China

Jane Golley, Yu Sheng and Yuchun Zheng

INTRODUCTION

Two of the processes underpinning China’s economic growth and devel-

opment during the three decades between 1978 and 2008 have been

industrialization and urbanization. As one of the key inputs into these

two processes, Chinese crude steel demand has been strong throughout

this period, outstripping domestic production and making China a net

importer through to 2006. Domestic consumption is the key determi-

nant of domestic production and, with China being the largest steel

producer in the world at the time of writing, this makes understand-

ing future trends in Chinese steel demand a matter of both national

and global importance. Although we have the benefit of hindsight with

regard to the relationship between economic growth and steel demand

for a number of advanced economies, such as the United States, Japan

and South Korea, it is unclear which of these relationships, if any,

is likely to be most relevant to understanding that relationship for

China.1 Indeed, it is most likely that China’s path will be unique because

of a range of specific characteristics that are simply not replicated

elsewhere.

As an alternative to using the experience of other countries to under-

stand the future trajectory of Chinese steel demand, provincial- level

analysis offers a fruitful line of research. While there are certainly some

other provincial- level characteristics that are likely to make future tra-

jectories for steel demand differ across provinces, just as they do across

countries, it seems reasonable to assume that China’s less- developed

provinces are more likely to replicate the past trends of its leading

provinces than those of countries elsewhere. With the use of time- series

data, moreover, we are able to overcome the problems of cross- sectional

analysis pointed out by Huw McKay in Chapter 2, in order to clarify the



driving forces behind dynamic trends in provincial- level steel demand

in the Chinese economy. In particular, we utilize the fact that there are

significant disparities in the levels of industrialization, urbanization

and per capita fixed- asset investment across China’s 31 provinces and

independent administrative metropolitan cities in order to address three

key questions. First, how can we address the provincial- level demand

for crude steel, given the absence of accurate, available data? Second,

what is the relationship between per capita income and the provincial

demand for crude steel per capita in China? Third, how do provincial

disparities in economic development impact on China’s total demand

for crude steel?

INDUSTRIALIZATION, URBANIZATION AND ECONOMIC GROWTH: IMPLICATIONS FOR REGIONAL DISPARITIES IN CHINA

In his seminal work on modern economic development, Simon Kuznets

(1965) identified industrialization as the central feature of the interrelated

set of structural transformation processes that accompany economic

growth.2 From an agricultural society to a modern industrial society,

the industrialization process is characterized by a number of common

features, including the following: an increase in the share of value- added

output created by secondary industry (mining, manufacturing and con-

struction, but particularly manufacturing) and a consequent fall in the

share of output created by primary industry (agriculture, forestry and

fisheries); an improvement in the technology base and the formation of an

integrated industrial system; rising levels of rural–urban migration and a

consequent increase in the urbanization rate; the establishment of tertiary

industry (services) and a rise in its contribution to national output; and

increasing per capita GDP.

While these features are common to the industrialization experience of

virtually all countries, the pace and extent of change varies substantially

from country to country and from era to era. One of the reasons behind

these cross- country growth differentials is that late- industrializing coun-

tries have been able to receive help from advanced countries in terms

of investment funding, technology and market access. Another notable

feature is that economic growth rates decline towards the end of the indus-

trialization phase – for example, most of the early industrializers expe-

rienced slower growth rates in 1973–98 than in 1950–73. Thus, Japan’s

growth slowed from the 1970s onwards, Singapore’s from the 1980s and

Korea’s from the late 1990s.3


Economic growth, regional disparity and core demand 47

The manufacturing sector is the key supporter of high- speed growth

during the industrialization process, although each country will clearly

have quite different industrial structures both within manufacturing and

across all industrial sectors. Meanwhile, the agricultural population is

a key limiting factor during the process of industrialization, with the

speed of industrialization being impacted not only by population size

but also by factors affecting the speed at which rural–urban migration

can unfold. China’s hukou system and its vast population suggest, for

example, that the process for the country on the whole will take longer

than in a country with a small population and no restrictions on popula-

tion movements. This has been evident in China’s urbanization process

during the past two decades. While over 150 million people have moved

from rural to urban areas since the late 1970s, the share of the popula-

tion living in urban areas had reached just 46 per cent in 2006, which is

well below the urbanization level for countries with China’s level of per

capita income according to the Chenery–Syrquin (1975) standard. Zhao

and Zhang (2008) note that in addition to the hukou system, low levels

of per capita natural resources have also restrained the pace of urbaniz-

ation in China.

Chenery et al. (1986) analysed a wide range of data from 137 coun-

tries, including income per capita, industry structure and urbanization

rates. This enabled them to divide industrialization into six stages,

summarized in Table 3.1.4 These stages reflect an extremely broad

spectrum of economic development, as suggested by the fact that Haiti,

Indonesia, Brazil, Poland, South Korea and Luxembourg fall into

stages 1 to 6, respectively. Table 3.2 lists GDP per capita, the proportion

of output produced by primary, secondary and tertiary industries, and

the urbanization rate for each of China’s provinces in 2006. According

to the national average (in the bottom row), China’s per capita GDP of

US$2214 in 2006, its urbanization rate of 44 per cent and its primary

share of output of 12 per cent suggest that it is somewhere between

stages 2 and 3.5 However, its share of secondary industry is higher than

any country in the entire range, while its share of tertiary industry places

it in stage 1. The inability to place China clearly in any one of these cat-

egories is indicative that using international experience to understand

China’s development path can be problematic. The task remains prob-

lematic at the provincial level as well, but for illustrative purposes an

attempt at allocating the provinces across the six industrialization stages

is made in Table 3.3, indicating that China’s provinces range from stage

1, or pre- industrialization, to stage 4, the last stage of industrialization.

A common deviation for all provinces from the standard pattern in

Table 3.1 is that provincial shares of tertiary industry are relatively low


48

Table

3.1

T

he

six

sta

ges

of

indust

riali

zati

on

Basi

c in

dex

Pre

-

ind

ust

riali

zati

on

Ind

ust

riali

zati

on

Po

st-i

nd

ust

riali

zati

on

12

34

56

GN

I p

er c

ap

ita (

US

$)

1964

100–200

200–400

400–800

800–1500

1500–2400

2400–3600

2006

699–1300

1301–2599

2600–5000

5001–10 0

00

10 0

01–25 0

00

.25 0

01

Sh

are

of

pri

mary

(P

), s

eco

nd

ary

(S

) an

d t

erti

ary

(T

) in

du

stry

in G

NI

(%)

P .

25

S ,

30

T ,

50

15 ,

P ,

25

25 ,

S ,

35

50 ,

T ,

60

6 ,

P ,

15

30 ,

S ,

40

50 ,

T ,

60

4 ,

P ,

8

30 ,

S ,

40

60 ,

T ,

65

P ,

4

30 ,

S ,

40

60 ,

T ,

70

P ,

3

S ,

30

T .

65

Urb

an

po

pu

lati

on

(U

) (%

)U

, 3

530 ,

U ,

50

40 ,

U ,

60

50 ,

U ,

70

U .

65

U .

70

Sourc

es:

Ch

ener

y e

t al.

(1986)

for

1964,

con

ver

ted

to

2006 c

urr

ent

pri

ces

by Z

hen

g Y

uch

un

. S

hare

s o

f p

rim

ary

, se

con

dary

an

d t

erti

ary

in

du

stry

an

d u

rban

po

pu

lati

on

are

base

d o

n t

he

ran

ge

of

share

s fo

r th

e co

un

trie

s u

sed

by C

hen

ery e

t al.

, as

giv

en i

n W

orl

d B

an

k (

2006).



while shares of secondary industry are relatively high. This may be one

reason why China’s steel demand is and may remain on a higher trajec-

tory than other countries, given that secondary industry is relatively

steel- intensive.

Table 3.2 Provincial GDP per capita, industry shares and urbanization

rates in 2006

GDP per

capita ($)

Primary Secondary Tertiary Urban

population

Shanghai 7237 0.9 48.5 50.6 88.7

Beijing 6331 1.3 27.8 70.9 84.3

Tianjin 5164 2.7 57.1 40.2 75.7

Zhejiang 3998 5.9 54.0 40.1 56.5

Jiangsu 3614 7.1 56.6 36.3 51.9

Guangdong 3554 6.0 51.3 42.7 63.0

Shandong 2985 9.7 57.7 32.6 46.1

Liaoning 2733 10.6 51.1 38.3 59.0

Fujian 2693 11.8 49.1 39.1 48.0

Inner Mongolia 2515 13.6 48.6 37.8 48.6

Zhejiang 2128 13.8 52.4 33.8 38.4

Heilongjiang 2032 11.9 54.4 33.7 53.5

Jilin 1972 15.7 44.8 39.5 53.0

Xinjiang 1882 17.3 48.0 34.7 37.9

Shanxi 1772 5.8 57.8 36.4 43.0

Henan 1670 16.4 53.8 29.8 32.5

Hubei 1668 15.0 44.4 40.6 43.8

Hainan 1587 32.7 27.4 39.9 46.1

Chongqing 1563 12.2 43.0 44.8 46.7

Shaanxi 1523 10.8 53.9 35.3 39.1

Hunan 1499 17.6 41.6 40.8 38.7

Ningxia 1486 11.2 49.2 39.6 43.0

Qinghai 1475 10.9 51.6 37.5 39.3

Jiangxi 1355 16.8 49.7 33.5 38.7

Sichuan 1323 18.5 43.7 37.8 34.3

Tibet 1308 17.5 27.5 55.0 28.2

Guangxi 1292 21.4 38.9 39.7 34.6

Anhui 1261 16.7 43.1 40.2 37.1

Yunnan 1125 18.7 42.8 38.5 30.5

Gansu 1098 14.7 45.8 39.5 31.1

Guizhou 726 17.2 43.0 39.8 27.5

Average 2341 13.0 47.1 39.9 46.4

Source: National Bureau of Statistics (2007).



STEEL CONSUMPTION AND ECONOMIC GROWTH: A PROVINCIAL- LEVEL ANALYSIS

Traditionally, stable long- run economic growth was considered a suf-

ficient condition for stable long- run metal demand growth, through its

ongoing impact on metal- intensive sectors such as capital equipment,

transport and consumer durables. However, in the early 1970s, many

developed economies experienced a permanent slowdown in metals con-

sumption growth, despite continued economic growth overall.6 This gave

rise to the notion of an inverted- U- shaped long- term relationship between

GDP growth and metal consumption growth, or equivalently between per

capita GDP and per capita metal consumption. This relationship emerges

as a consequence of economic growth and development, two major com-

ponents of which are industrialization and urbanization. At low levels

of per capita GDP – that is, in the pre- industrialization stage described

above – national output is concentrated largely in primary industry,

which is characterized by relatively low per capita metal consumption. As

per capita GDP rises and the economy enters the industrialization stage,

changing consumer preferences drive a gradual shift towards more metal-

intensive products, including infrastructure and housing construction,

manufacturing, consumer durables and capital equipment. Urbanization

rates rise significantly during this stage as well, underpinning much of the

change in consumer preferences and production structure. During this

stage, metal consumption growth exceeds GDP growth and so per capita

metal consumption increases. In the post- industrialization stage, while

per capita income continues to rise, urbanization rates tend to plateau

Table 3.3 China’s mainland provinces in different industrialization stages

Provinces Population, millions

1 Tibet, Guangxi, Anhui, Yunnan, Gansu, Guizhou (6) 219.6 (16.7%)

2 Inner Mongolia, Hebei, Heilongjiang, Jilin, Xinjiang,

Shanxi, Hubei, Henan, Hainan, Chongqing, Shaanxi,

Hunan, Ningxia, Qinghai, Jiangxi, Sichuan (16)

637.3 (48.5%)

3 Zhejiang, Jiangsu, Guangdong, Shandong, Liaoning,

Fujian (6)

389.7 (29.6%)

4 Shanghai, Beijing, Tianjin (3) 44.7 (3.4%)

5 – –

6 – –

Source: Summarized by the authors.



and this, combined with the ongoing shift towards non- metal- intensive

services and high- technology products, drives per capita metal consump-

tion down.

Historical data for the early industrializers illustrate the idea of an

inverted- U- shaped relationship between steel consumption per capita

and per capita income, although it is clear that the relationship varies

over time and place. In 1974 the United States reached a peak of steel

consumption per capita of 674 kg with GDP per capita of US$20 050.7

Steel consumption per capita remained above 500 kg until 1980, and

then fell as low as 293 kg in 1982, thereafter fluctuating between that

level and 434 kg. Japan peaked in the same year as the US but with a

higher per capita steel consumption of 717 kg and a lower GDP per

capita of US$14 170. In 1970 the UK reached its peak steel consump-

tion at a lower per capita steel consumption than the US with a lower

per capita income, while Germany’s peak steel consumption was higher

and occurred at a lower per capita income level. In contrast, steel con-

sumption per capita in South Korea and Brazil, so- called newly indus-

trializing economies, has not yet revealed any downturn at the time of

writing. By 2004, South Korea’s crude steel consumption per capita

reached 981 kg, more than ten times the level in 1974. During this period

per capita income increased fivefold to reach US$18 840 – close to the

turning point for the United States but well beyond Japan’s. By 2005

Taiwan’s per capita steel consumption of 870 kg was higher than any

of the early industrializers’ peak levels, as was its per capita income of

US$21 446.

In line with the substantial differences in terms of economic develop-

ment among China’s provinces, there are also enormous differences in

crude steel consumption per capita, as illustrated in Figure 3.1. Shanghai’s

apparent crude steel consumption per capita of 769 kg in 2006 was 6.7

times higher than Guizhou’s mere 114 kg per capita. GDP per capita in

Shanghai was 9.4 times higher than in Guizhou. While Shanghai’s per

capita consumption was already past the peak of both the UK’s and the

United States’, its per capita income was not nearly as high as either one

was at the turning point. On average, it is even more obvious that China’s

provinces have not yet entered the post- industrialization stage referred

to in Table 3.2, and it seems very unlikely that the country on the whole

is even close to its own turning point, regardless of the corresponding

values of steel consumption and income for other countries. That is not

to say that a turning point will not emerge in the future, as the subsequent

analysis reveals.



ESTIMATING THE PROVINCIAL- LEVEL CORE DEMAND FOR CRUDE STEEL, 1979–2004

Although the relationship between per capita steel consumption and per

capita GDP at the provincial level can provide useful information on pro-

jecting Chinese total demand for crude steel, few studies have been carried

out owing to the data problems related to steel consumption at the provincial

level. The problems associated with using the official reported data on appar-

ent steel consumption are at least twofold. First, the data go back only as far

as the early 1990s, which is problematic in terms of the panel data estima-

tion techniques that are most appropriate for dealing with the issue at hand

– namely, long- term dynamics of Chinese steel demand. More crucially,

apparent crude steel demand for each province is calculated from official sta-

tistics simply by subtracting exports from total crude steel production. This

method is problematic because it does not consider interprovincial trade

of crude steel owing to data availability, and so estimating consumption

will be biased by production data that are affected by central planning and

government preference (via state ownership of large steel enterprises). For

0

100

200

300

400

500

600

700

800

900

Provinces

App

aren

t cru

de s

teel

con

sum

ptio

n(k

ilogr

am p

er c

apita

)

0

1000

2000

3000

4000

5000

6000

7000

8000

GD

P p

er c

apita

(U

S$

2000

pric

e)

Gui

zhou

Gan

suY

unna

nA

nhui

Gua

ngxi

Tib

etS

ichu

anJi

angx

iQ

ingh

aiN

ingx

iaH

unan

Sha

anxi

Cho

ngqi

ng

Hen

anH

aina

n

Hub

eiS

hanx

iX

injia

ngJi

linH

eilo

ngjia

ngH

ebei

Inne

r M

ongo

liaF

ujia

nLi

aoni

ngS

hand

ong

Gua

ngdo

ngJi

angs

uZ

hejia

ngT

ianj

inB

eijin

gS

hang

hai

Steel consumption per capita, (kg)GDP per capita, (US$)

Source: China Iron and Steel Statistical Yearbook, various years.

Figure 3.1 Apparent steel consumption per capita and GDP per capita,

2006



example, in Figure 3.1 above the apparent consumption of crude steel per

capita of Inner Mongolia was more than Liaoning’s and Shandong’s while

Ningxia’s was more than Jilin’s and Anhui’s. These figures are unconvincing

and reflect the high concentration of state- owned steel production in Inner

Mongolia and Ningxia rather than high consumption. Failure to deal with

this problem may lead to pseudo- regressions if GDP per capita is also cor-

related with steel production at the provincial level.

In order to deal with these problems, we propose an econometric

method for estimating core provincial- level steel consumption or under-

lying provincial- level steel demand by using information on industri-

alization, urbanization and the fixed- asset investments that result from

economic growth.8 Henceforth, we use the terms ‘core’ and ‘underlying’

interchangeably. Specifically, we estimate provincial demand for crude

steel between 1978 and 2004 by regressing steel production on measures

of industrialization, urbanization and fixed assets investment per capita.

We can do so for two reasons. First, owing to the fact that steel’s low-

value- added status makes it unprofitable for long- distance transportation,

crude steel production at the provincial level captures crucial information

about crude steel consumption at the provincial level.9 Second, given that

industrialization, urbanization and fixed- asset investment are demand-

related factors that are largely independent of supply, they can be used

as instruments for separating crude steel consumption from production.

This method enables us to construct time- series data that more accurately

reflect the underlying pattern of crude steel consumption across provinces.

Equation (3.1) specifies the regression model that is used for our first-

stage estimation:

ln (ProdSteelit) 5 b0 1 b1Industrializationit 1 b2Urbanizationit

1 b3 ln (FixedAssetInvit) 1 ui 1 eit, (3.1)

where ProdSteelit is the output of crude steel per capita in province i at

time t; Industrializationit, Urbanizationit and FixedAssetInvit are an indus-

trialization index (namely, the share of secondary and tertiary industry in

total output value), the urban share of the population and the amount of

fixed- asset investment per capita at 2000 constant prices; and ui represents

the time- invariant specific effects of each province. bn represents the coef-

ficient to be estimated and eit is the residual.

Data are drawn from a variety of sources. The production of crude steel

by provinces is available for 26 provinces between 1979 and 2004 in various

issues of the China Iron and Steel Statistical Yearbook.10 The industrializa-

tion index, urban population shares and fixed- asset investment ratios are



also available for the same 26 provinces, and the same time period, in

China Comprehensive Data Collection 55 Years: 1949–2004 (CNBS, 2010).

To eliminate province- specific and time- specific effects, we adopt the

panel data regression technique with random effects to estimate Equation

(3.1), with the results presented in Table 3.4.11 From this table, it is clear

that industrialization, urbanization and fixed- asset investments each play

an important role in affecting the demand for crude steel, since their coef-

ficients are all positive and significant at the 1 per cent level.

Combining the estimated coefficients of the industrialization index, the

urbanization index and the fixed- asset investment index with their cor-

responding real value, we can generate the underlying demand for crude

steel per capita at the provincial level from 1979 to 2004:

ln(DSteelˆit) 5 b1Industrializationit 1 b2Urbanizationit

1 b3 ln (FixedAssetInvit) , (3.2)

where DSteelˆit is the predicted underlying demand for crude steel per

capita in province i at time t. The average estimated core demand level

for each year is presented in Table 3.5 and Figure 3.2, while Table 3.6 and

Figure 3.3 give the estimated consumption for each province in 2004. There

Table 3.4 Estimation of ‘core’ demand for crude steel per capita at the

provincial level, 1979–2004

Random-effect model Fixed-effect model

Industrialization Index 0.029*** 0.028***

(0.004) (0.004)

Urbanization index 0.006*** 0.004

(0.003) (0.003)

ln_fixedassetinvestments 0.300*** 0.307

(0.025) (0.025)

Constant −0.235 −0.156

(0.251) (0.212)

Number of observations 676 676

Number of groups 26 26

Number of years 26 26

Adjusted R-squared 0.566 0.552

Note: ***, ** and * represent significance at the 1, 5 and 10 per cent levels, respectively. Numbers in parentheses are standard errors.

Source: Authors’ own estimates.



are three significant features worth highlighting. First, China’s period of

rapid economic growth has been accompanied by a rising average core

demand for crude steel, from 30 kg per person in 1979 to 159 kg in 2003.12

Second, there are significant regional disparities across regions in core

demand for crude steel, ranging from 234 kg per capita in Shanghai to just

64 kg per capita in Guizhou in 2004. Third, the consumption of crude steel

per capita appears to be (unsurprisingly) higher in the eastern region than

in the central and western regions, a point to which we return below.

In Table 3.6, we compare the estimated core demand for crude steel per

Table 3.5 Production and core demand of/for crude steel per capita,

1979–2004 (kg/person)

Year Number of

regions

Production

of crude steel

Standard

error

Core demand

for crude steel

Standard

error

1979 26 61.72 101.05 30.41 23.21

1980 26 64.78 104.38 32.55 24.29

1981 26 60.61 99.61 32.02 25.76

1982 26 62.53 97.37 32.85 25.25

1983 26 70.42 99.94 34.52 26.80

1984 26 73.53 106.83 37.48 27.53

1985 26 78.36 111.03 45.01 31.49

1986 26 89.47 141.44 47.42 32.58

1987 26 95.62 150.36 50.79 34.26

1988 26 100.06 151.42 54.35 34.60

1989 26 100.06 143.95 54.37 33.75

1990 26 106.90 159.40 53.72 35.42

1991 26 114.77 174.04 60.28 37.73

1992 26 131.84 207.33 72.18 41.65

1993 26 145.26 224.03 85.21 49.06

1994 26 148.78 235.57 91.49 58.34

1995 26 163.71 258.48 97.03 65.55

1996 26 171.75 256.38 101.98 69.35

1997 26 180.93 268.84 108.85 71.06

1998 26 170.36 260.88 115.46 72.71

1999 26 181.12 265.50 119.65 73.89

2000 26 188.19 281.15 126.43 71.61

2001 26 207.77 287.09 136.17 74.06

2002 26 255.50 284.09 142.96 77.47

2003 26 278.48 304.84 159.06 81.61

2004 26 325.67 339.56 119.45 51.46

Source: Authors’ own calculation.



capita and the output of crude steel per capita (or official reported appar-

ent consumption) across provinces in 2004. As the table shows, the pat-

terns of two data series are quite different. There are numerous reasons for

this divergence. First, the output of crude steel per capita – i.e. the depend-

ent variable in Equation (3.1) – clearly contains information from the

production perspective that is not captured by the demand- related factors

in the regression, including province- specific and time- specific supply- side

0

20

40

60

80

100

120

140

160

180

1979

Pre

dict

ed in

dust

rial d

eman

d fo

rcr

ude

stee

ls (

kilo

gram

s pe

r ca

pita

)

1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003


Figure 3.2 Average estimated demand for crude steel, 1979–2004

(kg/person)

0

200

400

600

800

1000

1200

1400

1600

Beijing

Pro

duct

ion

and

cons

umpt

ion

ofcr

ude

stee

l (ki

logr

am p

er c

apita

) Estimated consumption of crude steel per capita Production of crude steel per capita

Hebei

Inne

r Mon

golia Jil

in

Shang

hai

Zhejia

ng

Jiang

xi

Henan

Hunan

Guang

xi

Yunna

n

Gansu

Ningxia


Figure 3.3 Estimated core demand and actual production of crude steel

for Chinese provinces in 2004 (kg/person)



factors captured by ui. Second, the core estimates will underestimate total

consumption to the extent that the independent variables in Equation (3.1)

are not the sole determinants of provincial demand. Third, disparities in

exports across provinces will account for some of the divergence between

total output and core demand (Liu et al., 2008). Critically, however, the

reasons for the divergence between our consumption estimates and the

actual production are not relevant to our key research question. What

is important to note is that our methodology here yields time- series data

that reflect the three dominant factors which not only determine steel con-

sumption in China but are also key factors in the economic development

Table 3.6 Production and estimated core demand for crude steel by

province, 2004 (kg/person)

Province Number of years Production Estimated core demand

Beijing 26 553.4 229.9

Tianjin 26 724.8 157.1

Hebei 26 828.5 115.6

Shanxi 26 355.2 120.7

Inner Mongolia 26 262.8 97.1

Liaoning 26 615.5 150.0

Jilin 26 152.8 87.9

Heilongjiang 26 62.4 118.9

Shanghai 26 1348.5 234.3

Jiangsu 26 299.0 186.2

Zhejiang 26 85.0 184.7

Anhui 26 151.5 87.3

Jiangxi 26 174.6 84.6

Shandong 26 202.5 169.6

Henan 26 100.3 102.5

Hubei 26 224.8 107.9

Hunan 26 120.1 87.2

Guangdong 26 91.9 188.0

Guangxi 26 65.4 65.7

Guizhou 26 52.1 64.1

Yunnan 26 79.1 74.9

Shaanxi 26 59.6 100.7

Gansu 26 106.4 69.6

Qinghai 26 632.7 65.8

Ningxia 26 991.6 67.6

Xinjiang 26 126.9 88.0




process that we are attempting to connect to that steel consumption.

Moreover, the methodology reduces the impact of the supply- side prob-

lems mentioned above. For example, as shown in Figure 3.2, Inner

Mongolia, Ningxia and Hebei, which have relatively high output of crude

steel due to their large- scale planned steel industries, have relatively low

per capita core consumption using our methodology.

DISPARITIES IN ECONOMIC DEVELOPMENT AND CRUDE STEEL DEMAND PER CAPITA

We use the provincial estimates detailed above to analyse the relation-

ship between the core demand for crude steel per capita and the level of

regional economic development. The basic model is based on a Kuznets-

curve function, where the demand for crude steel per capita is determined

by income per capita as shown in Equation (3.3):

ln (DSteelˆit) 5 g0 1 g1

[ln (GDPit)] 1 g2

[ln (GDPit) ]2 1 uit, (3.3)

where ln (DSteelˆit) is the logarithm of estimated crude steel consumption

per capita, and GDPit is provincial GDP per capita. To capture the pos-

sible non- linear relationship between crude steel demand and income per

capita, a squared term of ln (GDPit) is also included. g0 is the constant

and uit is the residual.

Although pooled ordinary least squares (OLS) can be used to estimate

Equation (3.3), it has been criticized for giving rise to two econometric

problems. First, there is the omitted variable problem. In addition to

provincial economic development, there are many other factors that may

affect provincial- level core demand for crude steel, such as provincially

varying government policies and history. If these factors are positively

(negatively) correlated to the level of GDP per capita, then the estimated

coefficients on these variables will be overestimated (or underestimated).

Second, there is a potential mis- specification problem. Core crude steel

demand in China may be related to many unobserved provincial char-

acteristics, such as specific industrial structures and local fiscal policies.

Even if these characteristics are well controlled from the perspective of

omitted variables, we may encounter a fake inverted- U- shaped relation-

ship between crude steel consumption per capita and income per capita

when the data are pooled together. For example, Guangdong province

is dominated by light industry and thus its steel consumption level would

be lower than that of Liaoning where heavy industry is dominant, all

else being equal. If unaware of this problem, we would observe that



Guangdong’s higher per capita income is associated with lower steel

consumption – contributing to the downturn in the inverted U – when

lower steel consumption actually had more to do with the dominance of

light industry in Guangdong instead.

To deal with the omitted variable problem, we assume that there are

province- specific unobservable factors (that can be either time- invariant

or time- variant), and adopt the first- difference (FD) regression technique.

Although other methods such as the panel data regression with random

or fixed effects can also be used to solve the problem of omitted variables

from a theoretical perspective, the FD model is likely to be more appropri-

ate in this case, given that crude steel consumption as it relates to indus-

trialization and economic transition is more likely to be associated with

region- specific characteristics rather than changing frequently over time,

since it is a long- term issue (as in Chenery et al., 1986). We confirm that

this is the case using appropriate statistical tests13 and therefore opt for the

FD model, which results from rearranging Equation (3.3) as:

d ln (IndDSteelˆit) 5 g0 1 g1d [ln (GDPit

)] 1 g2d [ln (GDPit) ]2 1 uit, (3.4)

where uit represents latent variables which cannot be observed but are

related to core crude steel consumption. To deal with the mis- specification

problem, we test the model specification both with and without the square

term of GDP per capita, which confirms that the appropriate specification

does include the square term.14

The estimates for the pooled OLS and FD models are presented in Table

3.7. Columns (1) and (2) provide the estimated results from the pooled

OLS regression (adjusted for heteroscedasticity and provided for com-

parative purposes). While the estimated coefficients suggest an inverse-

U- shaped curve, they cannot be relied on for the reasons discussed above.

Thus, the FD regression technique is used instead and the results from this

regression are presented in columns (3) and (4). Based on the estimated

results from the FD model, the coefficient in front of GDP per capita and

its squared term are positive (2.613) and negative (−0.197), and both sig-

nificant at the 1 per cent level. This result, combined with the comparison

between the fitness of the specifications with and without the square term

of GDP per capita, demonstrates the existence of an inverse- U- shaped

relationship between GDP per capita and industrial demand for crude

steel per capita. The estimated turning point for the core demand for crude

steel at the provincial level on average in China is US$4728 at 2000 prices.

Although the above discussion provides some useful information on the

relationship between core crude steel consumption and GDP per capita at

the provincial level, the econometric results are valid only from an average



perspective. To further identify the impact of regional disparities on the

relationship between crude steel consumption and GDP per capita, we

incorporate two dummy variables representing Eastern China (includ-

ing Beijing, Shanghai, Guangdong, Hebei, Jiangsu, Liaoning, Shandong,

Tianjin and Zhejiang) and Western China (including Gansu, Guizhou,

Ningxia, Qinghai, Shaanxi and Yunnan), respectively, which we interact

with GDP per capita. Our hypothesis is that if the relationship between

the core crude steel consumption per capita and GDP per capita are sig-

nificantly different across regions, these dummy variables and their inter-

action terms with GDP per capita will be statistically significant. In the

model specification, we incorporate both the solo dummy variables and

their interaction terms. However, the FD regression method eliminates the

former as they do not change over time. This means that we are unable to

ascertain whether there are differences in initial steel consumption across

provinces. The coefficients in front of the interaction terms capture the

difference in the marginal impacts of GDP per capita on the core crude

steel consumption per capita. Thus, Equation (3.4) can be rearranged as:

d ln (IndDSteelˆit) 5 g0 1 g1d [ln (GDPit

) ] 1 g2d [ln (GDPit) ]2

1 g5d [Dummy_East * ln (GDPit) ] 1 g7d [Dummy_West * ln (GDPit

) ] 1 uit,

(3.5)

Table 3.7 Core provincial demand for crude steel per capita, 1979–2004

Pooled OLS First-difference

No square

term

With square

term

No square

term

With square

term

ln GDP78 1.101*** 2.454*** −0.126 2.621***

(0.014) (0.181) (0.092) (0.588)

ln GDP78sqr – −0.098*** – −0.196***

– (0.013) – (0.042)

Constant −3.508*** −8.106*** 0.068*** 0.069

(0.100) (0.622) (0.008) (0.008)

No. of observations 676 676 650 650

Adjusted R-squared 0.94 0.944 0.077 0.077

Note: ***, ** and * represent significance at the 1, 5 and 10 per cent levels respectively; the numbers in parentheses are standard errors.




where the reference group is assumed to be the central region. Dummy_East

takes a value of 1 if the province is in the eastern region and 0 elsewhere.

Dummy_West takes a value of 1 if the province is in the western region and

0 elsewhere.

As expected, the relationship between the underlying demand for crude

steel and GDP per capita does indeed vary significantly across regions. As

shown in Table 3.8, the coefficient on the interaction between the regional

dummy and GDP per capita from the FD model is positive for the eastern

region and negative for the western region, and both coefficients are signifi-

cant at the 1 per cent level. This result implies that the marginal underly-

ing demand for crude steel is significantly higher in the eastern region and

lower in the western region, which results in different turning points across

regions. According to our estimates, the turning point for the eastern region

measured at 2000 constant prices is US$5400, compared with US$4728 and

US$4053 for the central and western regions, as shown in Figure 3.4.

Table 3.8 Provincial demand for crude steel per capita with regional

disparities, 1979–2004

Pooled OLS First-difference (FD)

No square

term

With square

term

No square

term

With square

term

ln GDP78 1.106*** 3.290*** −0.035 3.623***

(0.026) (0.280) (0.280) (0.920)

ln GDP78sqr – −0.163*** – −0.272***

– (0.021) – (0.069)

Dummy_Eastern 0.672*** −0.828*** – –

(0.236) (0.297) – –

Dummy_Eastern *

ln GDP78

−0.113*** 0.106** 0.091 0.462***

(0.034) (0.043) (0.145) (0.144)

Dummy_Western −0.306 0.276 – –

(0.318) (0.314) – –

Dummy_Western *

ln GDP78

0.091* 0.000 −0.332 −0.482*

(0.049) (0.048) (0.239) (0.243)

Constant −3.553*** −10.722*** 0.064*** 0.065***

(0.174) (0.936) (0.007) (0.007)


Adjusted R-squared 0.897 0.905 0.055 0.115

Note: ***, ** and * represent significance at the 1, 5 and 10 per cent levels, respectively. Numbers in parentheses are standard errors.




SIMULATING TRENDS IN CHINA’S NATIONAL CRUDE STEEL DEMAND

As Figure 3.4 illustrates, the relationship between per capita income and

per capita core steel consumption differs across regions. This implies that

the relationship at the national level will depend critically on the relative

income growth rates across provinces. Ignoring this is problematic in

terms of projecting China’s future aggregate demand for steel. To under-

stand how national- level trends will be affected by this range of disparities,

we use the estimates that determined the regional patterns in Figure 3.4 to

consider three hypothetical inter- regional growth scenarios.

Constant Growth Across Regions

If we assume that per capita income growth is the same across all regions,

say 8 per cent per annum, by applying this growth rate to each province

given initial income levels in 2004, we can determine how the regional and

national levels of per capita income will change over time. We then take

the per capita income level for each region in each year, beginning in 2005,

and use the regional estimates on the coefficients of per capita income and

its squared term to determine per capita core steel consumption for each

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

675

GDP per capita (US$ 2000 constant price)

Cru

de s

teel

con

sum

ptio

n (t

on p

er c

apita

) Centre Region Eastern Region Western Region

Latest actual, 2004: US$1899 per capita

1351 2026 2702 3377 4053 4728 5404 6079 6755 7430


Figure 3.4 Simulated turning point for regional core crude steel demand

in China



region. Then we take the weighted sum of these to determine national per

capita core steel consumption (where the weights reflect regional popu-

lations, which we assume will stay constant over time). This process is

repeated for each year through to 2020, with the aggregate projection illus-

trated in Figure 3.5. Even with constant income growth across regions,

what we expect and what we observe is a path that is quite different from

the national average projection depicted in Figure 3.5.

Convergence of Regional Per Capita Incomes (the ‘Effective Western

Development Strategy Scenario’)

In this case we assume instead that per capita income growth in the eastern

region stagnates – for simplicity, at zero per cent per annum – while the

western region records the most rapid growth at 8 per cent per annum and

the central region’s growth lies somewhere in between, at say 4 per cent per

annum. As Figure 3.5 shows, more rapid growth in the west generates a

much steeper trajectory and indicates that the aggregate turning point may

happen much later – that is, at a higher level of per capita income. This

result is intuitive given that the western provinces are much lower down on

their own curves and so are yet to experience much of the rise in per capita

steel consumption. Their rapid growth ensures that this rise happens

quickly. Simultaneously, slow (zero) growth in the east reduces the pace at

7.0

7.5

8.0

8.5

9.0

9.5

10.0

10.5

11.0

11.5

12.0

1800

Pro

ject

ed c

onsu

mpt

ion

of c

rude

ste

el(1

00 k

g pe

r ca

pita

)

E8 C8 W8 E8 M4 W0 E0 M4 W8

2300 2800 3300 3800 4300


Source: Authors’ calculation.

Figure 3.5 Projected aggregated relationship between crude steel

consumption and GDP per capita, 2004–2013



which eastern provinces reach their own peak and hence extends the per

capita income aggregate turning point for declining per capita core steel

demand.

Ongoing Divergence of Regional Per Capita Incomes (the ‘Failed Western

Development Scenario’)

Here, we assume that per capita income growth in the eastern region

outpaces the rest of the country, at 8 per cent, compared with 4 per cent

for the centre and 0 per cent for the west. In contrast with the above sce-

narios, more rapid growth in the east precipitates an earlier aggregated

turning point, as these provinces are already further along their own curve.

Simultaneous, slower growth in the west dampens their contribution to the

upward trend in per capita steel consumption.

Given that there are an infinite number of possible combinations of

provincial growth rates in the future, it follows that the national aggregate

path of per capita steel consumption could essentially take on any kind

of shape in the next few decades – possibly but not necessarily reaching

the turning point, and possibly but not necessarily concave. How, then,

can our empirical analysis be used to understand China’s aggregate crude

steel demand? To demonstrate, consider the relationship between Chinese

demand for crude steel and GDP per capita at the national level in the

period since the late 1970s, a trajectory which shows an increasing trend

with no sign of a turning point (Figure 3.6(a)). One could attempt to make

use of these time- series data to predict the future demand for crude steel

consumption per capita, simply by extrapolating along the past trendline,

which is done in Figure 3.6(b). This trend seems to be inconsistent with

our previous discussion that the core crude steel demand per capita may

decrease as GDP per capita increases (in Figures 3.4 and 3.5). However, the

inconsistency can be easily explained if we consider the impact of regional

disparities and aggregation. In particular, were we to use the past rates of

growth of income across provinces in our model, the increasing trajectory

of crude steel demand per capita could be simulated.15 Similarly, fore-

casts of per capita income growth that take into account interprovincial

growth disparities will provide a more accurate prediction for the future,

given that the aggregate- level relationship between crude steel consump-

tion per capita and GDP per capita as projected using past national- level

data (as in Figure 3.6(b)) does not provide any information regarding the

turning point for any province or for the nation on the whole.



CONCLUSIONS

Rapid economic growth and development since the late 1970s has her-

alded significant changes in the structure of the Chinese economy. China’s

underlying demand for crude steel has recorded strong growth as a result

of this structural change, and in particular as a result of the twin processes

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

1800

GDP per capita (RMB) 1978 constant price

Cru

de s

teel

con

sum

ptio

n(1

00 k

g pe

r ca

pita

)

2000 3000 4000 5000 6000 7000 8000

0

50

100

150

200

250

300

350

400

450

0

Con

sum

ptio

n of

app

aren

tcr

ude

stee

l (m

mt)

500 1000 1500 2000


a

b

Source: China Iron and Steel Statistical Yearbook, various years; authors’ projections.

Figure 3.6 Apparent core consumption of crude steel and GDP per capita

in China. (a) Historical relationship between national- level

apparent consumption of crude steel and GDP per capita,

1960–2004. (b) Simulation of the relationship between

national- level core apparent consumption of crude steel and

GDP per capita after 2006



of industrialization and urbanization. Whether and when the consequent

rise in China’s per capita steel consumption will cease is a question of

importance not only for Chinese policy- makers and steel- makers, but for

all participants in the global steel market as well. Projecting future trends

in Chinese steel demand is complicated by the fact that China is charac-

terized by significant inequalities across its 31 provinces, many of which

are the size of large countries themselves, and further by the fact that

accurate time- series provincial- level data for steel demand are not readily

available.

The method adopted in this chapter has utilized provincial- level data

on steel production, in combination with three factors – the levels of

industrialization, urbanization and fixed- asset investment – to estimate a

time series of provincial- level core steel consumption or underlying steel

demand. We claim not that these estimates are precise, but rather that

they capture the dynamics of potential core steel consumption disparities

across regions (as determined by the major processes of structural change

that accompany ‘modern economic growth’). It is this component of per

capita core steel demand that is most likely to follow the Kuznetsian

inverted- U- shaped path as per capita income rises.

Armed with these provincial estimates, our econometric results con-

firmed that such a path exists at the provincial level in China. Given the

imprecision of our estimates, we attempted to provide a precise projection

neither of the future total demand for crude steel in China nor of when

the turning point for per capita steel demand might be reached. Instead,

we offered simulations to demonstrate that these potential estimates

would necessarily be influenced by the relative growth performances of

different provinces, because of the vast disparities in levels of develop-

ment across those provinces and hence their very different positions along

the predicted path. Failing to take these disparities into account risks

failing to understand just how dependent the future national- level trend

in Chinese steel demand will be on future patterns of interprovincial

economic growth. We hope this chapter has provided a small step in the

right direction.

NOTES

1. See Chapter 2 for a full discussion on this point. 2. See also Syrquin (1988). 3. Authors’ calculations based on GDP data from Maddison (2007). 4. This analysis excluded countries with special features, such as those with superior

natural resources and energy supplies, because it seemed clear that these countries were unlikely to follow even the general patterns observed elsewhere.



5. All values throughout this chapter are reported in constant US dollars based on the national accounts and current exchange rates.

6. This paragraph is based on Crompton (1999). 7. Data in this paragraph are drawn from the IMF. The information for the UK,

Germany and Chinese Taiwan comes from the International Iron and Steel Institute (2007). All per capita incomes reported are in 2000 prices.

8. Based on the above discussion, and also on Liu et al. (2008), there are three main factors affecting the per capita demand for crude steel in China: industrialization, urbanization and the fixed- asset investments that result from economic growth. The importance of the last of these three factors stems from the observation that levels of investment will increase as the industrial structure becomes more capital- intensive and as the demand for infrastructure associated with urbanization rises. While other factors, such as con-sumer preferences and the availability of substitutes for steel will also play a role in determining steel demand, Liu et al. (2008) estimate that the above three factors deter-mine between 60 and 70 per cent of crude steel consumption in China at the aggregate level and are largely representative of the demand- side perspective, which can be used to index changes in steel consumption.

9. See Chapter 5 on the point that the steel industry is not particularly well suited to export orientation, but rather that production tends to concentrate in countries where demand is high. Given that many – indeed most – of China’s provinces are themselves the size of large countries, this point carries over to the provincial level, suggesting that most provincial steel production will be consumed ‘domestically’, that is, within the province.

10. Sichuan (because Chongqing started to split in 1995), Fujian, Hainan and Tibet (the latter three have no complete data sets).

11. We also run a fixed- effect regression to test our model specification, and the Hausman test (used to test the null hypothesis that the fixed- effect model is not preferred to the random- effect model, is not rejected at the 10 per cent level) leads us to conclude that the estimates using the model with random effects are more accurate.

12. In 2004, macroeconomic adjustment policies from the central government reduced the industrialization, urbanization and fixed- asset investment measures for all provinces, which caused significant declines in their underlying demand for crude steel. This makes 2004 an outlier.

13. In particular, the Hausman test and the Breusch–Pagan test.14. If Guangdong’s lower per capita steel consumption had more to do with its industrial

structure than its higher level of per capita income, there would be no reason for Guangdong’s data to follow an inverted- U shape rather than a linear path. The fact that we find that the squared term is significant confirms that this is not an issue we need to worry about. We also run a random coefficient regression, which aims to identify the local curvature of the marginal contribution of per capita GDP to per capita steel consumption for each individual province. This confirms that the inclu-sion of the squared term is appropriate as all provinces follow an inverse- U- shaped trajectory.

15. This would take a substantial amount of effort as we would need information not only on provincial GDP growth rates across time but also on the simulated starting point for each province (which is not available because the FD regression eliminates the solo dummy variable), and so on. This is beyond the scope of this chapter, but the key point remains that the past combination of provincial growth rates should generate a trajec-tory that matches Figure 3.6, given that national steel demand is, by definition, the sum of provincial steel demand at any point in time.



REFERENCES

Chenery, H. and M. Syrquin (1975), Patterns of Development: 1950–1970, Oxford: Oxford University Press.

Chenery, H., S. Robinson and M. Syrquin (1986), Industrialisation and Growth: A Comparative Study, World Bank research publication, New York: Oxford University Press.

China Iron and Steel Association (various years), China Iron and Steel Statistical Yearbook, Beijing: China Iron and Steel Association Press.

China National Bureau of Statistics (CBNS) (2010), China Comprehensive Data Collection 55 Years: 1949–2004, Beijing: China Statistical Press.

Crompton, P. (1999), ‘Forecasting steel consumption in South- east Asia’, Resources Policy, 25 (2), 111–23.

International Iron and Steel Institute (2007), Steel Statistical Yearbook, Brussels: International Iron and Steel Institute, accessed at www.worldsteel.org/.

Kuznets, S. (1965), Economic Growth and Structure: Selected Essays, London: Heinemann.

Liu, H., H. He, L. Chen, L. Ma, Y. Zheng, W. Yuan and H. Shi (2008), ‘Prediction on China’s demand for iron and steel in the medium- and long run’, research report for National Development and Reform Commission, China Steel Industry Development Research Institute, Beijing, June.

Maddison, A. (2007), ‘Contours of the world economy and the art of macromeas-urement 1500–2001’ International Association for Research in Income and Wealth Ruggles Lecture’, accessed at www.ggdc.net/maddison/.

National Bureau of Statistics (2007), China Statistical Yearbook, Beijing: China Statistics Press.

Syrquin, M. (1988), ‘Patterns of structural change’, in H. Chenery and T.N. Srinivasan (eds), Handbook of Development Economics, vol. 1, Amsterdam: North- Holland, Chapter 7.

World Bank (2006), World Development Indicators Database, accessed at http://publications.worldbank.org/WDI/.

Zhao, M. and Y. Zhang (2008), ‘Development and urbanisation: a revisit of Chenery–Syrquin’s patterns of development’, Annals of Regional Science, accessed at www.springerlink.com/content/5281486103684772/fulltext.pdf.


69

4. China’s iron and steel industry performance: total factor productivity and its determinants


INTRODUCTION

The rapid expansion of China’s iron and steel industry (hereafter ‘the

industry’) since early in the twenty- first century has been remarkable in

terms of both speed and scale. Yet there is an issue regarding the ‘quality’

of the industry’s expansion – was the rapid growth driven primarily by

increases in inputs or by gains in productivity? There is no consensus as to

which factors have been more important for driving the current wave of

the industrial expansion. However, a more sustainable and healthy devel-

opment of the industry should be based on the continuation of firm- level

productivity growth – a representation of both technological progress and

efficiency improvement. Examining the change of firm- level productivity

and its determinants over the past decade therefore becomes an important

empirical question.

There have been many attempts made to quantify the productivity of

China’s iron and steel firms and its determinants by using microeconomic

(firm- level) data. Jefferson (1990) was the first to estimate the total factor

productivity (TFP) of the industry by using a log–linear function with

cross- sectional data from 120 large and medium- sized enterprises (here-

after LMEs) in 1986. Kalirajan and Chao (1993) and Wu (1996) adopted

the stochastic frontier analysis to distinguish between firms’ technical

efficiency and their technological progress using cross- sectional and panel

data of LMEs before 2000, respectively. Zhang and Zhang (2001) exam-

ined the technical efficiency of China’s iron and steel firms in the 1990s

using data envelope analysis, and Ma et al. (2002) and Movshuk (2002)

focused on the ownership reform undertaken in the late 1990s and its

impact on firms’ TFP in the industry after 2000.

These studies have provided important insights into the changes of

firms’ productivity in the industry and its determinants in the past.



However, their results proved quite diverse with respect to whether the

industry’s productivity and/or efficiency had increased or not over the

periods considered. For example, Zhang and Zhang (2001) found that

the average technical efficiency of China’s iron and steel firms had been

increasing in the 1990s, while Ma et al. (2002) and Movshuk (2002) found

to the contrary.

There are three possible explanations relating to both the methodology

and data issues for this inconsistency. The first is that studies estimat-

ing productivity via the stochastic frontier method (or the data envelope

analysis method) focus on technological efficiency by assuming that the

best- performing firms are at the production frontier. This assumption is

likely to generate results that are sensitive to sample choices. The second is

that LMEs (usually state- owned) were dominant across all samples (owing

to data availability). This means that some important information on

the prolific small- and- private- enterprise (hereafter SE) sector is excluded

from these studies. The third is that by utilizing data covering the period

from the late 1980s to the late 1990s, during which time many reforms in

the industry had not been fully implemented, or were yet to bear fruit,

these studies might therefore not have been able to capture more fully the

consequence of the reform. Thus it may not be surprising that the earlier

studies generated ambiguous results with respect to the impact of reform

on industry productivity.

This chapter seeks both to improve on the methods used in the previous

studies and to update the data set. We use some newly developed econo-

metric techniques to re- estimate Chinese iron and steel firms’ TFP by

using firm census data over the period 1998–2007. The approach adopted

here includes the Olley and Pakes (1996) and Levinsohn and Petrin (2003)

two- step method, and the generalized method of moment (GMM) method

proposed by Ackerberg et al. (2005) and Wooldridge (2009) to estimate

the industry’s production function with the gross output assumption.

These methods overcome the ‘endogeneity bias’ due to potential cor-

relation between capital usage and unobserved productivity (caused by

the assumption of exogenous inputs – capital – that plagues traditional

analysis) (Jefferson, 1990). As to the update of data, we believe that the

firm- level census data for the industry are the most recent data ever incor-

porated into a study of this type.

Three questions are to be addressed. First, how has firm- level pro-

ductivity in the industry changed over time? Second, what are the major

driving forces behind firm- level productivity growth in the industry over

the decade from 1998 to 2007? Third, are there any significant differences

in productivity growth among iron and steel firms with different char-

acteristics such as firm size, ownership type and geographical location?


Total factor productivity and its determinants 71

Answers to these questions show that productivities of firms of different

types in the Chinese iron and steel industry are not only different over the

period of 1998 to 2007, but also sensitive to different measures adopted

in carrying out the economic reforms in the industry. This implies that

further improvement in the productivity and quality of output of the

Chinese iron and steel industry may be enhanced by a range of policy

instruments targeting firms with different characteristics in the process of

restructuring the industry.

The remainder of the chapter is arranged as follows. The next section

describes briefly the development of China’s iron and steel industry

over the reform period. Some factors associated with changes in firms’

productivity in the industry, such as marketization reform, government-

sponsored investment and intensified competition, have been addressed.

The section that follows presents the model specification and the two- step

approach for estimating firms’ TFP and identifying its determinants. The

semi- parametric TFP estimation techniques and its related literature are

highlighted for their importance in dealing with the problem of ‘endoge-

neous input choice’. Data collection and summary statistics are presented

in the next section. The penultimate section discusses the estimation

results and the section after that concludes.

CHINA’S IRON AND STEEL INDUSTRY AND ITS MICROECONOMIC PERFORMANCE

While China’s iron and steel industry grew along with the rest of the

economy in the first decade of the reform era beginning in the late 1970s,

it was not until the early 1990s that the sector began to expand at a dra-

matic rate. During the period 1990–2007, China’s production of iron ore,

pig iron and crude steel has increased from 179 million tonnes (mt), 62 mt

and 66 mt to 582 mt, 404 mt and 422 mt, respectively, representing average

annual growth of 7.6 per cent, 12.4 per cent and 12.3 per cent. China’s

output of iron ore and crude steel rose to above one- third of the global

total, while its pig iron output rose to about half of world production.

The rapid expansion of output in the industry has been accompanied by

a significant industrial structural adjustment, characterized by a substan-

tial increase in the number of enterprises and an enlargement of scale at

individual firm level. The total number of firms in the industry increased

from 1589 in 1990 to 11 596 in 2007, while the average real output value

per firm (at 1990 constant prices) increased from US$17.2 million in 1990

to US$32.5 million in 2008.1 As a consequence, competition among firms

in the industry has been intensified and firms’ productivity has increased



rapidly over time. Figure 4.1 shows the positive relationship between the

real output value of China’s iron and steel industry (at 1990 constant

prices) and its average labour productivity between 1985 and 2006.

There are three factors that seem most relevant for assessing the rapid

increase of firms’ productivity. First, marketization reforms rendered

more autonomy to enterprises (especially state- owned ones), thereby

helping to increase their production efficiency. Second, the rapid increase

in fixed investment and the associated boost to average production capac-

ity has helped to foster firm- level technological progress. Third, the free

entry of SEs (motivated by profit incentives) reduced LMEs’ market

power and intensified competition in the industry. We consider each of

these factors in turn.

First, the iron and steel industry in China has historically been domi-

nated by the large, integrated state- owned enterprises (hereafter SOEs). By

integrated enterprises, we mean the ferrous metals firms, which produce all

items across the spectrum from iron ore to finished steel, rather than those

which specialize in producing a single product. In 1990, there were a total

of 1589 iron and steel enterprises in China, among which 163 were state-

owned or state- controlled. In terms of output value, the SOEs accounted

for more than 80 per cent of the industry total. Given that the SOE struc-

ture imposed a heavy burden on these firms in the form of non- productive

spending such as housing, pensions and other welfare expenses, this pro-

vided scant executive incentive to pursue productivity gains. This is the

chief reason why the management efficiency of those enterprises was weak.

0

50 000

100 000

150 000

200 000

250 000

0

50 000

100 000

150 000

200 000

250 000

300 000

1985

Labo

ur p

rodu

ctiv

ity (

yuan

per

per

son)

Out

put v

alue

(10

00 m

illio

n yu

an)

Output value (1990 constant price)

1990 1995 2000 2005 2006

Source: CISI (2008).

Figure 4.1 Output value and labour productivity in China’s iron and steel

industry, 1985–2006 (1000 million yuan; yuan/worker)



Since the early 1990s, a series of microeconomic reform policies aiming

to promote the marketization of SOEs have been implemented. These

include the reform of the profit distribution system; the provision of

incentives for increasing productivity; reform of the management system;

market- based reform, especially with respect to pricing; introducing

foreign direct investment; and a free entry policy. The most recent reform

is what has been termed the ‘modern enterprises system’ and ‘shareholding

structure reform’, which began in the early 2000s and is still underway for

a few very large enterprises. These reforms make the SOEs more indepen-

dent of the government with respect to both financial arrangements and

managerial appointments. In 2006, the share of output volume accounted

for by SOEs had fallen to 43.1 per cent while the number of SOEs had

declined to 67 (accounting for 5.2 per cent of the total number of firms). As

a consequence of these changes, productivity and management efficiency

at firm level have been improved.

Second, the rapid increase of investment in new enterprises and the

accompanying technological changes has assisted productivity gains at the

firm level. The industry has been characterized historically by a mixture

of old and advanced production technologies, with the weighted average

level of technology lagging far behind the conditions in the industrialized

countries. In 1990, the average continuous casting ratio in China’s iron

and steel industry was only 22.2 per cent, which is far less than the ratio of

above 95 per cent in the other main steel- making countries. Around 15 per

cent of crude steel was still being produced in open- hearth furnaces (OHF)

in China, which have effectively been scrapped in most steel- producing

countries.

To catch up with the world leaders in producing steel, a very large

amount of capital has been invested in production technology since the

1990s. These investments have been jointly funded by the central and pro-

vincial governments and a considerable amount of investment has come

from SOEs themselves and private sources. Between 1990 and 2005, the

average annual fixed- assets investment in China’s iron and steel industry

increased from US$2.7 billion to US$31.5 billion yuan (the exchange

rate used for deflating the series comes from China Statistical Yearbook

(CNBS, 2009)). This massive increase in investment has substantially

improved the standard of the industry’s production technology. In 2005,

the average continuous casting ratio had increased to 94 per cent and

crude steel produced from basic oxygen furnaces (BOF) and electric

arc furnaces (EAF) accounted for 88.1 per cent and 11.7 per cent of the

total, respectively. Such rapid improvements in production technology

imply that there should be significant gains in firm and industry levels of

productivity.



Third, the intensified competition due to free entry of SEs and its

associated reallocation of market share and resources within the industry

has favoured those with advanced production technology, promoting

productivity growth in the quest for profits. The industry is believed to

be one of the few sectors that can realistically expect increasing returns to

scale, given the large amount of sunk costs embedded in any steel enter-

prise. Thus, firms aiming to obtain higher productivity through increasing

returns to scale must seek to achieve both gains in market share and the

expansion of productive capacity, as well as securing additional access to

intermediate inputs including finance. With the increased number of firms

in the industry, the share of national crude steel production accounted for

by the top eight firms between 1998 and 2007 declined from 33 per cent to

17.9 per cent; the Herfindahl index of industrial concentration at the three-

digit level (defined as the squared share of the top eight firms’ market

sales revenue in the total revenue of the industry; see Brown and Warren-

Boulton, 1988) accordingly decreased from 33 in 1998 to 22.3 in 2007.

The discussions provide some background information on firm- level

productivity change and its potential drivers in the industry. The next step

is to detect the trend of the firm- level productivity and to identify the main

factors which determine the trend. In the following section, we start with

estimating firms’ TFP by using the newly developed endogeneous input

usage method.

MODEL SPECIFICATION: ENDOGENOUS INPUT USAGE AND FIRMS’ PRODUCTIVITY

Estimating productivity as a residual after accounting for measurable

inputs and then decomposing that TFP into its proximate determinants

is a long- standing preoccupation of empirical economists, going back to

the seminal paper by Solow (1957). While the Solow ‘growth account-

ing’ framework has been widely applied for carrying out economy- wide

analysis, the technique is easily adapted to microeconomic analysis. The

standard approach is to assume a Cobb–Douglas, quadratic or translog

production function with an additive, time- consistent firm effect and to

solve the unobserved endogeneity problem by using fixed- effect general

least squares (GLS) estimation. Unfortunately, the fixed- effect estimator

still assumes strict exogeneity of the inputs (that is, labour, capital and

various intermediate inputs), which is conditional on firms’ heterogeneity

in productivity (Wooldridge, 2002). This assumption requires that inputs

must not be chosen in response to productivity shocks – a severe and

unrealistic restriction on firm behaviour.



To deal with this problem, econometricians have resorted to using

the instrumental variable method (say, using lagged inputs as instru-

ments for inputs) to relax the strict exogeneity assumption for inputs.

For example, Arellano and Bover (1995) and Blundell and Bond (2000)

use this approach to correct the estimation of their production functions.

Although this method works well in some cases, it is open to two criti-

cisms. The first is that introducing lags into the regression (or differenc-

ing) removes much of the variation in the explanatory variables and can

exacerbate the measurement error of the inputs. The other is that the

instruments available after differencing are often only weakly correlated

with the differenced explanatory variables.

Olley and Pakes (1996; hereafter OP) arrived at an alternative way

to deal with the endogenous input problem. Rather than allowing for

time- constant firm heterogeneity, OP show that, under certain assump-

tions, investment can be used as a proxy variable for unobserved, time-

varying productivity. In other words, productivity can be expressed as an

unknown function of capital and investment (when investment is strictly

positive). This, for the first time, took the simultaneity problem explicitly

into account when estimating a production function by introducing an

estimation algorithm. Following this innovation, Levinsohn and Petrin

(2003; hereafter LP) later proposed a modification of OP’s method to

address the problem of lumpy investment. They suggested the use of

intermediate inputs as a proxy for unobserved productivity, a method that

generated a better result than the use of an investment variable.

Generally, both the OP and LP methods suggest a two- step process to

consistently estimate the coefficients on variable inputs. In the first stage,

semi- parametric methods are used to estimate the coefficients on the vari-

able inputs along with the non- parametric function linking productivity to

capital and investment. In a second step, the parameters on capital inputs

can be identified under the assumptions on the dynamics of the produc-

tivity process. Both the OP and LP methodologies have been widely used

in the recent literature on firm- level heterogeneity for derivation of TFP

measures, though the LP method is more preferred to the OP method in

practice since it can save more observations when firms, as is common, do

not carry out long- term investment on an annual basis.

More recently, Ackerberg et al. (2008; hereafter ACF) argued that,

while there are some solid and intuitive identification ideas in the papers

by Olley and Pakes (1996) and Levinsohn and Petrin (2003), their two-

step semi- parametric techniques may suffer from a potential problem

with identification of parameter in the first- stage estimation – if all inputs

(including labour usage) are determined by a TFP shock (and thus opti-

mally chosen by the firm), then they all enter the deterministic function



of unobserved productivity and stated variables. As a consequence,

the coefficient on the variable input is non- parametrically unidentified.

ACF showed that specifying popular functional forms for the produc-

tion process does not help. In fact, in the Cobb–Douglas case (and some

others), labour disappears after substituting unobserved productivity as a

function of inputs (Wooldridge, 2009). This problem is more serious for

the LP estimation since the potential collinearity between intermediate

inputs and labour is usually strong in practice.

To deal with this problem, ACF proposed a hybrid of the OP and LP

approaches, along with the assumptions on the timing of decisions concern-

ing input choice. Specifically, ACF resolved the potential lack of identifica-

tion by using a two- step estimation method that does not attempt to identify

any production parameters in the first stage. Later, Wooldridge (2009)

further extended the estimation method by using a unified GMM estima-

tion, which allows for the possibility that the first stage of OP or LP actually

contains identifying information for parameters on the variable inputs, such

as labour. Since the Wooldridge method is a one- step GMM procedure, it

can use the cross- equation correlation to enhance efficiency, and the optimal

weighting matrix efficiently accounts for serial correlation and heteroscedas-

ticity. Thus, the Wooldridge GMM method for production function estima-

tion is the most preferred regarding its consistency and effectiveness.

In this chapter, we use the Wooldridge GMM method to estimate firms’

TFP, while checking the robustness of our results by using the OP and LP

methods as well as some other traditional measures. For simplicity, we

assume that the production function of China’s iron and steel firms takes

the Cobb–Douglas form with endogenous capital and labour usage.2

Yit 5 Ait L

blit K bk

it M bm

it , (4.1)

where Yit represents the physical output of firm i in period t; Lit, Kit and

Mit are inputs of labour, capital and intermediate inputs, respectively; and

Ait is the Hicks neutral efficiency level of firm i in period t. Taking natural

logs and differentiating the equation yields a linear production function

as follows:

yit 5 ln Ait 1 bllit 1 bkkit 1 bmmit, (4.2)

where lower- case letters refer to natural logarithms and ln (Ait) 5 .it 1 eit,

and .it measures the mean of firm- level TFP over time; eit is the time- and

producer- specific deviation from that mean, which can then be further

decomposed into an ‘observable’ (or at least predictable) and ‘unobserv-

able’ component. Under this assumption, firms’ TFP can be written as:



tfpit 5 yit 2 bllit 2 bkkit 2 bmmit, (4.3)

where bl, bk and bm are estimated using the Wooldridge GMM method as

well as other methods including the OP, LP and ordinary least squares

(OLS) regressions. With these estimation results, the relationship between

firms’ productivity and its determinants, including marketization reform,

changes in market share, exports and so on, can be examined on the basis

of:

tfpit 5 g0 1al

glXlit 1 ui 1 vit, (4.4)

where Xlit is a vector containing the determinants of firms’ TFP; ui is

the firm- specific unobserved effects; and vit is the residual. To estimate

Equation (4.4), the panel data regression technique with random- and

fixed- effect assumptions can be used to account for the firm- specific unob-

served effects.

DATA COLLECTION AND SUMMARY STATISTICS

The data used in this study are taken from the annual firm census carried

out by the National Bureau of Statistics (hereafter NBS) during the

period 1998–2007. The survey covers all enterprises above a designated

size (with annual sales reaching at least 5 million yuan) regardless of

ownership status. Iron and steel firms are defined as the firms registered

with the sector of ‘smelting and pressing of ferrous metals’ (namely, the

thirty- second category according to the two- digit Chinese Industrial

Classification Code). Discarding enterprises with incomplete data left

33 778 observations, which covered 1654 firms in 1998 to 4929 firms in

2007. These firms have accounted for more than 70 per cent of the total

number of enterprises in the industry and their combined output and asset

shares were around 90 per cent of the total. Table 4.1 shows a statistical

summary of these firms. Compared with data used in previous studies,

our sample is more representative, as it covers not only the LMEs owned

by the state but also a large number of SEs and private firms. This helps

reduce the selection bias significantly.

The output of iron and steel firms is defined as the total output value

discounted by the producer price index at the firm level. Our reasoning for

this choice of deflator is that the industry is composed of multiple types

of enterprise with different output structures. The vertically integrated

enterprises produce all of the products in the product chain, including iron

ore, metallurgical coke, pig iron, ferroalloys, refractories and finished steel



products. Others may produce only one or two items in the chain of iron

and steel production. From this perspective, output values are much better

than physical output as a means of comparison. All the output values are

benchmarked to the 1990 price level.

Capital usage is defined as the value of net fixed assets, which is equal

to total fixed assets less accumulated depreciation, deflated using the

fixed asset investment price index for the industry. Although it is argued

that net fixed assets provide a problematic measure of the total capital of

China’s iron and steel enterprises (Jefferson, 1990), there is little we can

do to adjust this due to data limitations at the firm level. Labour usage is

defined as the number of employees working in the industry at the end of

each calendar year rather than the total of all labour employed during the

course of the year. The reason for this is that there is a certain proportion

of employees who are not directly involved in productive activity in the

industry, especially in LMEs.3 In this study, we did not make distinctions

between skilled and unskilled workers, owing to lack of consistent data

over time.

Intermediate inputs are defined as the total output value (current price)

minus value added plus the value added tax, which is consistent with the

approach of the NBS. To eliminate the impact of inflation, a ‘single defla-

Table 4.1 Descriptive statistics of iron and steel enterprises in the sample,

1998–2007

Year Number

of

firms

Total

output value

(billion

yuan,

current

price)

Total

number

of

employees

(million

persons)

Total fixed

capital

assets

net value

(billion

yuan)

Total sales

revenue

from export

(billion yuan,

current price)

Total value

added

(billion

yuan,

current

price)

1998 1654 281.4 20.2 351.3 17.5 71.4

1999 1859 318.4 21.0 373.3 17.2 84.3

2000 2025 400.4 22.0 373.3 27.1 110.3

2001 2297 496.9 21.3 441.6 20.5 110.3

2002 2481 583.1 21.0 474.4 22.8 163.3

2003 2769 861.0 21.5 543.7 30.0 248.6

2004 4898 1338.6 21.9 606.2 63.4 316.7

2005 4952 1757.7 22.9 631.8 96.9 476.1

2006 5158 2097.5 23.6 890.5 150.3 562.0

2007 4929 2784.1 24.7 1055.4 216.8 739.9

Source: Authors’ own calculations based on firm census data from NBS.



tion’ approach, which assumes identical deflators for all intermediate

materials and valued added, is used to adjust the impact of price changes

on the estimation of intermediate input quantity. In other words, a firm-

level ratio of real to nominal gross industrial output is calculated and used

to deflate the intermediate input values.

Finally, we define a series of variables that may reasonably be

expected to have impacted on firms’ productivity. They include: (1) the

Herfindahl index, defined as the squared share of the top eight firms’

sales revenue in the industry total at three- digit- level CICC sectors; (2)

an index for marketization, defined as the share of non- state- ownership

in firms’ real capital; (3) a R&D proxy index, defined as the share of

revenue from selling new products; (4) the scale indices, defined as

dummies distinguishing between small, medium and large firms; and (5)

firms’ export ratio, which is defined as the share of revenue from firms’

exports.

FIRM- LEVEL TFP ESTIMATION AND ITS DETERMINANTS

Table 4.2 reports the estimated production function coefficients for

China’s iron and steel enterprises obtained using different methodologies.

All reported estimates are obtained for the unbalanced panel data during

the period 1998–2007. Each column reports a set of estimators obtained

by using a specific method. The focus is principally on the column headed

GMM with other columns (in particular the OP and LP estimation) for

comparison.

The comparison between the estimated results obtained from using the

non- parametric methods (including OP, LP and GMM) with those from

the OLS, first- differencing and fixed- effects methods shows that coeffi-

cients obtained with the non- parametric methods are lower in magnitude

for both labour and intermediate inputs but higher for capital. In particu-

lar, the marginal contribution of capital is estimated to be significantly

larger than that of labour in the GMM estimation. This implies that

capital usage plays a more important role than labour in the production

of China’s iron and steel firms. This finding is consistent with the key

characteristic of this industry, which is its capital intensity. This also sug-

gests that the problem of ‘endogenous input usage’ applied in the previous

studies reviewed is likely to have caused an underestimation of capital’s

contribution and a corresponding overestimation of intermediate inputs’

contribution to output (noting that the coefficient assigned to labour input

is not changed significantly in GMM vis- à- vis OLS or the fixed- effects



estimation). Thus, the application of the GMM estimation method is

appropriate in this context.

In all of the regressions, the estimated elasticity of intermediate inputs

ranges from 0.89 to 0.94 and is statistically significant at the 1 per cent

level. On average, they account for around 90 per cent of contributions

to the growth of total output. This finding shows that intermediate input

usage plays an important role in the value of production of China’s iron

and steel enterprises. This suggests that China’s iron and steel production

is focusing mainly on producing low- value- added products such as sec-

tions and wires, where output growth has been mainly driven by increasing

material inputs. As an example, long products accounted for 51.9 per cent

Table 4.2 Estimates of the Cobb–Douglas production function of iron and

steel firms with total output, 1998–2007

OLS First-

differencing

Panel (fixed

effects)

OP LP GMM

Dependent variable (ln Y): log of total output value in 1990 constant prices

Log of labour 0.042*** 0.081*** 0.056*** 0.037*** 0.038*** 0.042***

(0.002) (0.007) (0.006) (0.002) (0.002) (0.011)

Log of capital 0.014*** 0.011*** 0.025*** 0.031*** 0.058*** 0.145***

(0.002) (0.003) (0.003) (0.004) (0.006) (0.012)

Log of

intermediate

inputs

0.942*** 0.860*** 0.928*** 0.926*** 0.887*** 0.890***

(0.002) (0.008) (0.006) (0.003) (0.009) (0.010)

Constant −0.324*** 0.045*** −0.360*** – – −0.550***

(0.008) (0.002) (0.027) – – (0.042)

No. of

observations 33 022 22 646 33 022 6173 33 022 33 022

R-squared 0.975 0.82 0.975 – –

Arellano–

Bond test

AR(2) – – – – – −2.04

Saggan/

Hansen

test of

exogeneity

of

instruments – – – – – 2.38

Wald test for

IRTS

Rejected Rejected Rejected Not

rejected

Not

rejected

Not

rejected

Note: ***, ** and * represent the results statistically significant at the 1 per cent, 5 per cent and 10 per cent levels, respectively. Numbers in parentheses are standard errors.




of China’s steel product output in 2005 (CISA, 2008). This figure, though

declining over time, is still far greater than the corresponding share in

Germany (23.8 per cent), the United States (28.7 per cent), Japan (37.8 per

cent) and South Korea (43.3 per cent) more than a decade ago (Labson

et al., 1995). These countries are all important producers of flat products,

where value added in production is much higher. China’s relatively weak

penetration in flat products reflects that the fact that its industry is more

highly input- intensive relative to its relevant peers, especially in those

industrialized countries.

As a rough measure of returns to scale by adding up the elasticities

over all inputs, the GMM estimation shows that the production function

exhibits some characteristics of constant returns to scale or mildly increas-

ing returns to scale as shown by the bottom row in Table 4.2, which is

labelled ‘Wald test for IRTS’ (standing for ‘increasing return to scale’).

Although the hypothesis of constant returns to scale in the estimation with

the OLS, first- differencing and panel fixed- effects methods is significantly

rejected at the 1 per cent level, it is not rejected in the LP estimation (at the

1 per cent level). This suggests that even when SEs are taken into account

for estimating the production function, the industry still exhibits the sig-

nificant characteristic of constant returns to scale, or mildly increasing

returns to scale, when capital is correctly accounted for in the production

function by using the GMM estimation method. Thus, mergers and acqui-

sitions, especially those initiated by market impulses, should be further

encouraged to obtain the potential benefits from economies of scale in the

industry.

Based on the preceding analysis of the estimates under the GMM

method, we can use Equation (4.3) to extract an estimate of firm- level

TFP and examine the determinants. Figure 4.2 show the changes in the

mean and variance of China’s iron and steel firms’ TFP over time. Between

1998 and 2007, there was a significant increasing trend in average produc-

tivity at the firm level, with an annual growth rate of 2.1 per cent. This,

compared with the annual growth rate of firms’ average output of 7.8 per

cent, suggests that firm- level productivity growth accounted for 27 per cent

of output growth during that decade. A further analysis of the relationship

between the estimated TFP and some approximate determinants, such

as marketization reform, firms’ R&D investment, market structure and

firms’ export behaviour, shows that these factors play different roles in

affecting the productivity of China’s iron and steel enterprises, depending

among other things on different types (including firm size, R&D invest-

ment, ownership and exporting behaviour), as shown in Tables 4.4 and 4.5.

Based on the entire sample estimation, we can see that firms’ TFP is

generally positively related to R&D investment, firm size, market share



and marketization reform, while negatively related to market monopoly

power (measured by the Herfindahl index for the top eight firms), and firms’

capital/labour ratio. As is shown in columns (6) and (7) in Table 4.3, the esti-

mated elasticities for the firms’ R&D index, market share, the scale dummy

and the marketization index are all positive and statistically significant at the

1 per cent level, while the estimated elasticity of firms’ capital/labour ratio

and market monopoly level are negative and also significant at the 1 per cent

level (in both the random- effects and fixed- effects frameworks). The results

are robust to TFP estimations with the OP and LP methods. These results

imply that the operation of China’s iron and steel firms has been relatively

labour- intensive. This, together with firm size, a high proportion of private

ownership and strong market share positions, has contributed positively to

the improved level of productivity. However, exporting firms are less likely

to have relatively high productivity vis- à- vis non- exporting ones. Also, to

our surprise, the impact of R&D (the new products share index) on average

had no significant impact on TFP levels. This may be consistent with the

fact that a large number of iron and steel firms, especially non- state SEs,

were still using rather old and outdated technologies in their production,

and these firms had less spending on R&D.

When we split the whole sample into two categories characterized by

firm size – LMEs and SEs (as shown in Table 4.4) – we find that the

drivers for improving productivity differ substantially with firm size. For

the LMEs, different degrees of privatization generally have no significant

Average TFP levelBest practice for the year

Firm

-leve

l TF

P

1998 1999 2000 2001 2002 2003 2004 2005 2006 2007


Figure 4.2 Changes in the TFP level of China’s iron and steel firms,

1998–2007



impact on their TFP performance (as shown in the fixed- effect model),

although new products do contribute to the improvement in productivity.

This result implies that the large state- owned steel enterprises have already

been competitive as compared with large private enterprises, because of

the reform measures implemented in the state sector (Ma et al., 2002).

However, for the SEs, those with high levels of privatization have signifi-

cant higher TFP than other types of SEs (in both the random- and fixed-

effect model), implying that the marketization reform was still important

for the large number of SEs in the industry.

In terms of the relationship between firm ownership and productivity,

Table 4.3 Determination of TFP in China’s iron and steel firms (all

firms), 1998–2007

Olley–Pakes model Levinsohn–Petrin

model

GMM model

Random

effects

Fixed

effects

Random

effects

Fixed

effects

Random

effects

Fixed

effects

Dependent variable: ln TFP

ln (K/L) −0.031*** −0.037*** −0.005*** −0.003*** −0.111*** −0.105***

(0.002) (0.003) (0.002) (0.001) (0.002) (0.003)

R&D share 0.000 0.000 0.000 0.000 −0.000 0.000

(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

Market share 0.082*** 0.094*** 0.049*** 0.070*** −0.061*** 0.066***

(0.012) (0.021) (0.009) (0.016) (0.016) (0.015)

Herfindahl

index

−0.329*** −0.268*** −0.293*** −0.192*** −0.400*** −0.240***

(0.019) (0.029) (0.018) (0.029) (0.020) (0.030)

D scale 0.126*** 0.115*** 0.075*** 0.084*** −0.053*** 0.060***

(0.006) (0.010) (0.006) (0.010) (0.008) (0.011)

Marketization

index

0.001*** 0.000*** 0.000*** 0.000** 0.001*** 0.000***

(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

Export share −0.022* −0.003 −0.022* −0.002 −0.047*** −0.010

(0.012) (0.022) (0.011) (0.023) (0.013) (0.023)

Constant −0.126*** −0.129*** −0.221*** −0.242*** −0.558*** −0.580***

(0.008) (0.012) (0.008) (0.012) (0.009) (0.012)

No. of

observations 26 215 26 215 26 215 26 215 26 215 26 215

R-squared 0.056 0.025 0.035 0.008 0.271 0.103

Notes:1. ***, ** and * represent the results statistically significant at the 1 per cent, 5 per cent

and 10 per cent levels, respectively. Numbers in parentheses are standard errors.2. ln TFP is defined as ln TFP 5 ln Y 2 bl ln L 2 bk ln K 2 bm ln M.

Source: Authors’ own estimation.



our estimation results show that SOEs are more likely to obtain productiv-

ity improvement through increasing R&D innovation and enlarged scale

of production from large capital investments; while the non- SOE LMEs

are more likely to obtain their productivity gains through exports. As

shown in Table 4.5, the estimated coefficient of exports for the non- SOEs

is positive and statistically significant at the 1 per cent level.

The next factor to consider is the industrial location. Iron and steel

firms have been physically distributed rather unevenly across different

regions. To examine different drivers of firms’ productivity related to loca-

tions, we have split the sample into three subgroups: the eastern, central

and western regions; the estimation results are reported in Table 4.6. We

Table 4.4 Determination of TFP in China’s iron and steel firms by firm

size, 1998–2007

Small firms Large and medium firms

Random

effects

Fixed

effects

Random

effects

Fixed

effects


ln (K/L) −0.110*** −0.105*** −0.110*** −0.070***

(0.002) (0.004) (0.002) (0.011)

R&D share −0.000 0.000 −0.000 0.001***

(0.000) (0.000) (0.000) (0.000)

Market share −0.088*** 0.065*** −0.088*** 0.036**

(0.024) (0.025) (0.024) (0.016)

Herfindahl index −0.377*** −0.225*** −0.377*** −0.298***

(0.021) (0.032) (0.021) (0.094)

Marketization index 0.001*** 0.000** 0.001*** 0.000

(0.000) (0.000) (0.000) (0.000)

Export share −0.052*** −0.017 −0.052*** 0.131

(0.013) (0.024) (0.013) (0.102)

Constant −0.561*** −0.571*** −0.561*** −0.789***

(0.009) (0.013) (0.009) (0.040)

No. of observations 25 022 25 022 25 022 1193

R- squared 0.226 0.100 0.380 0.097

Notes:1. ***, ** and * represent the estimation results statistically significant at the 1 per cent,

5 per cent and 10 per cent levels, respectively. Numbers in parentheses are standard errors.

2. ln TFP is defined as ln TFP 5 ln Y 2 bl ln L 2 bk ln K 2 bm ln M.




highlight the following findings. Although the general impacts of the

firms’ capital/labour ratio and scale on their TFP are similar, the impact

of market share on productivity is much stronger in the eastern region

while the impact of marketization and market power is most evident in

the western region. This finding is consistent with the fact that market-

oriented reforms have been more thoroughly done in the eastern than both

the central and western regions as measured by the relatively low share of

SOEs in total industry in the eastern region. The increased competition

makes the changes in market share an important factor in influencing

firms’ productivity in the eastern region.

For the same reason, further reform in deepening the process of

Table 4.5 Determination of TFP in China’s iron and steel firms by

ownership, 1998–2007

SOEs Non- SOEs

Random

effects

Fixed

effects

Random

effects

Fixed

effects


ln (K/L) −0.110*** −0.103*** −0.113*** −0.107***

(0.002) (0.004) (0.005) (0.010)

R&D share 0.000 0.001** −0.001* −0.001

(0.000) (0.000) (0.000) (0.001)

Market share −0.084*** 0.072** −0.038* 0.059***

(0.024) (0.031) (0.021) (0.016)

Herfindahl index −0.336*** −0.143*** −0.545*** −0.392***

(0.023) (0.036) (0.037) (0.053)

D scale −0.047*** 0.057*** −0.104*** 0.043**

(0.009) (0.011) (0.016) (0.019)

Export share −0.058*** −0.027 −0.024 0.042***

(0.013) (0.025) (0.035) (0.012)

Constant −0.493*** −0.556*** −0.522*** −0.608***

(0.007) (0.010) (0.015) (0.019)

No. of observations 19 938 19 938 6440 6440

R- squared 0.252 0.104 0.256 0.093







marketization plays a more important role in those central and western

regions which have a less competitive environment owing to the relatively

slow progress in reform. Finally, although increasing firm size is positively

correlated to firm performance in TFP for all the three regions, it gener-

ates a much larger impact (as measured by the magnitude of the coefficient

estimates) in the western region than elsewhere.

CONCLUSIONS

This chapter has aimed to fill the gap left by previous studies in the

field that had not been able to reach a consensus on the level and pos-

Table 4.6 Determination of TFP in China’s iron and steel firms by region,

1998–2007

Eastern region Central region Western region


ln (K/L) −0.102*** −0.099*** −0.125***

(0.004) (0.006) (0.011)

R&D share −0.000 −0.000 0.001**

(0.000) (0.001) (0.000)

Market share 0.064*** 0.109 0.059

(0.021) (0.067) (0.051)

Herfindahl index −0.365*** −0.106** 0.050

(0.038) (0.053) (0.105)

Firmscale_dummy 0.041*** 0.095*** 0.103***

(0.013) (0.023) (0.035)

Marketization index 0.000 0.000 0.001***

(0.000) (0.000) (0.000)

Export share −0.002 −0.032 0.006

(0.032) (0.036) (0.050)

Constant −0.522*** −0.591*** −0.765***

(0.015) (0.021) (0.039)


R- squared 0.106 0.100 0.109




Source: Authors’ own estimate.



sible drivers of TFP growth in the Chinese iron and steel industry. Our

approach was to correct some of the econometric problems that might

have constrained those previous studies by adopting the newly developed

econometric approaches. We then applied these approaches to a more up-

to- date sample covering the period 1998–2007.

The estimation results suggest that the previous studies have in all likeli-

hood underestimated the contribution of capital to industry output and

have correspondingly overestimated the contribution from intermediate

inputs resulting from the ‘endogenous input’ problem evident in previous

studies. Furthermore, our decomposition of derived TFP suggests that the

key drivers of productivity improvement differ substantially with respect

to differences in firm size, ownership type and geographical location.

Notably, the productivity of SEs is positively related to market share and

negatively related to R&D. For SOEs, firm- level productively is relatively

insensitive to market share and R&D, but more responsive to techno-

logical upgrading and marketization reform. The non- state large firms are

more likely to obtain their productivity gains through exporting. Finally,

increasing firms’ size is generally positively correlated to firms’ perform-

ance in TFP, and it is more so in the less- developed western than in the

eastern or central regions.

A policy implication from this study is that to further improve the pro-

ductivity and quality of Chinese iron and steel enterprises, different policy

instruments targeting firms with different characteristics in the process of

restructuring the industry may be desirable. For example, policy measures

aimed at market entry will work well for relatively small firms; further

progress on technological upgrading and marketization reform such as

development of shareholding will be more conductive to large SOEs; more

opportunities for trade will help improve productivity for non- state large

firms; and an increase in firms’ scale of production will be advantageous

for firms located in the western region.

NOTES

1. Ma et al. (2002) outline the increasing trend in the growing scale of existing firms.2. The estimation method can also be used for adopting other types of production func-

tions, provided some basic requirements are met (Ackerberg et al., 2005).3. This is not as straightforward a decision as it seems prima facie. As the marketiza-

tion reforms may have reduced the amount of bureaucracy employed in the industry without having a direct role in the production process, this may be an interesting effect to capture. However, we settled on this abstraction as we are most interested in proxying the ‘blue- collar’ workforce.



REFERENCES

Ackerberg, D.A., K. Caves and F. Garth (2005), ‘Structural identification of pro-duction functions’, available at http://ideas.repec.org/e/c/pac11.html.

Ackerberg, D.A., L.C. Benkard, B. Steven and A. Pakes (2008), ‘Econometric tools for analysing market outcomes’, in J. Heckman and E. Leamer (eds), Handbook of Econometrics, Amsterdam: North- Holland.

Arellano, M. and O. Bover (1995), ‘Another look at the instrumental variable estimation of error components models’, Journal of Econometrics, 68 (1), 29–51.

Blundell, R. and S. Bond (2000), ‘Initial conditions and moment restrictions in dynamic panel data models’, Journal of Econometrics, 87 (1), 115–43.

Brown, D.M. and F.R. Warren- Boulton (1988), ‘Testing the structure–compe-tition relationship on cross- sectional firm data’, US Department of Justice Economic Analysis Group discussion paper 88- 6, Washington, DC.

China National Bureau of Statistics (2006), China Statistical Yearbook, Beijing: China Statistical Press.

China National Bureau of Statistics (CNBS) (2009), China Statistical Yearbook, Beijing: China Statistical Press.

China Iron and Steel Association (CISA) (2008), China Steel Industry Yearbook, Beijing: China Iron and Steel Association.

Jefferson, G.H. (1990), ‘China’s iron and steel industry’, Journal of Development Economics, 33 (2), 329–55.

Kalirajan, K.P. and Y. Cao (1993), ‘Can Chinese firms behave like market entities: the case of Chinese iron and steel industry’, Applied Economics, 25 (12), 1071–80.

Labson, S., P. Gooday and A. Manson (1995), ‘China steel: China’s emerging steel industry and its impact on the world iron ore and steel market’, Australian Bureau of Agricultural and Resource Economics Research Report no. 95(4), Canberra.

Levinsohn, J. and A. Petrin (2003), ‘Estimating production functions using inputs to control for unobservables’, Review of Economic Studies, 70 (2), 317–41.

Ma, J., G.E. David, J.F. Robert and F.S. Donald (2002), ‘Technical efficiency and productivity change of China’s iron and steel industry’, International Journal of Production Economics, 76 (3), 293–312.

Olley, S.G. and G. Pakes (1996), ‘The dynamics of productivity in the telecommu-nications equipment industry’, Econometrica, 64 (6), 1263–97.

Solow, R.M. (1957), ‘Technical change and the aggregate production function’, Review of Economics and Statistics, 39 (3), 312–20.

Wooldridge, J.M. (2002), Introductory Econometrics: A Modern Approach, 2nd edn, Mason, OH: Thomson Learning.

Wooldridge, J.M. (2009), ‘On estimating firm- level production functions using proxy variables to control for unobservables’, Economics Letters, 104 (3), 112–14.

Wu, Y. (1996), ‘Technical efficiency and firm attributes in the Chinese iron and steel industry’, International Review of Applied Economics, 10 (2), 235–48.

Zhang, X. and S. Zhang (2001), ‘Technical efficiency in China’s iron and steel industry: evidence from the new census data’, International Review of Applied Economics, 15 (2), 199–211.


89

5. The technical efficiency of China’s large and medium iron and steel enterprises: a firm- level analysis


INTRODUCTION

The expansion of China’s iron and steel industry since the 1990s has

been driven largely by the strong domestic demand resulting from the

accelerated pace of urbanization and industrialization. A key question is

whether the rapid expansion of the industry is also accompanied by any

significant gain in efficiency in the large and medium state- owned enter-

prises (SOEs) through institutional and ownership reform. The question is

significant in that the improvement of productivity in the industry could

be an indication that there has been some impact on those SOEs from the

series of marketization reforms in the industry including privatization and

corporate restructuring. To demonstrate that these reforms work in the

industry, it is important to connect the improvement in SOEs’ firm- level

productivity with the marketization reform from an empirical perspective.

This is the task of this chapter.

Historically, China’s iron and steel industry has been dominated by

the large and medium SOEs. In 1999, the industry consisted of 3042

enterprises, of which only 793 were state- owned or majority- state- owned

holding companies (accounting for 26 per cent of the total). However,

the total output value and total assets of these SOEs accounted for 74 per

cent and 89 per cent of the whole industry, respectively. The significant

advantages of SOEs in firms’ scale, market share and capital stocks over

their private counterparts in the industry were not accompanied by high

productivity and profitability. This is partly because the institutional

arrangement in those SOEs associated with their ownership led to inef-

ficiency in investment and management compared with the non- state-

owned enterprises.



In order to solve the problem of inefficiency, the central government

has implemented a series of policies aiming to strengthen marketization

reforms of SOEs and encourage the entry of privatized enterprises to the

industry. As a consequence, more than one- third of SOEs were privatized

during the 1990s and into the new century, and the market share of non-

SOEs in the industry significantly increased. Between 1999 and 2005, the

total number of state- owned and state- holding enterprises reduced from

793 (accounting for 26 per cent) to 407 (accounting for 6.1 per cent). The

share of the output value of the state- owned and state- holding enterprises

over the industry total also reduced from 74 per cent to 47 per cent. In

particular, in the iron ore mining sector, the share of the total output value

of the state- owned and state- holding enterprises over the industry total fell

to 20 per cent in 2005.

Continuing marketization and privatization reform promoted market

competition in the industry and helped to improve SOEs’ production effi-

ciency and profitability, which in turn has further supported the expansion

of the top firms’ production levels. Between 1999 and 2007, the top 60 iron

and steel enterprises all expanded their production capacities, so that, by

2007, ten enterprises were each producing more than 10 million tonnes

of crude steel (after an annual growth rate of 22.9 per cent); 13 were pro-

ducing 5–10 million tonnes (after an annual growth rate of 9.3 per cent);

and 34 were producing 2–5 million tonnes (after an annual growth rate of

11.5 per cent). Reflected in the aggregate performance of the industry, the

total output of pig iron, crude steel and steel products in 2007 was 0.47

billion tonnes, 0.49 billion tonnes and 0.56 billion tonnes, respectively,

with annual growth rates of 20.3 per cent, 22.4 per cent and 24.7 per cent,

respectively.

This chapter examines the impact of marketization reforms on firm-

level productivity for large and medium enterprises during the period

1999–2005. Rather than using the data for the whole industry, this study

focuses on the data from 60 major SOEs in China’s iron and steel industry

to test the hypothesis that these reform measures affect firms’ performance

by changing the institutional arrangements which tend to improve the

technical efficiency of these firms. In particular, we distinguish between

the impacts of different ownership arrangements on enterprises’ techni-

cal efficiency and highlight the important role that iron ore has played in

affecting the technical efficiency in SOEs’ production. The latter point has

some important policy implications for securing the long- term supply of

iron ore for China’s steel mills and for firms’ strategy of investing in over-

seas mining operations.


Technical efficiency of large and medium enterprises 91

MARKETIZATION REFORM AND LARGE STATE- OWNED ENTERPRISE: LITERATURE REVIEW

The marketization reform in China’s iron and steel industry started in

the late 1970s, but the main changes in ownership at the enterprise level

did not take place until the early 1990s. The initial purpose of the reform

was to establish a modern management system within the large iron and

steel enterprises so as to improve firm- level productivity and profitability.

Between 1994 and 1998, there had been 12 SOEs (including Wu Gang,

Ben Gang, Tai Gang, Cong Gang, Ba Yi, Tianji Steel Tube and Da Zhi)

involved in the national pilot reform programme and 57 SOEs (including

Han Gang, Fushun Gang, Tianjin Gang and Jiuquan Gang) in the local

pilot reform programme.

Since 1998, further marketization reforms have been extended to

restructure all SOEs in the industry. A series of reform measures have been

carried out (following the new Cooperation Law), including attempts to

clarify the property rights, strengthen the principal–agent relationship (or

ownership) and set up the modern management and corporate finance

systems. This helped to eliminate institutional barriers for SOEs in the

industry which were associated with the legacies of central planning under

which there was very little autonomy at the firm level with respect to

decision- making.

In 2003, the National Development and Reform Commission (NDRC)

implemented the ‘About Restricting Iron and Steel Firms’ Rush

Investment’ and ‘Iron and Steel Industry Development Strategy’ (NDRC,

2003) to regulate firms’ production and market competition. Through

strengthening government policy direction, raising the threshold for

market entry, and tightening the arrangements around bank loans, the

new policy succeeded in preventing 345 projects – recognized as duplicate

or redundant construction by NDRC – from entering the industry. The

government also managed to close down 12.9 million tonnes and 13.1

million tonnes of outdated production capacity for iron and steel in that

year. These policies, followed by ‘Accelerating Structural Change in the

Iron and Steel Industry’ implemented by the NDRC in 2006 (NDRC,

2006), helped to restrict the duplication of investment in the industry and

provide an improved market environment for large and medium iron and

steel enterprises.

In recent years, some domestic SOEs in the industry have started to

enter the international market through public listing (for fund- raising) and

investment in overseas iron ore operations. Through mergers and associa-

tions with some upstream and downstream enterprises, some SOEs in the

industry have succeeded in building up their competitive advantages. As a



consequence, importing iron ore from, and exporting iron and steel prod-

ucts to, the international market have become significant characteristics

of these enterprises. These characteristics reflect the fact that the iron and

steel industry in China has become more deeply integrated with the world

market, providing some evidence that the predominant character of the

production upon which China’s comparative advantage lies has begun to

shift from labour- intensive to capital- intensive. Much as in other sectors,

the market integration process helps to improve the productivity at firm

level.

Marketization reform in China’s iron and steel industry has been suc-

cessful and helped to improve enterprises’ productivity, particularly for

those SOEs during the period under study. There are two main channels

through which the positive impacts of marketization reforms impact on

firms’ performance. First, marketization reform can improve SOEs’ tech-

nical efficiency through strengthening the within- firm incentive mecha-

nism. Second, marketization reform can regulate the market environment

and intensify market competition, leading to the reallocation of market

share to more efficient enterprises. This will help to nurture within- firm

innovation and maintain the long- term productivity growth of SOEs.

These two impacts of marketization reform are frequently referenced in

the literature on privatization. In this chapter we provide the empirical evi-

dence as to whether these positive impacts of marketization are working

for the large and medium SOEs in the steel industry in China.

There have been a large number of studies on China’s iron and steel

industry, including, among others, Jefferson (1990); Kalirajan and Cao

(1993); Wu (1996, 2000); Zhang and Zhang (2001); Nolan and Yeung

(2001); Ma et al. (2002); Movshuk (2004); and Sun (2005). These studies

focus mainly on the firm- level analysis from three aspects. Jefferson (1990)

was the first to use both the Cobb–Douglas and the log–linear production

functions to estimate the multifactor productivity for China’s iron and

steel industry with 120 large and medium enterprises in 1986. Thereafter,

Kalirajian and Cao (1993) and Wu (1996) adopted a stochastic frontier

production function to estimate technical efficiency with 1988 data for 97

and 87 enterprises; and Zhang and Zhang (1999) used the 1995 national

industrial census data to examine the production frontier of iron and steel

firms.

As more time- series data were released in China’s Iron and Steel

Industrial Yearbook, Ma et al. (2000) used data envelope analysis with

panel data for 88 enterprises during the period 1989–1997, and Movshuk

(2004) used the stochastic frontier analysis with 82 enterprises for the

period 1988–2000 to examine firms’ productivity and technical efficiency

and discuss the impact of economic reform on China’s iron and steel



industry. Nolan and Yeung (2001) and Sun (2005) did their case studies on

Shou Gang Group in Beijing and Bao Shan Group in Shanghai, respec-

tively, these companies being the two largest iron and steel enterprises in

China respectively. The authors analysed changes in the productivity and

competitiveness of China’s iron and steel industry and its catch- up with

the advanced level of world production.

However, these reviewed studies did not provide consistent findings

with respect to the impact of marketization on SOE firms’ performance,

and in some cases contrary results emerged. For example, Movshuk (2004)

found that there had been no significant increase in the technical efficiency

of China’s iron and steel industry, particularly in the four largest enter-

prises, and argued that the impacts of major reform measures on enter-

prises’ technical efficiency were weak during the 1990s. Ma et al. (2000)

found however that the estimated technical efficiency of China’s iron and

steel firms increased from 58 per cent in 1989 to 66 per cent in 1997. There

have also been some differences with respect to the performance of the

four largest enterprises according to Ma et al. (2000) and other studies

such as Nolan and Yeung (2001) and Sun (2005).

Further studies are therefore needed to re- examine the relationship

between marketization reform and the productivity and technical effi-

ciency of SOEs in China’s iron and steel industry, especially after 2000. In

this chapter, we apply the stochastic frontier model with the unbalanced

panel data of 68 large and medium- sized SOEs. Contributing to the lit-

erature in the field, we incorporate intermediate goods into the log–linear

production function so as to improve the accuracy of the productivity esti-

mation with a better control of the returns- to- scale effect in production.

The results show that marketization reform has significantly promoted

SOEs’ productivity and technical efficiency during the decade to 2005,

although such an impact varies across different types of reform.

STOCHASTIC FRONTIER METHOD AND EMPIRICAL MODEL SPECIFICATION

Traditional production theory is based on the assumption that the behav-

iour of production units is optimal. Under the conditions of perfect

competition, a production unit will produce at the most efficient point

that satisfies the objective of profit maximization. It is assumed that pro-

duction units optimize from a technical or engineering perspective by not

wasting resources, and that they operate up to their maximum potential

output with available input resources. Production units are also assumed

to optimize from an economic perspective by solving allocation problems



that involve prices, – that is, they locate the input resource effectively so as

to operate on, rather than above, their minimum cost boundary.

However, in practice, for various reasons, not all production units

succeed in working at optimal levels, since firms may have the incomplete

nature of knowledge of best practice and other organizational factors.

Therefore, it is important to have a way of analysing the degree to which

production units fail to optimize, and the extent of the departure from the

most efficient level. In response to these needs, advanced econometric and

mathematical methods have been developed, among which the stochastic

frontier method has emerged and attracted much attention in the applied

analyses.

In order to examine the production efficiency of the Chinese iron and

steel industry, we estimated a stochastic frontier model as described

by Battese and Coelli (1995), or the BC model for short. While early

stochastic frontier models were devised for cross- sectional data, the BC

model is formulated for panel datasets that may be unbalanced. The

model not only estimates inefficiency levels of particular enterprises,

but also explains their inefficiency in terms of potentially important

explanatory variables. The model decomposes TFP growth into three

components: technological growth (essentially, a shift of the production

possibility frontier, set by best- practice enterprises); inefficiency changes

(that is, deviations of actual output level from the production possibility

frontier); and scale- mix effects (output change due to increasing returns

to scale).

Conventionally, stochastic frontier models contain a production func-

tion f ( . ) and the disturbance term ei,t:

ln (Yit) 5 ln [f (Xit; b)] 1 eit, (5.1)

where Yit is the production for the ith company in year t, Xit is the vector

of independent variables (inputs, etc.), b is the corresponding vector of

unknown parameters to be estimated, and f ( . ) denotes a production

function (in the form of a translog, or Cobb–Douglas, production func-

tion, or similar).

The disturbance term is defined by eit 5 vit 2 uit, where vit is a conven-

tional systematic random disturbance term, associated with the impact of

omitted variables on the output variable, and uit is a non- negative random

term, representing various inefficiencies in production. The random dis-

turbance term vit is assumed to be following an independent and identical

normal distribution (i.i.d.) with mean zero and variance s2v, while uit is

obtained by non- negative truncation of the normal distribution with mean

mit and variance s2u.



In the BC model, the mean of the inefficiency term uit is defined by:

mit 5 zitd, (5.2)

where zit is the vector of variables that explain technical inefficiency and d

is the corresponding vector of unknown parameters to be estimated.

The model was estimated by applying the method of maximum

likelihood, using the computer programme frontier (version 4.1) of

Coelli (1996), with variance parameters expressed by s2 5 s2v 1 s2

u and

g 5 s2u/s

2. Technical efficiency TEit of the ith enterprise in year t equals

the ratio of observed output level to the estimated frontier output level:

TEit 5Yit

exp [ f (Xit; b) ]5 exp (2uit

) . (5.3)

The inefficiency component uit in Equation (5.3) is not observable, but can

be estimated by using the minimum squared error predictor of uit:

E [exp (2uit0eit

) ] 5 [exp (2m*it 1 12s

2*) ]

[�(m*it/s*) 2 s* ]

�(m*it/s*), (5.4)

where m*it 5 [s2v(d rzit

) 2 s2u(eit

) ] /s2,s2* 5 s2

vg, and �( . ) is the cumulative

distribution function of a standard normal variable.

Using the estimates of TEit from Equation (5.4), the index of technical

efficiency change DTE for the ith enterprise between time periods t and s

was calculated by:

DTE 5TEit

TEis

. (5.5)

Following Coelli et al. (1998), the index of technical change DTCh

between periods t and s was obtained from

DTCh 5 e c1 10f (Xis, s, b)

0sd 3 c1 1

0f (Xit, t, b)

0td f 1/2

, (5.6)

where the index of TFP growth DTFP will be calculated.

In production function f(.), production Yit was dependent on two

inputs, capital Kit and labour Lit, as well as time t. We estimated two con-

ventional production functions – a quadratic form in Equation (5.7) and a

more restricted Cobb–Douglas form in (5.8):



ln (Yit) 5 b0 1 bk ln (Kit

) 1 bl ln (Lit)1 bm ln (Mit

) 1 12 3 [bkk ln (Kit

) 2

1 bll ln (Lit) 2 1 bmm ln (Mit

) 2 1 btt 1 D1999 1 (vit 2 uit) ; (5.7)

ln (Yit) 5 b0 1 bk ln (Kit

) 1 bl ln (Lit) 1 bm ln (Mit

)

1 btt 1 D1999 1 (vit 2 uit) . (5.8)

The hypothesis testing was performed by the generalized likelihood

ratio (LR) statistic:

l 5 22 ln cL(H0)

L(H1)d , (5.9)

where L(H0) and L(H1

) are the values of likelihood function under the

null and alternative specifications. The l statistic is non- negative, and

follows c2r distribution under the null hypothesis, where r denotes the

number of restrictions.

To examine various potential determinants of inefficiency in the specifi-

cation (5.2), we considered several sets of zi. They are (1) capital intensity

to account for firms’ specific characteristics; (2) scale to capture the rela-

tive size of the firms; (3) the marketization index; and (4) firms’ long- term

investment.

Overall, the inefficiency model (5.2) was specified as follows:

mit 5 d0 1 d1Ageit 1 d2 ln kit 1 d3Scaleit 1 d4Mktit

1 d5Linvit 1 d6Dexpit 1 d7Drestaxit 1 d8R4 1 .it, (5.10)

where Ageit denotes the enterprise’s age; kit denotes the capital labour

ratio; Scaleit denotes the relative scale of enterprises; Mktit the marketiza-

tion index; and Linvit the ratio of the long- term investment over total fixed

assets. Dexpit, Drestaxit and R4 are three dummy variables, representing

the enterprise’s exports, resource tax and whether they are the four largest

iron and steel enterprises. A particular feature of the specification (5.10) is

that its parameters dj (j 5 1, 2, 3, 4, 5, 6, 7, 8) measure the impact of those

listed variables on firms’ technical inefficiency.

DATA AND ESTIMATION RESULTS

The primary source of firm- level data used for this study was the Annual

Financial Report of Large and Medium Iron and Steel Firms (CISA, various



years). The annual reports provide data on output with both current

and constant prices; the total and net fixed assets; total and working (or

‘congye’) labour at the end of year; and various financial and other indi-

cators for about 130 large enterprises. We compiled an unbalanced panel

dataset, with data reported for most variables covering the seven- year

period 1999–2005. Table 5.1 reports the basic statistics relating to the

dataset. Overall, these enterprises accounted for about half of gross output

and valued added in China’s iron and steel industry, and 53 per cent of

total employment and 48 per cent of total fixed assets.

The annual report contains different measurements for gross output

value over time: gross industrial output value (both in real terms with 1990

constant prices and in nominal terms) for the period 1999–2003, and gross

industrial output value (only in nominal terms) for the period 2004–2005.

In order to make these data consistent over time, we use the ex- factory

price index to adjust the gross output value for both 2004 and 2005.

We use the method provided by the Chinese National Bureau of

Statistics (CNBS) to estimate the value of the intermediate inputs, which

is equal to the gross output value minus value added and then plus value-

added tax. Since major intermediate inputs in the iron and steel industry

are coal (or electricity) and iron ore, the price index of energy and mate-

rials is used to make the adjustment for inflation. For capital input, we

used the net value of fixed assets (in nominal terms) as the initial value.

Table 5.1 Descriptive statistics of sample iron and steel enterprises,

1999–2005

Year/Item Number

of firms

Output

value

(constant

price)

Total

labour

(end

year)

(1000

persons)

Total

fixed

assets

(net)

(million

yuan)

Long- term

investment

(million

yuan)

Total

fixed

assets

(million

yuan)

Total

value

added

(million

yuan)

1999 60 62.61 80.62 56.51 36.59 57.85 55.43

2000 68 75.00 81.15 71.76 78.79 73.32 77.74

2001 56 32.14 38.77 38.53 32.37 38.38 36.73

2002 64 39.38 42.55 40.03 31.53 40.22 40.55

2003 63 41.23 42.24 40.45 36.32 40.53 37.88

2004 68 41.21 43.01 42.26 32.04 42.26 41.74

2005 66 43.24 44.94 43.47 32.65 43.30 39.84

Average 64 47.83 53.33 47.57 40.04 47.98 47.13

Source: Authors’ own calculation based on the firms’ annual financial reports.



The accumulated price index of investment in fixed assets was calculated

for 2000 with the data from the China Statistical Yearbook (CNBS, 2006),

and used for discounting the net fixed assets in 2000. For latter years, we

used the net fixed assets in 2000 with constant prices plus the increased

part of the changes in the net fixed assets adjusted with the current- year

price index of investment of fixed assets. Finally, all of the capital is mul-

tiplied by 0.817 to generate the capital used for direct production, since

iron and steel enterprises in China usually have some proportion of assets

which are not productive (Jefferson, 1990).

For labour input, we use the total congye (on- job) labour at the end

of year for the period 2002–2005. For the years before 2002, we use the

average ratio between working labour and total labour for the period

2002–05 as an index to discount the total labour at the end of the year

for the period 1999–2001. The reason we made this approximation was

to make sure that the ratio between working labour and total labour was

consistent over time.

To capture the impact of marketization reform on SOEs’ technical effi-

ciency, we further define two groups of indices (for marketization reform)

based on firms’ registered capital: one group is the share of non- state-

owned capital in the total registered capital and the other group is the

shares of each type of non- state capital in the total registered capital. The

former index is used to examine the average impact while the latter is used

to explore the relative impact of different types of marketization reforms.

We use three shares to roughly capture the three types of marketization

reforms: legal person share for mutual purchase or firms’ restructuring

reform; individual share for shareholding reform; and FDI share for the

openness reform. Finally, we also distinguish SOEs with different export-

ing and importing behaviours by using the dummy variables.

MARKETIZATION REFORM, INTERNATIONAL TRADE AND SOES’ TECHNICAL EFFICIENCY

Technological Progress vs. Technical Efficiency

Based on Equations (5.7) and (5.8), we estimate the technical efficiency of

the large and medium- sized SOEs in China between 1999 and 2005. Table

5.2 shows the estimation results based on the Cobb–Douglas and quad-

ratic production function.

An important finding from the estimation is that the values of g

(defined as sigma- v/sigma- u) in the two sets of models range from 0.92 to

0.96, which is close to unity. This suggests that technical inefficiencies are



significant in the sample firms and the application of the technical inef-

ficiency model would be appropriate. Moreover, the log- likelihood test

(LR) for the null hypothesis that the interaction terms in the quadratic

specification are jointly insignificant is 27.36, which is larger than the

critical value of 17.75 at the 1 per cent level (Kodde and Palm, 1986). This

implies that the quadratic function form is more suitable than the Cobb–

Douglas form.

Table 5.2 Estimation results from the production function model,

1999–2005

Cobb–Douglas function Quadratic function

On- average

model

Comparable

model

On- average

model

Comparable

model

Dependent variable: ln y

ln L −0.020 −0.013 0.000 0.054

(0.017) (0.016) (0.057) (0.058)

ln L squared term – – −0.001 −0.005

– – (0.004) (0.004)

ln K 0.109*** 0.116*** 0.145*** 0.127***

(0.014) (0.013) (0.026) (0.027)

ln K squared term – – −0.005* −0.002

– – (0.002) (0.003)

ln M 0.857*** 0.847*** 0.721*** 0.728***

(0.012) (0.012) (0.038) (0.038)

ln M squared term – – 0.011*** 0.009***

– – (0.003) (0.003)

Year 0.002*** 0.002*** 0.003*** 0.003***

(0.000) (0.000) (0.000) (0.000)

ln sig 2v (constant) −2.765*** −2.811*** −2.773*** −2.814***

(0.075) (0.067) (0.071) (0.064)

Sigma- v 0.251*** 0.245*** 0.250*** 0.245***

(0.009) (0.008) (0.009) (0.008)

Wald chi(2) 121 564 152 077 125 589 156 081


Notes:1. The null hypothesis of preferring the Cobb–Douglas function is rejected at the 1 per

cent level, since the likelihood ratio (LR) test is 27.36, which is larger than the critical value 17.75 (Kodde and Palm, 1986).

2. *, ** and *** represent the coefficient estimates statistically significant at the 10 per cent, 5 per cent and 1 per cent levels, respectively. Numbers in parentheses are standard errors.




Based on the quadratic production function, we introduce the time

trend variable into the production function model estimation to capture

changes in the technological progress of SOEs. Table 5.2 reports the esti-

mated coefficients of time trends in the production functions which are all

positive and statistically significant at the 1 per cent level, suggesting that

SOEs have made some positive technological progress over the period

under review. After controlling for the technological progress, a further

estimation shows that firms’ technical efficiency has also improved.

Between 1999 and 2005, the average technical efficiency of SOEs in our

sample was 80.9 per cent and the average coefficient of returns- to- scale for

the quadratic production function with intermediate inputs is 1.63. This

implies that our sample firms generally enjoy increasing returns to scale,

but there is still a large potential for them to increase their technical effi-

ciency. Two important implications of the changing technical efficiency of

SOEs can be made below.

First, although the average technical efficiency of enterprises was still

relatively low, it increased during the period 1999–2005 (Figure 5.1).

Second, the increase of enterprises’ technical efficiency came mainly

from the catch- up effect of the original low- efficiency enterprises. This

can be seen in the dramatic decrease in the number of low- efficiency

enterprises during the period 1999–2005. In terms of the determinants

of the technical efficiency of China’s iron and steel enterprises, different

factors have different effects and their various impacts can be summa-

rized below.

0.85

0.86

0.87

0.88

0.89

0.90

0.91

1999

Tec

hnic

al e

ffici

ency

2000 2001 2002 2003 2004 2005

Source: Authors’ own calculation based on the estimation results.

Figure 5.1 Mean technical efficiency of iron and steel enterprises in

China, 1999–2005



Marketization Reform and its Impacts on Technical Efficiency

In theory, the technological progress in large and medium SOEs can result

from continuous investment and R&D innovation, while technical effi-

ciency gain can be the result of marketization reform. Yet from the previ-

ous literature as reviewed earlier the empirical evidence on the positive

causal relationship between marketization reform and technical efficiency

is mixed. We intend to clarify this by estimating further the technical inef-

ficiency model (Table 5.3).

Two groups of marketization indices are incorporated into the second-

step technical inefficiency model following the production function esti-

mation. In Table 5.3, columns (2) and (4) report the results from the

regressions with the total non- state- owned share of capital, and columns

(3) and (5) report the results from the regressions with each individual

non- state- owned share of capital. The share of aggregated non- state-

owned capital in the total registered capital stock is used as an approxima-

tion of the overall marketization reform, and the estimated elasticity of the

variable is positive, but statistically insignificant at the 10 per cent level.

This result seems to suggest that marketization reforms have made no

significant contribution to the improvement of SOEs’ technical efficiency

(the finding is consistent with those reported in the existing literature).

However, when the non- state- owned capital share is split according to

ownership type, to capture the relative impact of different marketiza-

tion reforms, the estimated results show that the elasticity of legal person

share is positive and statistically significant at the 1–5 per cent level while

those of individual share and FDI share are negative and significant at the

1–5 per cent level. This implies that different marketization reforms may

produce some differing impacts on the large and medium SOEs’ technical

efficiency in the industry. In particular, the shareholding reform (repre-

sented by the individual share) and the openness reform (represented by

the FDI share) have helped to reduce the production inefficiency of SOEs,

while the mutual purchase or firms’ internal restructuring (represented by

the legal person share) does not. This mixed result is probably the main

reason why the impact of marketization reform at the aggregate level on

firms’ technical efficiency is not significant.

As for the impact of other factors on firms’ technical efficiency, the

elasticity of firms’ capital/labour ratio ranges from 0.77 to 1.08 and is

statistically significant at the 1 per cent level. This implies that a 1 per

cent increase in firms’ capital/labour ratio may lead to a 0.77–1.08 per

cent decline in technical efficiency, even though an increase in the capital/

labour ratio may enhance technological efficiency. This finding implies

that making use of the comparative advantage with respect to labour is



still an important factor for improving SOEs’ technical efficiency. The

result is consistent with the insignificant estimates of labour elasticity in

the production function. Finally, the coefficients of exports and resource

tax are negative and statistically significant at the 1 per cent level. This

suggests that openness to trade and market- oriented administration (for

example, the imposition of the resource tax) may nurture entrepreneurship

and tend to reduce the inefficiency of SOEs.

Table 5.3 Estimation results from the technical inefficiency model,

1999–2005

Cobb–Douglas function Quadratic function

On- average

model

Comparable

model

On- average

model

Comparable

model

Dependent variable: ln u (technical inefficiency obtained from the production

function model)

Year −0.175*** −0.205*** −0.151*** −0.180***

(0.046) (0.046) (0.048) (0.048)

Non- state- owned share 0.000 – 0.001 –

(0.004) – (0.004) –

Legal person share – 0.100** – 0.008**

– (0.045) – (0.004)

Individual share – −0.008** – −0.076**

– (0.004) – (0.038)

FDI share – −0.038*** – −0.045***

– (0.014) – (0.018)

ln (K/L) ratio 0.846*** 1.078*** 0.765*** 1.063***

(0.231) (0.209) (0.229) (0.229)

Export dummy −1.974*** −1.880*** −2.165*** −1.962***

(0.484) (0.449) (0.550) (0.470)

Resource tax −0.002*** −0.002*** −0.001*** −0.001***

(0.000) (0.000) (0.000) (0.000)

Constant 348.549*** 407.867*** 299.421*** 358.214***

(92.543) (92.015) (95.952) (96.290)

Sigma- v 0.251*** 0.245*** 0.250*** 0.245***

(0.009) (0.008) (0.009) (0.008)

Wald chi(2) 121 564 152 077 125 589 156 081


Note: *, ** and *** represent the coefficient estimates statistically significant at the 10 per cent, 5 per cent and 1 per cent levels, respectively. Numbers in parentheses are standard errors.




IMPORT OF IRON ORE AND SOES’ PRODUCTION EFFICIENCY

How will imports of iron ore from the world market affect China’s iron

and steel industry’s production efficiency? Owing to the endogeneity and

the reverse causality problem, we cannot simply run a regression to find

the causal relationship between imports of iron ore and firms’ efficiency.

An alternative way of dealing with this issue was to split the sample enter-

prises into three groups: those which used only domestically supplied iron

ore, those which relied on imports of iron ore for less than 50 per cent of

total demand, and those which depended on imports of iron ore for 50 per

cent or more of total demand. The estimated technical efficiencies in all

these three types of enterprises were then compared.

We found that the SOEs with imports of iron ore of 50 per cent or

more of their total demand were more technically efficient than those with

imports of iron ore less than 50 per cent of demand and those with no

imports of iron ore at all. Figure 5.2 shows the comparison of technical

efficiency among the three groups of enterprises from both a comparative

static perspective and a dynamic one. Figure 5.2(a) shows that the average

technical efficiency of enterprises with imports of iron ore of 50 per cent or

more was 82.1 per cent, which was higher than those with imports of less

than 50 per cent and those with no imports (80.9 per cent and 80.1 per cent,

respectively). In particular, from a dynamic perspective, the enterprises

with positive imports of iron ore were significantly more efficient than

those with no imports of iron ore (Figure 5.2(b)). This implies that imports

of iron ore from the international market tended to improve technical effi-

ciency of enterprises in China’s iron and steel industry. A possible expla-

nation is that intermediate inputs played an important role in increasing

the total output, and the imports of iron ore from the international market

were usually of better quality than the domestic supplies in terms of iron

content. Those enterprises which exclusively used domestically supplied

iron ore might incur much higher costs of sifting iron ore and producing

iron and steel products that would affect their level of efficiency.



CONCLUSIONS

This chapter has examined the technical efficiency of large and medium

SOEs in China’s iron and steel industry between 1999 and 2005, apply-

ing a stochastic frontier model with the unbalanced panel data. We

found that SOEs in the iron and steel industry have generally experienced

rapid improvement in production (technical) efficiency resulting from the

marketization reform which has been implemented over the same period

as part of China’s overall economic transformation towards a market

economy. Although different types of reforms may bring about different

impacts on enterprises’ performance, those marketization reform meas-

ures such as shareholding reform and openness to trade and investment

Tec

hnic

al e

ffici

ency

0.50

0.55

0.60

0.65

0.70

0.75

0.80

0.85

0.90

0.95x = 0 0 < x < 50 x > 50

1999 2000 2001 2002 2003 2004 2005

0.80

0.80

0.81

0.81

0.82

0.82

0.83

M = 0

Firms’ import share

Tec

hnic

al e

ffici

ency

0 < M < 50% M > 50%

a

b


Figure 5.2 Impact of imports on technical efficiency in China’s iron and

steel industry, 1999–2005. (a) Comparison of the average

technical efficiencies among enterprises which import: more

than 50 per cent; less than 50 per cent; and no iron ore.

(b) Changes of technical efficiencies of three types of

enterprise over time



have helped to promote the improvement of SOEs’ technical efficiency

more than other measures. Finally, firms’ trade orientations such as

exports of iron and steel products and imports of iron ore have also played

an important role in affecting enterprises’ technical efficiency, which

deserves some further exploration in future.

REFERENCES

Battese, G.E. and T.J. Coelli (1995), ‘A model for technical inefficiency effects in a stochastic frontier production function for panel data’, Empirical Economics, 20 (2), 325–32.


China Iron and Steel Industry Association (CISA) (various years), ‘Annual finan-cial report of large and medium iron and steel firms’, mimeo.

Jefferson, G.H. (1990), ‘China’s iron and steel industry’, Journal of Development Economics, 33 (2), 329–55.

Kalirajan, K.P. and Y. Cao, (1993), ‘Can Chinese state enterprises perform like market entities: productive efficiency in the Chinese iron and steel industry’, Applied Economics, 25 (8), 1071–80.

Kodde, D.A. and F.C. Palm (1986), ‘Notes and comments: Wald criteria for jointly testing equality and inequality restrictions’, Econometrica, 54 (5), 1243–8.

Ma, J., D.G. Evans, R.J. Fuller and D.F. Stewart (2002), ‘Technical efficiency and productivity change of China’s iron and steel industry’, International Journal of Production Economics, 76 (3), 293–312.

Movshuk, O. (2004), ‘Restructuring, productivity and technical efficiency in China’s iron and steel industry, 1988–2000’, Journal of Asian Economics, 15 (1), 135–51.

National Development and Reform Commission (NDRC) (2003) ‘About restrict-ing iron and steel firms rush investment and iron and steel industry development strategy’, NDRC policy document 2003, Beijing.

National Development and Reform Commission (NDRC) (2006) ‘Accelerating structural change in iron and steel industry’, NDRC Policy document 2006, Beijing.

Nolan, P. and G. Yeung (2001), ‘Large firms and catch- up in a transitional economy: the case of the Shougang group in China’, Economic Planning, 34 (1–2), 159–78.

Sun, P. (2005), ‘Industrial policy, corporate governance and competitiveness of China’s national champions: the case of the Shanghai Baosteel group’, Journal of Chinese Economic and Business Studies, 3 (2), 173–92.

Wu, Y. (1996), ‘Technical efficiency and firm attributes in the Chinese iron and steel industry’, International Review of Applied Economics, 10 (2), 235–48.

Wu, Y, (2000), ‘The Chinese steel industry: recent developments and prospects’, Resource Policy, 26 (3), 171–78.

Zhang, X. and S. Zhang (2001), ‘Technical efficiency in China’s iron and steel industry: evidence from the new census data’, International Review of Applied Economics, 15 (2), 199–211.


106

6. The backward and forward linkages of the iron and steel industry in China and their implications


INTRODUCTION

A unique feature of the iron and steel industry is that it has some close

relationships both upstream and downstream, in that the rapid expan-

sion of the industry may influence the performance of both the upstream

and downstream industries, possibly through cross- industry productiv-

ity spillover. Capturing this spillover effect is the subject of this chapter.

Following the recent literature analysing firm- level productivity (Javorcik,

2004), the chapter examines the cross- industry productivity spillover of

the iron and steel industry using firm- level data in the Chinese manufac-

turing sector over the period 2000–03. After controlling for the potential

endogeneity problem, we find that increases in the average productivity

of firms in the iron and steel industry may promote firms’ productivity

downstream but might not be conducive to firms’ productivity upstream.

When firms’ heterogeneity in operation size and productivity are consid-

ered, the results show that medium and small firms with low productivity

downstream are likely to benefit more from the productivity growth of the

Chinese iron and steel industry.

This chapter makes the following contributions. To begin with, the

study is the first to use firm- level data to explore the cross- industry link-

ages in China’s iron and steel industry. This allows us to examine more

closely the impact of China’s iron and steel industry on other related man-

ufacturing industries, with some useful policy implications being drawn.

Second, the data used in this chapter are unbalanced panel data, coming

from the Annual Manufacturing Enterprise Census. This data set covers

all firms (including entering and exiting firms) for each industry and each

year, enabling us to avoid the selection bias problem (usually suffered by

a number of studies in the literature which relied primarily on the survey

data). Third, we introduce firms’ heterogeneity in exploring the cross-


Backward and forward linkages of the Chinese steel industry 107

industry spillovers in the iron and steel industry. Adopting this approach

assists an understanding of the channels through which the backward and

forward linkages of the iron and steel industry are transmitted.

CROSS- INDUSTRY LINKAGE OF THE IRON AND STEEL INDUSTRY IN CHINA

It is found in the literature that the rapid development of a firm or an

industry may not only generate positive spillovers to firms and industries

in its neighbourhood, but also affect firms and industries operating in

the upstream and downstream sectors through the cross- industry linkage

(Krugman, 1991, 1998; Venables, 1996). To see how such a cross- industry

linkage works, we assume that a firm or industry has a strong input–output

linkage with other firms and industries through purchasing the interme-

diate inputs from the upstream industries and selling the outputs to the

downstream ones. On one hand, as the productivity of the firm or industry

increases, it may generate additional demand for inputs from the upstream

firms and industries. The improved demand may intensify market compe-

tition of the upstream firms and thus increase their productivity. On the

other hand, the improved productivity of the firm and the industry may

also raise the quality of outputs for the downstream firms and industries

so as to promote their productivity through decreasing production costs

and nurturing new products. Furthermore, an increase in productivity of

the firm or industry is likely to encourage larger- scale operations which

strengthen the standardization in the manufacturing process of indus-

trial goods, promoting the productivity of all industries. These effects

are usually defined as the backward and forward linkages of a firm or an

industry in relation to its upstream and downstream firms or industries.

As one of the most important pillar industries, the iron and steel

industry in China has a strong linkage with other industries throughout

the manufacturing sector and beyond. It provides the basic materials for

most manufacturing sectors producing goods sold in both the domestic

and international markets. According to the China National Bureau of

Statistics (2008), in 2007 around 423 million tonnes of crude steel were

produced and consumed by the manufacturing industries such as metal

producers, machinery manufacturers, construction and so on which

account for 30 per cent of the total outputs of these industrial sectors.

In addition, with China’s underlying comparative advantage shifting

towards more capital- intensive products, the steel industry plays an

important role in supporting China to become a ‘manufacturing factory

of the world’ (UNSD, 2011).1 The industry also supports the primary



industries upstream such as coal- mining, iron ore mining, electricity gen-

eration, and so on. In terms of the backward linkage, the iron and steel

industry in China consumes 85 per cent of coking coal, 20 per cent of

electricity and almost all of the iron ore supplied from both domestic and

overseas sources. The strong cross- industry linkage between the Chinese

iron and steel industry and other manufacturing industries suggests that

any increase in productivity of the steel industry may spill over to those

upstream and downstream industries.

However, it remains to be tested empirically to what extent such cross-

industry linkage effects would exist between the Chinese steel industry and

other industries. Some previous studies have explored the backward and

forward linkage of foreign direct investment (FDI) in China’s manufac-

turing sector from an empirical perspective. For example, Lin and Saggi

(2004) used firm- level data to examine the backward and forward link-

ages of FDI inflow at the two- digit China Industry Classification Code

(CICC) level in the Chinese manufacturing sector and found that there

have been significant positive backward and forward linkages between

both FDI and domestic firms in terms of technological spillovers. Hu

and Jefferson (2002), Hu et al. (2008) and Harrison et al. (2008) use the

2002 input–output table and firm- level data to re- examine the backward

and forward linkages of FDI in China at the three- digit International

Standard Industry Code (ISIC) level. They found that the backward and

forward linkages of FDI in China’s manufacturing sector are weak and

part of the reason is due to the large proportion of the processing indus-

tries which trade on international markets. However, there have been no

studies examining the cross- industry linkages of the Chinese iron and steel

industry with other industries. To fill the gap, this chapter aims to examine

systematically such linkages by using firm- level data and the recent input–

output table.

DATA COLLECTION, VARIABLE DEFINITION AND DESCRIPTIVE STATISTICS

The data used in this study are the firm- level data constructed by using

the Annual Manufacturing Enterprise Census, conducted by the China

National Bureau of Statistics (CNBS), for the period 2000–03. The

Annual Enterprise Census covers all state- owned firms and non- state-

owned enterprises with annual sales above 5 million yuan in Chinese

manufacturing sectors across all 32 provinces and metropolitan cities.

These enterprises accounted for more than 95 per cent of the total value

of Chinese industrial output during this time. The sample used is an



unbalanced dataset at the firm level, including those firms which enter

and exit each year. The total number of observations in the sample varies

from 134 130 in 2000 to 169 810 in 2003. At the two- digit (CICC) level,

the sample covers 26 different industries (with the CICC ranging from 13

to 42), which is split into two groups. The first group is the iron and steel

industry, including three manufacturing sectors: smelting and pressing of

ferrous metals (33), smelting and pressing of non- ferrous metals (34), and

manufacture of metal products (35); the second is its upstream and down-

stream industries including the 23 other manufacturing industries.

Table 6.1 provides some descriptive statistics of the sample: the number

of firms, the average value of output, and firm inputs such as labour,

capital and the intermediate inputs for both the iron and steel industry

and its upstream and downstream industries. The real output of firms, Y,

is defined as the total value of the sample firms’ output, deflated by the

producer price index at the firm level. Labour input, L, is defined as total

employment. As employment data are not available for 2003, we use regis-

tered labour (‘zaigang’) as a substitute. Although there are a large number

of non- productive workers in Chinese firms, there was a strong correlation

(about 95 per cent) between total employment and zaigang labour at the

firm level in 2000. Therefore, using zaigang workers as a proxy for total

employment in 2003 is appropriate. Capital, K, is defined as the value of

fixed assets at the end of the year, deflated by using the price index for

investment at the industry level. Intermediate inputs, M, is the value of

total output less value added, plus the net value- added tax, deflated by the

intermediate input deflator at the industry level. All value variables are

deflated, with 1990 the base year.

The key variable of interest in this study is the measure of firm’s pro-

ductivity, i.e. total factor productivity (TFP). Following the standard

literature in the field of growth accounting, we assume that the TFP

of representative firm i in industry j and region r at time t takes the para-

metric form of

ln TFPijrt 5 ln Yijrt 2 b̂jl ln Lijrt 2 b̂jK ln Kijrt 2 b̂jM ln Mijrt, (6.1)

where Yijrt is firm i’s output, and Lijrt, Kijrt and Mijrt are labour, capital and

intermediate inputs used in production. bjl, bjk and bjm are the estimated

elasticities of labour, capital and intermediate inputs to output by indus-

try. Although the simple ordinary least squares (OLS) regression technique

with the adjustment of heteroscedasticity can be used to provide estimates

of various input elasticities, its results are criticized for the potential over-

estimation due to the positive correlation between firms’ choice of capital

and their unobserved productivity level. As Olley and Pakes (1996) and


110

Table

6.1

S

om

e des

crip

tive

sta

tist

ics

on i

ron a

nd s

teel

indust

ry a

nd a

ll m

anufa

cturi

ng i

ndust

ries

in C

hin

a,

2000–03

2000

2001

2002

2003

All

fir

ms

Panel

A:

Fir

ms

in a

ll m

anufa

cturi

ng s

ecto

rs

A

ver

age

tota

l o

utp

ut

valu

e (m

illi

on

yu

an

)50 8

16.4

53 8

29.4

60 5

52.8

71 0

11.0

59 0

52.4

(343 9

90)

(429 0

73)

(490 5

71.9

)(6

59 5

06.4

)(4

80 7

85.3

)

A

ver

age

lab

ou

r em

plo

yed

(p

erso

ns)

325

292

285

276

294

(1141)

(1041)

(981)

(923)

(1022)

A

ver

age

cap

ital

sto

ck (

mil

lio

n y

uan

)12 1

08.5

11 7

45.0

11 7

46.2

11 6

24.3

11 8

06.0

(112 4

44.4

)(1

21 4

45.6

)(1

22 8

63.6

)(1

16 7

19.4

)(1

18 3

68.2

)

A

ver

age

inte

rmed

iate

in

pu

ts u

sage

(mil

lio

n y

uan

)39 4

69.3

41 9

40.8

46 9

14.2

54 7

80.5

45 7

76.2

(278 0

58.3

)(3

42 6

16)

(395 9

52.5

)(5

44 9

28.1

)(3

90 3

88.7

)

N

um

ber

of

ob

serv

ati

on

s134 9

52

148 9

61

155 9

22

170 8

84

610 7

19

Panel

B:

Fir

ms

in t

he

iron a

nd s

teel

ind

ust

ry

A

ver

age

tota

l o

utp

ut

valu

e (m

illi

on

yu

an

)115 8

04.5

128 1

76.8

145 4

97.5

168 2

28

141 1

69

(689 4

23.9

)(7

77 6

82.9

)(8

95 5

64)

(970 8

10.3

)(8

49 5

77.3

)

A

ver

age

lab

ou

r em

plo

yed

(p

erso

ns)

865

755

705

647

735

(5360)

(4758)

(4395)

(3883)

(4579)

A

ver

age

cap

ital

sto

ck (

mil

lio

n y

uan

)61 3

97.6

61 0

30.7

63 4

10.2

60 0

53.7

61 4

18.6

(559 4

75.3

)(6

20 1

61.7

)(6

67 8

59.9

)(6

05 2

54.7

)(6

15 5

29.7

)

A

ver

age

inte

rmed

iate

in

pu

ts u

sage

(mil

lio

n y

uan

)90 1

35.3

8100 5

26.2

112 9

66.9

128 7

03.7

109 3

63.4

(492 5

41.3

)(5

86 8

62.3

)(6

4 5

253.5

)(6

88 9

80.0

)(6

16 5

16.6

)

N

um

ber

of

ob

serv

ati

on

s2948

3340

3345

3832

13 4

65

Note

: N

um

ber

s in

pare

nth

eses

are

sta

nd

ard

dev

iati

on

s.

Sourc

e:

Au

tho

rs’

ow

n c

alc

ula

tio

n.



Levinsohn and Petrin (2003) point out, firms’ inputs like capital should

be considered as endogenous since managers choose their usage rates of

machinery based on production cost and productivity considerations that

are observed only by producers and not by econometricians. Without

properly accounting for firms’ endogenous input, choices may lead to

biased estimates of inputs elasticities and thus firms’ TFP. To deal with

this problem, we adopt the Levinsohn and Petrin semi- parametric method

to estimate the elasticity of labour, capital and intermediate inputs in each

two- digit (CICC) level industry and use Equation (6.1) to calculate the

firm- level TFP (see the appendix for more details).

To examine the cross- industry productivity spillovers of the iron and

steel industry in China, we defined two variables called the forward and

backward productivity linkage, respectively, following Javorcik (2004).

The forward productivity linkage of the iron and steel industry is

defined as the aggregated firms’ average productivity in the three com-

ponent manufacturing sectors defined above, weighted by the propor-

tion of each if these three sectors’ output supplied to the specific sector

Up_Steeljt 5 Sk,k2 jajkTFPkt, where ajk is the proportion of sector j’s output

supplied to sector k. The backward productivity linkages of the iron and

steel industry are defined as the aggregated firms’ average productivity

in the three sectors weighted by the proportion of the specific sector’s

domestic output (total output minus exports) supplied to each of the three

Down_Steeljt 5 Sk,k 2 jsjk[Si,i[kTFPikt*(Yikt2Xikt

)/(Si,i[kYikt 2 Xikt) ], where

sjk is the share of inputs purchased by sector j from sector k in total

intermediate inputs sourced by sector j. Both ajk and sjk are taken from

the 2002 input–output matrix at the two- digit industry level. The two vari-

ables of forward and backward productivity linkage capture the possible

linkage between the industry and its upstream and downstream sectors.

Tables 6.2 and 6.3 present the backward and forward linkages between

the three component sectors in the iron and steel industry in China and

other manufacturing industries, in terms of firms’ productivity and its

change over time reflected in the input–output table. There are three key

features that can be observed: (1) the forward linkage is more significant

than the backward linkage, though both of them are increasing over time.

The average forward linkage across the 26 manufacturing industries is

around 2.2 per cent, which is more than twice the forward linkage, at 1.0

per cent. (2) In terms of the cross- industry distribution of their impacts,

the forward linkage focuses on some high value- added industries, while

the backward linkage focuses on the low- value- added industries. The top

four industries being affected by the forward linkages include: metal prod-

ucts, general and special purpose machinery, communication equipment,

and computers and other electronic equipment. Those affected by the


112

Table

6.2

B

ack

ward

and f

orw

ard

lin

kages

bet

wee

n t

he

iron a

nd s

teel

indust

ry a

nd o

ther

indust

ries

in C

hin

a,

2000–03

(per

cen

t)

Ch

ina I

nd

ust

rial

Cla

ssif

icati

on

Co

de

Fo

rward

lin

kage

Back

ward

lin

kage

2000

2001

2002

2003

Aver

age

2000

2001

2002

2003

Aver

age

13

0.0

20.0

20.0

20.0

20.0

20.0

00.0

00.0

00.0

00.0

0

14

0.0

50.0

50.0

50.0

60.0

50.0

00.0

00.0

00.0

00.0

0

15

0.1

20.1

20.1

20.1

30.1

20.0

00.0

00.0

00.0

00.0

0

17

0.0

50.0

50.0

50.0

50.0

50.0

10.0

10.0

10.0

10.0

1

18

0.0

90.0

90.0

90.0

90.0

90.1

60.1

60.1

60.1

70.1

6

19

0.0

70.0

70.0

70.0

80.0

70.0

00.0

00.0

00.0

00.0

0

20

0.0

60.0

60.0

70.0

70.0

70.0

60.0

60.0

60.0

70.0

6

21

1.7

61.7

81.7

91.8

91.8

10.0

30.0

30.0

30.0

30.0

3

22

0.0

40.0

40.0

40.0

40.0

40.0

40.0

40.0

40.0

40.0

4

24

1.1

41.1

51.1

61.2

21.1

70.0

40.0

40.0

40.0

40.0

4

25

0.2

70.2

70.2

70.2

90.2

83.5

03.5

33.5

63.7

63.5

9

26

0.1

10.1

10.1

10.1

20.1

10.3

30.3

30.3

30.3

50.3

4

27

0.0

60.0

60.0

60.0

60.0

60.0

00.0

00.0

00.0

00.0

0

28

0.0

40.0

40.0

40.0

50.0

50.0

00.0

00.0

00.0

00.0

0

29

1.2

31.2

41.2

51.3

21.2

60.4

50.4

50.4

50.4

80.4

6


113

30

0.0

50.0

50.0

50.0

50.0

50.1

30.1

40.1

40.1

40.1

4

31

1.0

41.0

51.0

51.1

11.0

61.9

01.9

21.9

32.0

41.9

5

32

12.0

012.1

112.2

012.8

912.3

312.0

012.1

112.2

012.8

912.3

3

33

0.6

50.6

60.6

60.7

00.6

72.2

22.2

42.2

52.3

82.2

8

34

11.4

211.5

211.6

112.2

611.7

30.8

30.8

40.8

40.8

90.8

5

35

6.0

76.1

36.1

76.5

26.2

41.3

31.3

41.3

51.4

31.3

7

36

7.4

77.5

47.6

08.0

37.6

70.9

20.9

30.9

40.9

90.9

4

37

3.9

64.0

04.0

34.2

64.0

70.4

40.4

40.4

50.4

70.4

5

40

0.2

50.2

50.2

50.2

60.2

50.0

40.0

40.0

40.0

50.0

4

41

2.3

92.4

12.4

32.5

62.4

50.6

40.6

50.6

50.6

90.6

6

42

0.3

80.3

80.3

90.4

10.3

96.7

96.8

56.9

07.2

96.9

6

To

tal

2.1

72

.21

2.2

12.3

62.2

41.0

21.0

31.0

41.0

81.0

4

Note

: F

or

mo

re d

etail

s o

f C

hin

a I

nd

ust

rial

Cla

ssif

icati

on

Co

des

see

Tab

le 6

A.1

.

Sourc

e:

Au

tho

rs’

ow

n c

alc

ula

tio

n.



backward linkages were processing of petroleum, coking, processing of

nuclear fuel, other manufacturing industries, non- metallic mineral prod-

ucts, and general- purpose machinery. (3) The cross- industry input–output

linkages among the manufacturing sector have been weakening over time

(Table 6.3). Out of 26 industries, 20 decreased the relative share of pur-

chased intermediate inputs from domestic firms in their total inputs, and

21 decreased the share for outputs.

Finally, we define some firm- specific and industry- specific variables,

which are used to deal with the omitted variable problems due to the

potential correlation between these variables and firms’ productivity in

different regression models. More specifically, we define the Herfindahl

index of the industry as the output share of the top eight largest firms in

the industry to control for the negative relationship between the monopoly

power in market and firms’ productivity estimates. Also, we define the

number of products as a proxy for the firms’ production strategy relating

to output choice, so as to control the positive correlation between firms’

portfolio strategy in output and their productivity.

MODEL SPECIFICATION AND ESTIMATION STRATEGY

It is argued that an increase in firm productivity of the three component

sectors in the iron and steel industry helps improve firms’ productivity

in other industries outside the iron and steel industry through enforcing

market competition in the upstream industries and raising the quality

of the intermediate input supplied to the downstream firms. To measure

these backward and forward spillover effects of the iron and steel industry,

we start with applying two empirical specifications as below:

ln TFPijrt 5 b0 1 b1Forward_ISjt 1 b2Backward_ISjt 1altDt

1agrDr 1acjDj 1 eijrt (6.2)

ln TFPijrt 5 b0 1 b1Forward_ISjt 1 b2Backward_ISjt

1 b3Herfindahljt 1 b4New_Shareijrt 1 b5Exportijrt 1 b3FDIijrt 1altDt

1agrDr 1acjDj 1 eijrt (6.3)

where TFPijrt denotes the TFP of firm i operating in sector j and region

r at time t, and Forward_ISjt and Backward_ISjt denote the forward and



backward linkages of the iron and steel industry. Herfindahljt represents

the Herfindahl index for the top eight largest firms in the industry;

New_Shareijrt is the share of new products in total output; Exportijrt is the

dummy for firms’ export; and FDIijrt is the share of foreign investment

Table 6.3 Changes of domestic intermediate input shares in total inputs

and outputs, 1997–2002 (per cent)

China Industrial

Classification

Code

Intermediate/

Output share

Sign

change

1997–

2002

Intermediate/

Input share

Sign

change

1997–

20021997 2002 1997 2002

13 23.0 23.5 − 12.0 13.8 −

14 6.3 11.2 − 31.0 35.3 −

15 11.0 7.5 1 31.1 31.0 116 11.7 9.5 1 24.6 14.6 117 74.6 61.5 1 50.1 48.9 118 7.0 7.4 − 56.2 55.9 119 34.5 35.2 − 57.1 54.6 120 62.9 49.9 1 37.5 41.5 −

21 16.9 6.5 1 56.4 49.3 122 87.7 78.8 1 47.2 45.8 123 26.1 19.2 1 51.3 44.4 124 4.7 4.9 1 57.5 54.6 125 40.9 34.5 1 12.8 10.1 126 74.4 83.2 − 45.6 46.3 −

27 19.0 18.6 1 40.3 36.7 128 106.5 89.7 1 66.0 56.4 129 55.3 53.2 1 48.0 35.9 130 75.5 72.8 1 65.3 62.0 131 28.0 29.5 − 40.0 33.1 132 89.9 75.2 1 51.0 46.7 133 102.9 96.2 1 45.8 46.9 −

34 49.1 47.1 1 57.2 56.6 135 58.5 52.2 1 51.8 55.6 −

36 23.9 18.4 1 59.0 56.1 137 37.1 34.9 1 65.0 61.1 139 33.3 39.4 − 65.8 59.9 140 52.4 54.2 − 66.3 67.4 −

41 28.2 28.1 1 53.4 61.0 −

42 51.7 47.5 1 37.9 32.3 1Total 42.4 42.5 − 42.4 42.5 −

Source: Authors’ calculation with data from China Input–Output Table, 1997 and 2002, China National Bureau of Statistics.



in firms’ total capital. Dt, Dr and Dj are time- specific, region- specific

and industry- specific dummies, used to control for time- specific, region-

specific and industry- specific effects.

Equation (6.2) is a basic empirical specification, which regresses firms’

TFP directly on the backward and forward linkages of the iron and

steel industry after controlling for the time- specific, region- specific and

industry- specific effects. Equation (6.3), as a robust check, controls for

four more variables, including the Herfindahl index for the top eight

largest firms in the industry, firms’ new product share, exporting dummy,

and foreign investment share. The reason for using Equation (6.3) is

that the estimation based on Equation (6.2) may suffer from the omitted

variable problem: the potential correlation between some factors in error

terms and the independent variables may lead to over- or underestimation

of their coefficients. For example, market competitiveness is argued to be

positively related to firms’ productivity and their cross- industry linkages.

Without considering the impact of such a factor, this may lead to overesti-

mation of the impact of the iron and steel industry through the backward

and forward linkages. Also, there are some firm- level factors, such as

firms’ new product share, export behaviour and foreign investment share,

which play an important role in affecting their productivity as well as their

linkage to the iron and steel industry.

Estimation of Equations (6.2) and (6.3) by using the OLS technique may

suffer from the endogeneity problem. When the backward and forward

linkages of the iron and steel industry are correlated with the unobserved

firm, region, sector, time- variant and - invariant factors in the error term,

the estimated coefficients on the backward and forward linkages would be

biased. Usually, if the correlation happens to be positive, the results would

be overestimated; if the correlation happens to be negative, the results

would be underestimated. For example, the rapid macroeconomic growth

and the continuous microeconomic reform might promote improvement

in firms’ productivity in both the iron and steel industry and its upstream

and downstream industries. Failing to consider this issue may lead to

the overestimation of the impact of the Chinese iron and steel industry’s

development on the whole manufacturing sector. Meanwhile, the open-

door policy tends to promote firms’ productivity but decreases the cross-

industry linkages between the domestic iron and steel industry and other

manufacturing industries. Failing to take this issue into account may lead

to the underestimation of the impact of the Chinese iron and steel indus-

try’s development on the whole manufacturing sector. To deal with this

problem, we first adopt the first- differencing (FD) regression technique

to eliminate the time- invariant firm- , region- and industry- specific factors

from our estimation and then include the dummy variables of Dr and Dj



so as to control for those time- varying factors. Thus, Equations (6.2) and

(6.3) can be rearranged as follows:

d ln TFPijrt 5 b0 1 b1d ln Up_Steeljt 1 b2d ln Down_Steeljt

1altDt 1agrDr 1acjDj 1 uijrt; (6.4)

d ln TFPijrt 5 b0 1 b1d ln Up_Steeljt 1 b2d ln Down_Steeljt

1 b3dHerfindahljt 1 b4dNew_Shareijrt 1 b5dExportijrt 1 b3dFDIijrt

1altDt 1agrDr 1acjDj 1 uijrt, (6.5)

where d( . ) denotes changes in related variables over time.

To examine how firms’ characteristics, such as their scale of opera-

tion and productivity level, may affect their linkage to the iron and steel

industry, we further split our sample into subgroups: larger versus medium

and small firms, and the most productive versus the least and medium

productive firms. Regressions, based on Equations (6.2) to (6.5), are thus

reapplied to those subgroups:

ln TFPkijrt 5 b0 1 b1 ln Up_Steel

kjt 1 b2 lnDown_Steel

kjt

{1 b3Herfindahl kjt 1 b4New_Sharek

ijrt 1 b5Exportkijrt 1 b3FDIk

ijrt}

1altDt 1agrDr 1acjDj 1 eijrt; (6.6)

d ln TFPkijrt 5 b0 1 b1d ln Up_Steel

kjt 1 b2d ln Down_Steel

kjt

{1 b3dHerfindahl kjt 1 b4dNew_Sharek

ijrt 1 b5dExportkijrt 1 b3dFDIk

ijrt}

1altDt 1agrDr 1acjDj 1 uijrt, (6.7)

where k represents the subgroup of firms according to their operational

size and productivity levels.

However, the OLS and FD estimates may still overestimate the cross-

industry linkages of the iron and steel industry without making the correc-

tion for clustering effects. As Moulton (1990, p. 334) argued,

when one tended to use the aggregate market or public policy variables to explain the economic behaviour of micro units, it is possible that the standard



errors of estimated coefficients of those aggregate variables from the OLS or FD might be underestimated (or inefficient), which would lead to the over-stated significance of coefficients.

The presence of group- level variables in such a ‘structural’ model can

be viewed as putting additional restrictions on the intercepts in separate-

group models, which can cause the residual to deviate from the indepen-

dent and identical distribution (i.i.d.) assumption. Failure to address this

type of clustering effect may cause a serious downward bias in the esti-

mated errors, resulting in spurious findings of statistical significance for

the aggregate variable of interest (linkage to the iron and steel industry).

To deal with this problem, we control for both the inter- and intra- sectoral

variance in our OLS and FD regressions.

Finally, we carry out a robustness check by regressing the logarithm of

firms’ total output on their backward and forward linkages to the iron and

steel industry with the control of various inputs including labour, capital

and intermediate inputs. The purpose is to check whether the estimation

of firms’ TFP with the LP method may lead to any significant differences

in our regression results.

ESTIMATION RESULTS FOR THE BACKWARD AND FORWARD LINKAGE EFFECTS

Using Equations (6.2) to (6.5), we examine the relationship between

manufacturing firms’ productivity and their linkage to the iron and steel

industry at the aggregate level as well as by groups with different charac-

teristics as specified in the last section. The estimation results from both

the OLS and FD regression techniques with different model specifications

are presented in Tables 6.4 to 6.6.

Backward and Forward Linkages of the Iron and Steel Industry

Table 6.4 reports the estimation results at the national level obtained from

applying Equations (6.2) to (6.5). Columns (1) and (2) provide the OLS

estimates of the cross- industry linkages of China’s iron and steel industry,

with model I as the basic model specification controlling only for time-

specific, region- specific and industry- specific dummies, and model II as

the comparison model specification controlling for some additional firm-

specific and sector- specific factors such as the Herfindahl index, firms’

new product share, dummy for exporting, and foreign investment share.

The estimation results from both model specifications show that neither



the backward nor the forward linkages are statistically significant, which

seems to suggest that an increase in steel firms’ productivity had no signifi-

cant effects on the productivity of firms in the upstream and downstream

industries.

Although the OLS estimations could provide some useful information

on the cross- industry linkages of the iron and steel industry, they may suffer

from two potential problems: (1) the simultaneous bias problem whereby

Table 6.4 Iron and steel industry’s backward and forward linkages: LP

estimation

OLS regression First- differencing

regression

Model I Model II Model III Model IV


Up steel −0.273 −0.203 4.209 4.388

(2.032) (2.082) (1.018)*** (1.046)***

Down steel −3.684 −3.528 −9.374 −9.594

(2.200)* (2.238) (0.940)*** (0.915)***

Herfindahl Index – 0.004 – 0.300

– (0.060) – (0.209)

New product share – 0.100 – 0.024

– (0.018)*** – (0.008)***

Export share – 0.005 – 0.007

– (0.004) – (0.006)

FDI share – 0.063 – 0.006

– (0.007)*** – (0.005)

Sector dummies Yes Yes Yes Yes

(significant) (significant) (significant) (significant)

Region dummies Yes Yes Yes Yes


Year dummies Yes Yes Yes Yes


Constant 2.715 2.705 0.008 0.019

(0.022)*** (0.023)*** 0.013 0.013

Adjusted R- squared 0.813 0.818 0.045 0.048

No. of observations 558 702 553 584 314 589 310 180

Note: The TFP at the firm level is defined as ln TFP 5 ln Y 2 bL ln L 2 bK ln K 2 bM ln M. *, ** and *** represent the estimated coefficients statistically significant at the 10 per cent, 5 per cent and 1 per cent levels, respectively. Numbers in parentheses are standard errors.




macroeconomic growth may generate firms’ productivity improvement in

both the iron and steel industry and its upstream and downstream indus-

tries; and (2) the omitted variable problem whereby some unobserved

firm- specific factors are closely correlated with firms’ productivity and

that influence the effects of their linkages with the iron and steel industry.

To deal with these problems, we run the first- differencing regressions

following Equations (6.4) and (6.5) and report the estimation results in

columns (4) and (5) of Table 6.4.

After eliminating the simultaneous bias and omitted variable problems,

we find that the iron and steel industry in China has significant positive

forward linkages to firms’ productivity in the downstream industries but

negative backward linkages to firms’ productivity in the upstream indus-

tries. Specifically, a 1 per cent increase in firms’ average productivity in

the three component sectors may tend to increase firms’ productivity in

the downstream industries by 4.2–4.4 per cent, while decreasing firms’

productivity in the upstream industries by 9.4–9.6 per cent. This finding

suggests that the improvement of firms’ average productivity in the iron

and steel industry will help to foster productivity improvements in the

downstream industries but harm that in the upstream industries.

This finding seems to contradict the view that a rapid development of

the iron and steel industry helps promote firms’ productivity in the whole

manufacturing industry of a country through its cross- industry linkages.

This is because as the iron and steel industry is developed, it is expected

that the industry generates more demand for the upstream industries and

provides higher- quality products (as intermediate inputs) to the down-

stream industries. However, our empirical results show that such an effect

in China is more likely to take place through the forward than the back-

ward linkages. In other words, an increase in the productivity of the iron

and steel industry only tends to improve firms’ productivity in the down-

stream industries, while decreasing firms’ productivity in the upstream

industries. A possible explanation for this phenomenon is that the iron

and steel industry in China has depended so much on imported iron ore

and other materials from international markets that the productivity of

domestic suppliers in the upstream industry has been negatively affected.

WHO CAN BENEFIT MORE? LARGE vs. SMALL FIRMS OR HIGH vs. LOW PRODUCTIVITY FIRMS

The estimates reported refer to the average impact of the iron and steel

industry on firms’ productivity in other manufacturing industries. A

further question one may ask is whether these results mask the heteroge-



neity of the linkage effects across firms with different characteristics. To

answer this question, we now split our sample into subgroups according to

firms’ operational size and productivity level and re- estimate the linkage

effects of the iron and steel industry with different samples based on

Equations (6.2) to (6.5). The estimation results from the FD regressions

are shown in Tables 6.5 and 6.6.

As shown in Table 6.5, the estimated linkage effects from the FD regres-

sions with the large firms and with the medium and small firms show that

medium and small firms in the downstream industries are more likely

to benefit from the forward linkages of the iron and steel industry than

large ones. A 1 per cent increase in the industry’s average productivity

may increase the productivity of medium and small firms by 4.4–4.6 per

cent, which is around twice that for large firms (at 2.5–2.7 per cent). This

phenomenon may be explained by the fact that most small and medium

firms in the Chinese manufacturing sector had already been privatized as

part of the industrial transformation. These privatized firms are more flex-

ible in choosing their input mix and thus likely to make more use of the

innovation or positive externalities passing through the upstream iron and

steel industry to improve their own productivity. In terms of the negative

backward linkages, there is no significant difference for firms with dif-

ferent operation sizes. As shown in Table 6.5, a 1 per cent increase in the

average productivity of the iron and steel industry may decrease produc-

tivity for all sized upstream firms by 9.1–9.5 per cent. This finding suggests

that such a negative backward linkage is likely to be independent of firms’

operational scale.

The next question we may ask is whether the heterogeneity in firms’

productivity affects the potential gains in productivity from the iron and

steel industry through the backward and forward linkages. Table 6.6

presents the estimation results from the FD regressions for firms with

low (x , 25 per cent), medium (25 per cent , x , 75 per cent) and high

levels of productivity, where x is the measure of productivity, relative

to the full sample. Two interesting findings are apparent here. First, the

positive forward linkage effects of the iron and steel industry decrease with

changing levels of firm productivity. A 1 per cent increase in the average

productivity of the iron and steel industry may increase the productivity

of firms with low, medium and high levels of productivity by 5.3, 4.1 and

3.4 per cent, respectively. This result implies that firms with relative low

productivity are likely to benefit more from the forward linkages with the

iron and steel industry. Second, the negative backward linkage effects are

more severe for firms with a high level of productivity than for those with

medium and low productivity. A 1 per cent increase in the average produc-

tivity of the iron and steel industry may decrease the productivity of firms



with low, medium and high levels of productivity by 8.8, 8.1 and 16.9 per

cent, respectively. This result implies that productivity improvements in

the iron and steel industry are more likely to negatively affect those firms

with relative high productivity in the upstream industry.

The above findings seem to suggest that firms with low productivity in

the Chinese manufacturing sector may benefit more from the backward

and forward linkages of the iron and steel industry. It remains to be seen

how the ongoing reform in the iron and steel industry or even in the whole

industrial sector could further strengthen the cross- industry linkage of the

Table 6.5 Iron and steel industry’s backward and forward linkages by

firm size

Large firms Medium and small firms

Model I Model II Model I Model II


Up steel 2.549 2.727 4.372 4.551

(1.799)* (1.789)* (0.955)*** (0.989)***

Down steel −9.112 −9.236 −9.367 −9.591

(1.520)*** (1.509)*** (0.913)*** (0.882)***

Herfindahl Index – 0.337 – 0.289

– (0.213) – (0.217)

New product share – −0.010 – 0.033

– (0.015) – (0.009)***

Export share – 0.040* – 0.003

– (0.022) – (0.006)

FDI share – 0.002 – 0.003

– (0.022) – (0.005)

Sector dummies Yes Yes Yes Yes


Region dummies Yes Yes Yes Yes


Year dummies Yes Yes Yes Yes


Constant 0.042 0.042 0.019 0.007

(0.021)* (0.022)* (0.012) (0.013)

Adjusted R- squared 0.029 0.032 0.048 0.051

No. of observations 47 127 46 752 267 462 263 428

Note: *, ** and *** represent the estimated coefficients statistically significant at the 10 per cent, 5 per cent and 1 per cent levels, respectively. Numbers in parentheses are standard errors.



123

Table

6.6

Ir

on a

nd s

teel

indust

ry’s

back

ward

and f

orw

ard

lin

kages

by f

irm

s’ p

roduct

ivit

y

Lo

w p

rod

uct

ivit

y (

,25%

)M

ediu

m p

rod

uct

ivit

y (

25–75%

)H

igh

pro

du

ctiv

ity (

.75%

)

Mo

del

IM

od

el I

IM

od

el I

Mo

del

II

Mo

del

IM

od

el I

I

Dep

end

ent

vari

ab

le:

ln T

FP

Up

ste

el

4.7

05

5.3

39

3.9

84

4.0

66

3.3

25

3.4

26

(1.0

92)*

**

(1.1

22)*

**

(0.9

32)*

**

(0.9

76)*

**

(2.4

45)

(2.4

81)

Do

wn

ste

el−

8.0

88

−8.8

43

−7.9

71

−8.0

63

−16.7

71

−16.8

64

(1.4

10)*

**

(1.2

61)*

**

(0.8

32)*

**

(0.8

32)*

**

(2.1

55)*

**

(2.1

23)*

**

Her

fin

dah

l in

dex

–0.9

93

–0.0

80

–0.1

38

–(0

.468)*

–(0

.146)

–(0

.634)

New

pro

du

ct s

hare

–0.0

28

–0.0

27

–0.0

05

–(0

.012)*

**

–(0

.010)*

**

–(0

.016)

Exp

ort

sh

are

–0.0

02

–0.0

10

–0.0

08

–(0

.009)

–(0

.005)*

*–

(0.0

16)

FD

I sh

are

–0.0

00

–0.0

05

–0.0

14

–(0

.012)

–(0

.006)

–(0

.012)

Sec

tor

du

mm

ies

Yes

(si

gn

ific

an

t)Y

es (

sign

ific

an

t)Y

es (

sign

ific

an

t)Y

es (

sign

ific

an

t)Y

es (

sign

ific

an

t)Y

es (

sign

ific

an

t)

Reg

ion

du

mm

ies

Yes

(si

gn

ific

an

t)Y

es (

sign

ific

an

t)Y

es (

sign

ific

an

t)Y

es (

sign

ific

an

t)Y

es (

sign

ific

an

t)Y

es (

sign

ific

an

t)

Yea

r d

um

mie

sY

es (

sign

ific

an

t)Y

es (

sign

ific

an

t)Y

es (

sign

ific

an

t)Y

es (

sign

ific

an

t)Y

es (

sign

ific

an

t)Y

es (

sign

ific

an

t)

Co

nst

an

t−

2.0

91

−2.1

12

−0.6

68

−0.6

64

0.0

07

0.0

09

(0.1

47)*

**

(0.1

42)*

**

(0.0

15)*

**

(0.0

15)*

**

(0.0

14)

(0.0

14)

Ad

just

ed R

- sq

uare

d0.1

40

0.1

46

0.0

68

0.0

70.1

94

0.1

93

No

. o

f o

bse

rvati

on

s77 8

87

77 0

63

155 3

88

153 1

39

81 3

14

79 9

78

Note

: *,

** a

nd

*** r

epre

sen

t th

e es

tim

ate

d c

oef

fici

ents

sta

tist

icall

y s

ign

ific

an

t at

the

10 p

er c

ent,

5 p

er c

ent

an

d 1

per

cen

t le

vel

s, r

esp

ecti

vel

y.

Nu

mb

ers

in p

are

nth

esis

are

sta

nd

ard

err

ors

.

Sourc

e:

Au

tho

rs’

ow

n e

stim

ati

on

.



iron and steel industry with both the upstream and downstream industries

to the extent that the improved productivity of the iron and steel industry

spreads more positively to the other sectors of the economy.

ROBUSTNESS CHECKS

To check whether the above estimation results are sensitive to our specific

productivity estimation method, we have redone all estimations with the

logged output as the dependent variable (with the control of the three

inputs including labour, capital and intermediate inputs). The results show

that the new estimation results with the logged output as the dependent

variable are generally consistent with those with the estimated TFP as the

dependent variable, except that the standard errors of most coefficients

become larger. This result suggests that our initial estimation on the back-

ward and forward linkages of the iron and steel industry is stable.

CONCLUSIONS

Using the data from the Annual Manufacturing Enterprise Census, this

chapter has examined the cross- industry linkages of the iron and steel

industry in the Chinese manufacturing sector from an empirical perspec-

tive. After controlling for the potential endogeneity problems, we found

that a rapid increase in average productivity of the iron and steel industry

can promote firms’ productivity in the downstream industry but harm

firms’ productivity in the upstream industry, which can be partly explained

by the high degree of dependence of China’s iron and steel industries on

imported materials such as iron ore, which may negatively affect domestic

suppliers, especially the better- performing ones. This import substitution

policy has been implemented primarily because domestic supplies are

inferior in terms of both quantity and quality.

When firms’ heterogeneity in operational size and productivity are

considered, our results have shown that medium and small firms with low

productivity are more likely to benefit from the further development of

the Chinese iron and steel industry. A policy implication would be that

China should try to deepen the ongoing industrial reform in the iron and

steel industry as well as in the whole industrial sector in order to further

strengthen the cross- industry linkages of the iron and steel industry with

both the upstream and downstream industries. This would allow the

improved productivity of the iron and steel industry to spread more posi-

tively to the other sectors of the economy.



NOTE

1. China has become the largest manufacturing producer in the world, accounting for about 17 per cent of total world manufacturing output in 2010 (UNSD, 2011).

REFERENCES

China Iron and Steel Association (2007), China Iron and Steel Statistical Yearbook, Beijing: China Iron and Steel Association Press.


Harrison, A., L. Du and G. Jefferson (2008) ‘Does foreign direct investment spill over to domestic manufacturing firms? Investigation on vertical spillovers for China’, paper presented to the International Conference on Investments, Technology Spillovers and East Asian FTA, Fudan University, Shanghai, China, 10–11 October.

Hu, A.G., G.H. Jefferson and J.C. Qian (2005), ‘R&D and technology transfer: firm- level evidence from Chinese industry’, Review of Economics and Statistics, 87 (4), 780–6.

Javorcik, B.S. (2004), ‘Does foreign direct investment increase the productivity of domestic firms? In search of spillovers through backward linkages’, American Economic Review, 94 (3), 605–27.

Krugman, P. (1991), Geography and Trade, Cambridge, MA: MIT Press.Krugman, P. (1998), ‘What’s new about the new economic geography?’, Oxford

Review of Economic Policy, 14 (2), 7–17.Levinsohn, J. and A. Petrin (2003), ‘Estimating production functions using inputs

to control for unobservables’, Review of Economic Studies, 70 (2), 317–42.Moulton, B.R. (1990), ‘An illustration of a pitfall in estimating the effects of aggre-

gate variables on micro units’, Review of Economics and Statistics, 72 (2), 334–8.Olley, S.G. and A. Pakes (1996), ‘The dynamics of productivity in the telecommu-

nications equipment industry’, Econometrica, 64 (4), 1263–97.United Nations Statistics Division (UNSD) (2011), commodity trade statistics

database (COMTRADE), http://comtrade.un.org/db/default.aspx.Venables, A.J. (1996) ‘Equilibrium locations of verticality linked industries’,

International Economic Review, 37 (2), 341–59.



APPENDIX

Ordinary least squares (OLS) is inappropriate for estimating the impacts of

labour and capital on productivity, since they are factors of production and,

as such, should be treated as endogenous. Olley and Pakes (1996) (OP), fol-

lowed by Levinsohn and Petrin (2003) (LP), point out that inputs like capital

should be considered endogenous, since management chooses their levels or

usage rates based on cost and productivity considerations that are observed

by the producer but not by the econometrician. Productivity estimates may

be biased if the endogeneity of input choice is not taken into account.

To address this concern, we employ a semi- parametric estimation pro-

cedure suggested by LP. Compared with the OP approach, LP allows for

firm- specific productivity differences that exhibit idiosyncratic changes

over time, and use intermediate inputs rather than long- term capital invest-

ment to proxy for unobserved productivity. We follow the LP method for

two reasons: (1) a special feature of firms in China is that their investment

behaviour is highly influenced by government policy (such as policy loans

provided by the state- owned banks to SOEs), so that investment might

not be monotonic with respect to productivity; and (2) our data cover

only four years, which is not long enough for firms to make capital adjust-

ments, especially in regard to long- term investments in physical structure

and machinery. More specifically, we assume that the production takes the

form of a Cobb–Douglas production function whose natural- logarithmic

form after taking the first- order differentiation is:

yit 5 bc 1 bllit 1 bkkit 1 bmmit 1 .it 1 uit, (6.A1)

where bc measures the mean efficiency level across firms and over time,

.it represents firm- level productivity, and uit is a component following

an independent and identical distriubtion, which represents unexpected

deviations from the mean due to measurement error, unexpected delays

or other external circumstances. The three components are combined to

determine the time- specific and producer- specific outputs.

In order to estimate Equation (6.A1), we further assume that capital is

a state variable that is affected only by current and past levels of unob-

served productivity (.it) and is monotonic with respect to the intermediate

inputs. We define:

vit 5 gt(kit,.it

)

(t 5 1, . . ., T) , (6.A2)

where vit is a vector of proxy variables (that is, intermediate inputs) and

g(∙,∙) is time- invariant. Thus, the choice of intermediate inputs depends on



capital and productivity. Provided that the choice of intermediate inputs

is strictly increasing in productivity and conditional on capital, the rela-

tionship between vit and .it can be inverted. Thus, we have .it 5 ht(kit,vit

) ,

where ht(.,.) 5 g21

t( . , . ) . Substituting this information into Equation

(6.A1), we have:

yit 5 b0 1 bllit 1 bkkit 1 bmmit 1 ht(kit,vit

) 1 uit. (6.A3)

Estimation of Equation (6.A3) is carried out in two stages. In the first

stage, we define �(mit,kit) 5 b0 1 bkkit 1 bmmit 1 ht

(vit,kit) (in LP). Thus,

the OLS method can be used to estimate:

yit 5 bllit 1 �(vit,kit) 1 uit, (6.A4)

where �( . , . ) is approximated by a higher- order polynomial in vit and

kit (including a constant term). Estimation of Equation (6.A4) results in

a consistent estimate of the coefficients on labour. In the second stage,

assume that productivity follows a first- order Markov process, that is,

.it11 5 E(.it11 0 .it) 1 xit11, where xit11, representing the news compo-

nent, is assumed to be uncorrelated with productivity and capital in period

t 1 1. Thus, the estimation algorithm can be written as:

E [yit11 2 bllit11] 5 b0 1 bkkit11 1 E(.it11

0.it) 1 xit11 1 uit11,

(6.A5)

where E(.it110.it

) 5 q(�it 2 bkkit 2 bmmit) follows from the law of

motion for the productivity shock. As the first stage of the estimation

procedure has used a higher- order polynomial expansion in �̂it 2 b̂kkit

or �̂it 2 b̂kkit 2 b̂mmit to approximate g(∙,∙), the capital coefficients can

then be obtained by applying non- linear least squares (NLS) to Equation

(6.A6):

yit11 2 bllit11 5 b0 1 bkkit11 1 bmmit11 1 q(�it 2 bkkit 2 bmmit)

1 xit11 1 uit11. (6.A6)

By using the LP method, we can obtain accurate production function

estimates that can be used in turn to estimate domestic productivity.

For convenience, a table of concordance between CICC code and

industry names is also provided (Table 6.A1).



Table 6.A1 China Industrial Classification Codes and industry names

CICC Industry name

13 Processing of food from agricultural products

14 Foods

15 Beverages

16 Tobacco

17 Textile

18 Textile wearing apparel, footwear, and caps

19 Leather, fur, feather and related products

20 Timber, manufacture of wood, bamboo, rattan, palm, and straw

products

21 Furniture

22 Paper and paper products

23 Printing, reproduction of recording media

24 Articles for culture, education and sport activity

25 Processing of petroleum, coking, processing of nuclear fuel

26 Raw chemical materials and chemical products

27 Medicines

28 Chemical fibres

29 Rubber

30 Plastics

31 Non- metallic mineral products

32 Smelting and pressing of ferrous metals

33 Smelting and pressing of non- ferrous metals

34 Metal products

35 General- purpose machinery

36 Special- purpose machinery

37 Transport equipment

39 Electrical machinery and equipment

40 Communication equipment, computers and other electronic

equipment

41 Measuring instruments and machinery for cultural activity and office

work

42 Artwork and other manufacturing

Total All manufacture

Source: China National Bureau of Statistics.


129

7. China’s shift from being a net importer to a net exporter of steel and its implications

Haimin Liu and Ligang Song

INTRODUCTION

China for a long time prior to the 1990s was an importer of steel products,

reflecting the fact that its domestic supply could not meet its demand

in terms of both quantity and quality of steel products. From 1990, the

country began to export steel products, but in the following 15 years

remained a net importer. With the rapid growth of its exports, and the

local industry’s increasing ability to compete with imports, China became

in 2005 a net exporter of steel products (see Figure 7.1). The massive

increase in China’s steel exports since then reflects the fact that the coun-

try’s underlying comparative advantage has begun shifting from labour-

intensive to capital- intensive production, which coincides with the mid

phase of industrialization characterized by the level of China’s per capita

income in the second half of the first decade of the twenty- first century.

Such an increase in exports of steel products from China has caused

some trade friction, especially with the industrialized countries of Europe

and North America. Since China surpassed the United States to become

the world’s largest consumer of steel products in 2001 and has remained

the largest consumer of steel since then, one may ask why China has

shifted from being a net importer to a net exporter of steel products and

what the implications of such a shift might be for both China and its

trading partners in terms of economic restructuring. The purpose of this

chapter is to address this question and to discuss some implications associ-

ated with the shift.

To do so, we need to ask the following questions: On w h a t basis has

China become the world’s largest exporter of steel products? What

are China’s comparative and competitive advantages in producing and

exporting steel products? Will this comparative advantage endure, given

the current trend of rising production costs including labour and energy



costs and the need for structural change in response to the requirement for

addressing climate change in China? Why and how does the Chinese gov-

ernment try to control the exports of steel products? What does the future

hold with regard to China’s role as an exporter of steel products?

PATTERN AND TREND OF CHINA’S EXPORTS OF STEEL PRODUCTS

China developed a fairly strong base for its steel industry development

during the period of the central planning system (1950–76). However, as a

developing country, China’s steel industry has historically lagged behind

the global frontier in terms of its level of technology and equipment used

in production. As a result, despite a large quantity of steel being produced,

China was consistently a net importer of steel products prior to 2005

and relied heavily on importing high- quality steel products in particular

from Japan. During the period 1978–2004, China imported a total of 495

million tonnes of crude steel (equivalent),1 accounting for 17 per cent of

its apparent consumption of steel. China’s net imports of steel reached

369 million tonnes during the same period. The large quantity of China’s

demand for imported steel was extremely beneficial for the international

0.4

8

0.5

1

0.6

7

1.2

8

1.4

1

0.8

8

0.2

6

0.2

4

0.2

6

0.3

6

0.8

3

1.1

1

3.1

7

4.8

6

4.5

3

2.8

4

5.4

7

11.4

1

7.5

6

9.3

3

6.1

0

5.6

1

11.8

0

7.8

2

7.1

9

8.9

1

21.2

9

29.0

3

54.6

4

72.8

7

64.1

1

26.2

1

45.4

2

(12.6

7)

(11.7

9)

(6.8

9)

(4.4

8)

(5.1

6)

(12.9

2)

(17.7

5)

(27.0

7)

(24.8

2)

(15.3

1)

(10.9

6)

(10.6

4)

(5.0

8)

(4.3

8)

(10.3

6)

(42.4

0)

(28.9

5) (1

6.6

7)

(18.5

1)

(15.2

5)

(14.3

7)

(18.3

0)

(22.1

7)

(26.8

4)

(30.8

8)

(45.5

8) (35.0

7)

(28.7

8)

(20.0

0)

(18.1

3)

(16.6

1)

(23.3

5)

(18.1

6)

–50

–30

–10

10

30

50

70

1978

1980

1982

1984

1986

1988

1990

1992

1994

1996

1998

2000

2002

2004

2006

2008

2010

mt

Exports Imports

Sources: CISA (various years); trade statistics from China Customs.

Figure 7.1 China’s imports and exports of steel, 1978–2010 (million

tonnes)


The shift from net importer to net exporter 131

steel industry especially during the economic downturns of the OECD

countries in 1993 and 2003, respectively.2

Since 2004, China’s steel exports have increased at a faster rate than its

steel production, while steel imports have fallen continuously and have

consolidated at the level of about 20 million tonnes, which was maintained

during the period before 2000. In 2005, China ended its decades- long

history as a net importer of steel, while simultaneously replacing Japan as

the largest exporter of steel in the world in the period 2006–08.3 Despite

the sharp decrease in production and trade resulting from the global

financial crisis (GFC) in 2008/09, China’s steel trade has kept a favourable

balance in volume terms (Figure 7.1 and Table 7.1).

To measure a country’s dependence on exports of steel products, one

can use the steel export ratio, defined by export/production. China’s steel

export ratio reached its peak level at 15 per cent in 2007, which is far from

the highest as compared with many other steel- producing countries. For

example, in the same year, the steel export ratios of Taiwan, Canada,

Russia, Ukraine, Japan and South Korea all greatly surpassed China’s

(Figure 7.2). However, what is significant for China is the fact that the

absolute scale of China’s steel production (the denominator) reached 37

per cent of total w o r l d steel o u t p u t in that year. Due to the large quantity

of steel being produced in China, any additional increase in the export

ratio can have a destabilizing impact on the global market for steel.

The main destinations of China’s steel exports include its neighbouring

countries and regions such as South Korea, the ASEAN- 10 countries,

Table 7.1 China’s steel trade balance, 1978–2004 and 2005–10 (million

tonnes and per cent)

Year Production Imports Exports Net

exports

Apparent

consumption

Imports/

apparent

consumption

(%)

Exports/

Production

(%)

1978–

2004 2448 495.2 126.2 369 2817 17.58 5.15

2005 353 28.8 29.0 0.2 353 8.15 8.22

2006 419 20.0 54.6 34.6 384 5.20 13.04

2007 489 18.1 72.8 54.7 434 4.17 14.89

2008 503 16.6 64.1 47.5 456 3.65 12.74

2009 572 23.3 26.2 2.8 569 4.10 4.58

2010 626 18.1 45.4 27.3 599 3.03 7.24

Source: Calculated based on data from CISA (various years) and trade statistics from China Customs.



Hong Kong, Macao, Taiwan, India, the Commonwealth of Independent

States (CIS) and Middle Eastern countries. Among them, South Korea

and the ASEAN- 10 are the largest destinations for China’s steel exports.

In 2010, China’s steel exports to these countries accounted for 67 per cent

of China’s total steel exports (see Figure 7.3). Notably, China has also

exported steel products to both the European Union and North American

markets, as well as to markets such as the Middle East and India. Such

diversification of exporting destinations reflects the fact that the Chinese

steel industry is gaining competitive advantage as well as advantage in

terms of quality of products on world markets. It is therefore not surpris-

ing that the sudden surge in China’s steel exports has generated consider-

able anxiety in the world steel industry.

Chinese enterprises have an obvious price advantage in international

markets, but they have had a significant impact on similar industries and

enterprises in importing countries, which can easily trigger trade protec-

tionism in these countries. China has become the major target country of

anti- dumping suits around the world. According to the statistics released

by the WTO, between 1995 when the WTO was founded and June 2008,

there were 3305 anti- dumping investigations launched worldwide, of

which 640 cases were against China, or nearly a fifth of all cases (Li and

Song, 2011). Some countries such as the United States took anti- dumping

measures against the imports of steel products from China. Figure 7.3

reports the major destinations of China’s steel exports in 2010 (per cent).

55.9

48.1

37.432.4 31.2

28.2

17.1

14.3

13.14.0

0

10

20

30

40

50

60

050

100150200250300350400450500

Per

cen

t

mil

lio

n t

on

Taiwan

, Chin

aC

IS

South K

orea

Brazi

l

Japan

Turkey

EU27

Chin

a

India

NA

FTA

Production Extra-regional exports Export ratio (right)

Sources: Calculated using original data from WSA (2009); CISA (2009).

Figure 7.2 Ordering of export ratio by countries, 2007 (million tonnes

and per cent)



COST FACTORS AND THE COMPETITIVENESS OF CHINA’S STEEL INDUSTRY

China’s shift from being a net importer to a net exporter of steel has

been fundamentally determined by the cost factor, which in turn reflects

the changing pattern of China’s underlying comparative advantage.

Consistent with this changing pattern of comparative advantage, the

pattern and composition of China’s exports have also changed over

time – from the predominant reliance on primary goods such as petro-

leum and agricultural products at the beginning of the reform period; to

labour- intensive products such as textiles and clothing during the first

two decades of reform; further to capital- intensive products such as steel,

machinery and automobiles in the current phase of industrialization;

and increasingly to technology- intensive products such as some high-

tech equipment, bio- products and green technology (Li and Song, 2011).

This dynamic change in China’s comparative advantage in producing

and exporting capital- intensive goods has made the production of these

outputs such as steel products more competitive on international markets

through taking advantage of relatively low costs.4

We now analyse how the steel firms in China have increased their

Middle East10.3%

Other19.8%

Taiwan, Hong Kong& Macao 4%

South Korea19.9%

EU279.1%

India8.0%

NAFTA3.9% ASEAN10

19.4%

CIS3.3%

Japan1.9%

Source: Calculated from original data from China Customs.

Figure 7.3 Major destinations for China’s steel exports in 2010

(per cent)



competitive status on international markets through lowering the costs of

production and enhancing technological change in the industry. There are

two main factors which contribute to the industry’s increasing competitive

advantage in producing and exporting steel products. The first one is its

firms’ notable advantage in adopting low costs of production, especially

with respect to the low cost of labour; the other is the industry’s improved

productivity since the late 1990s, resulting from industry reform and tech-

nological change.

China’s steel industry undeniably enjoys the benefits from low labour

costs relative to its competitors. Coinciding with the period of reform,

China went through a phase of development during which it has benefited

tremendously from the rapid pace of urbanization, when cheap labour

flowed into the industrial sector from rural areas. The seemingly unlimited

supply of labour kept industrial wages at a relatively low level, thereby

lowering the costs of production, raising aggregate labour productivity

and adding to the competitiveness of the Chinese industries including the

steel industry. Over the same period, China also benefited from the ‘demo-

graphic dividend’ whereby the large share of the population that was of

working age contributed positively to the supply of labour and national

savings, thereby further strengthening the competitiveness of Chinese

industries. For example, the steel industry’s labour productivity, measured

by the average steel production per steel worker, was 550 tonnes for large

and medium steel enterprises in 2010, up from around 100 tonnes at the

end of the 1990s. Despite the progress being made in raising the labour

productivity over this period, the current average level of productivity in

China’s steel industry in 2010 was equal to only one half of the productiv-

ity level of large foreign steel mills. However, the average wage of Chinese

steel workers was still much lower, at around 20 per cent of levels prevail-

ing in advanced countries. The net effect was that Chinese steel- makers’

labour costs per unit of production were still about 50 per cent lower than

their American and European competitors’. In this regard, the steel indus-

try is not an outlier, as this comparison of costs of production is rather

consistent across the industrial sectors in China.

Similar to the situation in other industrial sectors, the improvement in

productivity in the steel industry has been due mainly to the reform and

industrial restructuring carried out as part of the overall industrial sector

reform. These reform measures include market entry, internal restructur-

ing such as shareholding reform, privatization, corporate government

reform and openness to trade. As a result, China’s steel industry not only

has increased its scale of production, but also has made rapid inroads into

the productivity gap which existed between firms in China and the world’s

most technologically advanced producers.



Increased competition has prompted firms to invest more in production

equipment, to increase greater energy efficiency, to enhance labour pro-

ductivity and to improve product quality. For example, within a period

of less than ten years, China’s ratio of continuous casting in total steel

production rose from 85.3 per cent in 2000 to 97.4 per cent in 2009, which

is above the world average of 94.1 per cent.5 Furthermore, technological

upgrading has enabled the average intensity of energ y consumption for

members of the China Iron and Steel Association (CISA)6 to fall from 885

kg of coal equivalent (kgce) in 2000 to 605 kgce7 per tonne of crude steel

in 2010, which is below the global a v e r a g e level of intensity of energy con-

sumption.8 Accordingly, the energy consumption in the main production

processes of these firms also decreased considerably (Table 7.2). These

efficiency gains, resulting from the industrial reform, increased competi-

tion and technological progress, added to the strong competitiveness of

the industry driven by the relatively low cost of production, especially with

respect to the cost of labour.

The industrial reform has also led to a fall in fixed expenses as a pro-

portion of the total sales for the industry. These fixed expenses include

the administrative, financial and operational costs (including marketing).

Therefore, the reduction of these expenses is indicative that the industry

has improved its administrative, financial and operational efficiency

(Figure 7.4). The data show that of the three measurements, the admin-

istrative costs fell most steeply, decreasing from 8.9 per cent of total sales

in 2000 to 2.9 per cent in 2010. This is significant, as high administrative

costs were associated with most of the state- owned enterprises in the past.

While the financial expenses increased (which may reflect the increasing

costs of inputs), the costs of marketing fell over this period, which is a clear

indicator of how firms have improved their level of efficiency in selling

their products.

Another source of competitiveness of the steel industry in China could

be that a considerable number of steel enterprises, especially small and

Table 7.2 Change of energy intensity in the steel- making processes

Indicator Unit 2000 2010 Change (%)

S intering kgce/tonne sinter 69 52 −24

Iron- making kgce/tonne iron 466 408 −12

Converter steel- making

and casting kgce/tonne crude steel 29 −0.2 −101

Steel rolling kgce/tonne finished steel 118 62 −48

Source: CISA (various years).



medium ones, have violated state regulations by improperly cutting the

expenditures required for environmental protection. This is of course not

an implicit subsidy to the firms provided by the central government, as

policy- makers at the central level are keen to stamp out such inopportune

behaviour. The problem lies at local/provincial government levels where

laxity in implementing the state environmental regulations arises owing to

the consideration that implementing such regulations may put local firms

in a disadvantageous position by increasing the cost of production. This is

therefore an area where local governments are operating in conflict with

the national interest. Generally, China’s steel industry has been rather

backward in terms of reducing emissions of carbon monoxide and dust

and their efforts in controlling sulphur dioxide only began at the begin-

ning of the current century, while in developed countries the old pollution

problems have been solved and much consideration is given to the treat-

ment of carbon dioxide and dioxins. It is likely that China’s performance

in this area will fall short of Beijing’s goals on reduction of emissions as

long as provincial governments act to protect local interests.

Furthermore, there are two policy factors that have played a role in

determining the exports of steel products by Chinese firms. First, a com-

monly cited factor is that the yuan has been undervalued, which gives the

Chinese firms a competitive edge in exporting their products to world

markets. In examining whether this perception is correct, one needs to

look at the long- term trends in changes in currencies and the level of

per capita income. Figure 7.5 portrays such a relationship, using cross-

country data, which show that there is in general a positive correlation

between the appreciation of a country’s currency and the increases in its

0

2

4

6

8

10

12

8.9

3.42.9

Administration/sale

Per

cen

t

2000 2008 2010

3.0

1.6

5.5

Financial/sale

2.01.1 0.8

Selling/sale

Source: CISA (various years).

Figure 7.4 Reduction of fixed expenses of CISA members, 2000 to 2010

(per cent of total sales)



level of per capita income. In the case of China, the value of its currency

versus its purchasing power parity (PPP) rate in 2008 (0.55) was basically

consistent with its level of per capita income (US$3259) in PPP terms as

compared with a PPP of unity for the United States with per capita level

of income reaching US$47 440 in 2008. Figure 7.5 also suggests that the

exchange rate of the yuan with the US dollar will not rise to 100 per cent

of its PPP level before China becomes a high- income country. Even were

the yuan to rise to this level, China, as a low- income country (in per capita

terms), would continue to enjoy its labour cost advantages, as the average

wage of Chinese steel workers is unlikely to catch up with those in devel-

oped countries in the foreseeable future. Furthermore, the Chinese steel

industry’s productivity will continue to converge towards the level of its

competitors in those advanced countries as a result of the ongoing reform

and technological progress. This will allow Chinese steel firms to continue

to enjoy the advantage of catching up with the most advanced technology

applied in the developed countries.

The second factor in determining the level of Chinese exports of steel

products is China’s demand for energy and resources, which has been

rapidly rising owing to the country’s rapid economic growth since 2003. In

response to this growing resource demand, China has adopted measures

to encourage imports and restrict exports for raw materials. In 2004, for

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

0

Cu

rren

t ex

cha

ng

e ra

te/P

PP

, p

er c

ent

GDP per capita

China: 3259, 0.55

USA: 47 440, 1.0

Trend line

10 000 20 000 30 000 40 000 50 0005000 15 000 25 000 35 000 45 000

Source: Calculated using original data taken from the database of the IMF: www.imf.org/external/data.htm.

Figure 7.5 Relationship between currency valuation and GDP per capita,

2008



the first time, the Chinese government applied the differentiated export

rebate of value added tax (VAT) as a policy instrument to restrict exports

of resource products. The VAT rebate rates for most primary products

exported were reduced and for some products they were eliminated com-

pletely. Further, in November 2006 China started to impose export tariffs

on some resource products on which the VAT rebate had been reduced to

zero, including pig iron, ferroalloy, coke, semifinished steel and common

long steel products. For example, a 40 per cent tariff was imposed on

coking coal exports without any rebate of VAT. This policy change seems

to have increased the export cost of coking coal by at least 40 per cent,

thereby disadvantaging foreign steel- makers using Chinese coking coal.9

In response to this policy change, steel- makers in America and Europe

claimed that taxing exports of coke equated essentially to ‘government

subsidies’ to the Chinese steel- makers as steel- makers in other countries

faced much higher costs of importing coke from China. The steel- makers

in those developed countries insisted that were there no subsidies then

China’s steel products would be far less competitive in international steel

markets. For example, a 2007 report from the German Steel Federation

claimed: ‘Unfair prices of this extent are principally possible because

China subsidises its state- owned steel companies, resulting in massive

overcapacities there’.10 In the same year, a report commissioned by a US

lobby group claimed that the Chinese government had subsidized the steel

industry by the equivalent of US$52 billion.11 In January 2009, a similar

report on behalf of the EUROFER (European Confederation of Iron and

Steel Industries) claimed that Chinese steel products would be unable to

enter the European market without government subsidies.12

It is true that, prior to China’s accession to the WTO at the end of 2001,

the Chinese government provided subsidies of various kinds to the then

state- owned enterprises including steel- makers. However, the Chinese

government took important steps in reforming the system of subsidies in

relation to the state- owned enterprises as part of its efforts to comply with

the requirements of the WTO prior to its accession. Since the accession, the

government has removed all of the direct subsidies to its enterprises, and

later unified company income taxes, which now apply to all kinds of firms,

including foreign firms operating in China. Even though there have been

no direct subsidies provided to the firms by the government, Chinese firms

may gain some competitive advantages from the partially reformed factor

market system, which allows the firms to utilize some of the key resources

such as energy and materials at below- market prices (Huang and Wang,

2010). As for the steel sector, the study by Liu (2008) shows that the key

factors underlying the strong competitiveness of the Chinese steel industry

come from the steel industry itself rather than the Chinese government.



THE POLITICAL ECONOMY OF CHINA’S STEEL EXPORTS AND PROSPECTS

Although exporting steel products reflects the fact that China’s underly-

ing comparative advantage has been shifting towards the production of

capital- intensive goods at the stage of industrialization current at the time

of writing, there are certain concerns (including environmental concerns)

with respect to whether China should continue to follow this comparative

advantage in exporting steel products on a large scale to world markets.

These concerns reflect a key feature in the Chinese economy, namely that

there are distinct interests associated with steel firms, the industry and

the government, respectively, and those interests may not necessarily be

consistent with each other. These different interests will influence how the

trade orientation of the Chinese steel industry evolves over time.

First, China faces tremendous challenges associated with its economic

growth, both with respect to the use of resources and with regard to

reducing its intensity of carbon emissions. There were shortages of raw

materials and energy, and a great deal of environmental degradation

accompanying China’s rapid economic growth in the 30 years 1978–2008.

Steel production features heavy energy consumption, heavy pollution and

resource- intensive production, especially as the blast furnace/converter

process still dominates steel- making in China. Taking coal as an example,

according to British Petroleum (BP), China’s proven reserves of coal

were 114.5 billion tonnes at the end of 2008, which accounted for 13.9

per cent of the world’s total proven coal reserves. However, with a huge

amount of coal being produced annually, its R/P ratio (proven reserves/

output of the year) was only 41 years, comparing with the world average

of 122 years.13 With regard to the environment, China became the largest

carbon- dioxide- emitting country in 2007, with a continuous increase in

fossil fuel use. China’s steel industry created 8.3 per cent of the country’s

total industrial value added and provided 4.6 per cent of total industrial

employment, whereas it used 25.1 per cent of the total energy consump-

tion and generated 10–15 per cent of pollution in the total manufacturing

industries in 2007. Furthermore, the rapidly increasing exports of steel

products have led to the growth of industrial demand for iron ore and

other resources such as coking coal. This increasing demand in turn has

pushed up the iron ore prices in both international and domestic markets,

contributing to the continuous deterioration of China’s terms of trade.

Therefore, continuing to export steel products may be profitable for

steel- makers, but that will make it harder for the industry to reduce energy

and resource intensity and comply with the environmental regulations,

and for China to avert the deterioration of its terms of trade.



Second, the increasing level of exports of steel products has reduced

the market pressure from excessive steel production capacity, resulting

in the failure of the Chinese steel industry’s efforts to control the rapid

expansion of its capacity. For example, had China not had net exports

of 40–50 million tonnes of steel annually over the period 2007–09, the

problem of excessive capacity in the industry could have been much

greater as those quantities of steel products would have had to be

absorbed in the domestic market. This could lower the domestic prices

for steel and squeeze the profit margins for the steel- makers. Considering

these adverse effects of exporting steel products, the Chinese govern-

ment has adopted measures to control steel exports, especially of low-

value- added products. According to Decree 35: China’s Steel Industry

Development Policy, issued by the National Development and Reform

Commission (NDRC) in July 2005, the basic role of the Chinese steel

industry was defined to ‘meet domestic demand’. From 2004, as men-

tioned earlier, the Chinese government adjusted tax regulations on steel

exports, reducing or abolishing the export rebate of VAT and collecting

tariffs on those exports.

The implementation of these regulations has made exporting costs

higher than the spot prices in the domestic market. The increased costs

vary for different steel products, but generally the costs are lower for

high- grade steel products and vice versa,14 depending on both suitable

rates of rebate and tariffs imposed on exports. For example, suppose 1

tonne of steel products is sold at the price of 5000 yuan excluding VAT or

5850 yuan including VAT15 in the Chinese domestic market. It could be

exported at the FOB (free on board) price of 5000 yuan if the VAT rebate

rate equals the collecting rate of VAT (complete rebate) and the exporter

would get similar profits as in domestic markets. If the VAT rebate is

different from the collecting rate of VAT by 10 per cent for example, the

exporter has to add 7 per cent at least, that is 350 yuan (5000 × 0.07) to the

original export price excluding VAT to get similar profits for the complete

rebate example. Furthermore, the price of FOB 5850 yuan (5000 1 5000

× 0.17) is equally beneficial to the exporter for steel products without any

rebate of VAT.

Let A 5 the price excluding VAT in the domestic market or export cost

under supposed export regulation of the complete rebate;

V 5 the suitable collecting rate of VAT;

Vr 5 the suitable rebate rate of VAT;

P 5 the actual export cost or actual export price (FOB), which

allows the exporter to get equal benefits as selling in domestic

markets.

Then, if there is a difference between V and Vr,



P 5 A 1 P × (V − Vr),

and thus A 5 P − P × (V − Vr). (7.1)

Now let us add in the export tariff. As we know, the export tariff is

added only to the steel products without VAT rebate. Then let C be the

rate of suitable export tariff; P is the actual export cost for steel products

with zero VAT rebate, and C allows the exporter to receive equal benefits

to selling in domestic markets.

Then, P 5 A 1 P ÷ (1 1 V) × V 1 P × C,

P 5 A ÷ [1 − V ÷ (1 1 V) − C],

A 5 P × [1 − V ÷ (1 1 V) − C], (7.2)

D 5 an increase of export costs in amount 5 P − A,

E 5 an increase of export costs in percentage 5 P ÷ A − 1.

Based on the methods above, we quote the actual data of China’s export

costs or prices in 2010 to see how the different export policies affect the

export costs for different steel products (Table 7.3).

The results show that the policies adopted by the government have

been quite effective, be they in the form of either the VAT rebate or the

tariff imposed on steel exports as a measure to control exports of steel and

other resource products. Exports of plate, sheet, tube and pipe – so- called

‘high- grade steel products’ – have increased in both relative and absolute

terms, while the proportion of ‘low- grade products’ such as bar and wire

has decreased by half, and exports of semifinished steel have decreased to

almost zero in the five years to 2010. Certainly, the tendency to massive

Table 7.3 Effects of export regulations on exporting costs for different

steel products

Product P (FOB

in US$)

V

(%)

Vr

(%)

C

(%)

A

(US$)

D

(US$)

E

(%)

Slab and billet

(semifinished steel) 572 17 0 25 346 226 65.4

Reinforced bar 740 17 0 15 521 218 41.9

Hot rolled wire rod 655 17 0 15 462 193 41.9

Hot rolled coil 696 17 0 0 577 118 20.5

Cold rolled coil 691 17 9 0 636 55 8.7

Steel tube for oil- drilling 1180 17 13 0 1132 47 4.2

Source: Calculated using original trade data from China Customs.



steel exports has, on the whole, been arrested by the measures adopted by

the government (Table 7.4).

CONCLUSIONS

The complication in discussing the trade orientation of China’s steel

industry lies in the fact that it is difficult to judge how much of the shift

to exports since 2005 is a phenomenon resulting from disequilibrium that

will be corrected over time. However, one thing is clear – China will not

end its imports of steel anytime soon, as it will continue to rely on import-

ing high- quality and high- value- added steel products which the industry

cannot produce while exporting low- to- medium- quality products to world

markets. The large increase in net exports of steel since 2006 was associ-

ated with many structural problems as identified in this chapter.

Looking to the fut u r e , the Chinese government may have no inten-

tion of forsaking its policy of controlling excessive steel exports in

consideration of the need for industrial restructuring, the reduction of

resource intensity use by the industry, and the impact on the environ-

ment. However, a big challenge in implementing this strategy is how the

Chinese government could bring steel production back into line with

domestic demand without relying too much on exports. The application

of VAT and the imposition of export tariffs do not provide a permanent

solution, because both policies go against the principle of free trade. One

alternative is to leave the market to decide how much steel output the

industry needs to produce and export. In so doing, it may benefit the

Table 7.4 Changing pattern and structure of steel export, 2004 to 2010

(million tonnes and per cent)

Product 2004 2006 2008 2010

Mt % Mt % Mt % Mt %

Bar and wire 4.47 21.9 11.07 21.3 12.62 20.9 5.19 12.2

Sections 0.50 2.5 2.68 5.2 3.56 5.9 1.93 4.5

Plate and sheet 5.78 28.4 20.37 39.1 28.79 47.6 24.80 58.2

Tube and pipe 2.08 10.2 6.41 12.3 10.64 17.6 7.31 17.2

Steel for railway 0.09 0.5 0.24 0.5 0.55 0.9 0.48 1.1

Other finished steel 1.30 6.4 2.24 4.3 3.02 5.0 2.75 6.5

Semi- finished steel 6.15 30.2 9.08 17.4 1.32 2.2 0.14 0.3

Total 20.38 100.0 52.08 100.0 60.50 100.0 42.62 100.0

Source: Calculated based on data from China Customs.



steel- makers in terms of obtaining short- term profits, but it may worsen

the long- term structural problems for the industry, including the impact

on the environment.

There are a number of structural factors operating which could for

the industry eventually align the balance between demand and supply of

steel products without excessive reliance on exports. They include con-

tinual upgrading of production capacity and the retirement of outdated

steel mills; reduction of the current fragmentation of the industry as the

industry continues to raise its concentration; increasing energy efficiency

through improved management and technological change; real exchange

rate appreciation characterized by both a stronger yuan and a rising price

level; higher wages consistent with the previous point; and more effective

enforcement of state environmental regulations.

NOTES

1. China both exports and imports finished steel as well as semifinished steel billet and slab, that is crude steel, which has to be rolled into finished steel before end use, and there is wastage of materials in the rolling process. For comparison across different years, we customarily convert the finished steel of imports and exports into crude steel by the average yield rates of rolling for every year. For example, the rate for 2010 is about 94 per cent, so, 1 tonne of finished steel is equivalent to 1.064 (1/0.94) tonnes of crude steel in that year.

2. In both years, China’s imports of steel peaked, exceeding 40 million tonnes. This represented an increase of 409.3 per cent and 47.6 per cent from the previous years, respectively.

3. According to World Steel in Figures (2008, 2009 and 2010 editions) published by the World Steel Association (WSA), Japan exported 34.6, 35.5 and 36.9 million tonnes compared with China’s exports of 54.6, 72.8 and 64.1 million tonnes in 2006, 2007 and 2008, respectively.

4. Another factor which determines the competitiveness of an industry on international markets is the quality of products.

5. WSA (2010). 6. CISA is a national steel industry organization. The members mainly consist of steel

production enterprises, which account for 80 per cent of national steel output. Some trading firms, equipment manufacturers and construction firms as well as consulting companies are also members of CISA.

7. This is an energy unit which is used customarily in China. 1 kgce 5 7000 kcal (kilocalories) 5 29.27 MJ (megajoules). To convert from coal equivalent to oil equivalent, the amount must be multiplied by 0.7.

8. According to WSA (2010), the indicator of average energy intensity was 18 GJ/tonne (GJ 5 gigajoule) steel (equal to about 615 kgce/tonne steel) for participants submitting their questionnaires for both 2007 and 2008.

9. However, China has become a net importer of coking coal since 2009. Therefore, the imposition of export tariffs on coke by the Chinese government may not significantly affect the steel firms in other countries, as coke users in other countries may get ‘cheap’ coke from other sources just as Chinese steel- makers do.

10. Dieter Ameling, president of the German Steel Federation and Hans Jurgen Kerkhoff, General Manager of the German Steel Federation: ‘Dynamic Steel Market Faces New



Challenges’, for the Press Conference on 5 November 2007 on the occasion of stahl 2007 in Dusseldorf.

11. See Wiley Rein (2007).12. See China Research and Consulting (2009). This analysis covered production costs and

freight.13. See BP Group (2009).14. According to the specifications by the steel industry in China, sheet, plate, tube or pipe

are so- called ‘high- grade steel’ while long steel especially for construction such as rein-forced bar and wire rod are considered ‘low- grade steel’.

15. Note that 17 per cent is the common rate of VAT suitable for most goods including steel products in China. Applying this rate gives 5850 5 5000 × 1.17.

REFERENCES

BP Group (2009), Statistical Review of World Energy 2009, June, London: BP Group.

China Research and Consulting (2009), The State- Business Nexus in China’s Steel Industry – Chinese Market Distortions in Domestic and International Perspective, January, Munich, Germany: China Research and Consulting, accessed at http://www.eurofer.org/index.php/eng/content/view/full/911.

China Iron and Steel Association (CISA) (2009 and various years), Chinese Steel Industry Yearbook, Beijing.

China Customs (2010), Trade Statistics, Beijing: CISA.Huang, Y. and B. Wang (2010), ‘Cost distortions and structural imbalances in

China’, China and World Economy, 18 (4), 1–17.Liu, H. (2008), ‘Study on the cost competitiveness of China steel industry in

disagreement with Wiley Rein’s report’, Metallurgical Industry Management (China), 1 (January), 37–40.

Li, K. and L. Song (2011), ‘Technological content of China’s exports and need for quality upgrading’, in J. Golley and L. Song (eds), Rising China: Global Challenges and Opportunities, Canberra: Australian National University Press, pp. 69–84.

National Development and Reform Commission (NDRC) (2005) Decree 35: Steel Industry Development Policy, July, Beijing: NDRC.

Wiley Rein LLP (2007), ‘Money for metal – a detailed examination of Chinese government subsidies to its steel industry’, report sponsored by the American Steel and Iron Institute, July, accessed at htt p : / / w w w . w i l e y r e i n . c o m / r e s o u r c e s / d o c u m e n t s / p u 4 4 1 1 . p d f .

W orld Steel Association (WSA) (various years), World Steel in Figures, Brussels: WSA.


145

8. China’s iron ore import demand and its determinants: a time- series analysis


INTRODUCTION

China’s rapid economic growth and openness to trade dramatically

increased its production, consumption and export (directly or indirectly)

of iron and steel products during the two decades between 1986 and 2006.

The expansion of China’s iron and steel industry has lifted its demand for

iron ore – one of the major inputs in the production of iron and steel – and

driven up world prices of iron ore in recent years. During the period 2000–

06, China’s consumption of iron ore increased from 292 to 908 million

tonnes with an average annual growth rate of 20.8 per cent. In 2006, the

total consumption of iron ore in China accounted for 57.6 per cent of total

world production. China has become the largest consumer of iron ore in

the world and a key driver underlying the resource boom since the mid

2000s. However, as pointed out by Garnaut (2012), the increase in import

share of iron ore supply by China over the past 20 years is likely to be a

unique occurrence – a source of growth in imports that will not be there to

any significant degree in the future.

The dramatic increase in demand for iron ore in China has not only

encouraged the domestic suppliers to boost their production, but also caused

the rapid growth in Chinese imports from the international market. During

the period 2000–06, China’s total imports of iron ore increased from 70 to

326 million tonnes, giving an average annual growth rate of 29.3 per cent –

8.5 percentage points higher than the annual growth rate of total demand

and 11.9 percentage points higher than that of the demand from domestic

production. China’s dependency on imported iron ore (or the ratio between

imports and domestic production) reached 56.1 per cent in 2006. Given the

relatively constant world supply of iron ore, China’s high dependency ratio

implies that its increasing demand for iron ore from the international market

has exerted upward pressure on the world market price (Figure 8.5).



China’s increasing demand for iron ore raises two important issues.

First, what are the main driving forces and long- term trends behind

China’s increasing demand for iron ore from the international market?

Second, will world production of iron ore meet the future demand of

China’s expanding steel industry, and how will the world price of iron ore

change over time? Answering these questions is the purpose of this chapter.

This chapter applies time- series analysis to the industry- level data for

the period 1960–2005 to examine the trend of China’s imports of iron ore

from the international market and its determinants from a demand per-

spective. The analysis provides a useful context within which some policy

issues could be addressed. For example, one issue is whether government

intervention in imports of iron ore from the international market, through

restricting exports of pig iron and steel products by domestic steel- makers,

and imports of iron ore by small and medium firms, is necessary or not for

increasing the efficiency of China’s iron and steel industry, improving envi-

ronmental protection and stabilizing the world market price of iron ore.

There are three key empirical results: (1) The major driving force behind

China’s rapidly increasing import demand for iron ore is the increas-

ing domestic consumption of iron and steel products, and the demand

is generally price- inelastic, especially in the long run. (2) There exists a

significant substitution relationship between domestically supplied and

imported iron ore, but the quality (or richness) of domestic iron ore is

negatively related to the quantity of imports, suggesting that China’s

iron ore imports have played an important role in compensating for the

inadequacy of the domestic supply of iron ore in terms of both quality and

quantity. (3) Exports of pig iron and steel products do not have a signifi-

cantly positive relationship with the imports of iron ore in China, implying

that restricting exports of steel products might not be conducive to reduc-

ing the pressure on imports of iron ore.

CHINA’S IRON ORE IMPORTS: 1960–2006

China has been importing iron ore from the international market since

1960, despite the fact that the Chinese economy was operating under

a planned system between 1960 and 1978. However, the quantities of

imports had been rather small until the mid 1980s when they started to

increase steadily. In the two decades 1985–2006, China’s import of iron

ore increased rapidly from 10.1 million tonnes to 326.3 million tonnes, and

its domestic supply of iron ore also increased substantially over the same

period (Table 8.1). Figure 8.1 shows the changes in the shares of China’s

iron ore production and trade in the world total. After more than ten years


The iron ore import demand 147

Table 8.1 Iron ore production and trade: China and the world compared,

1960–2006 (million tonnes)

Year World iron ore

production (Mt)

World iron ore

exports (Mt)

China’s iron ore

production (Mt)

China’s imports

of iron ore (Mt)

1960 522.0 151.9 112.8 0.6

1961 503.8 149.2 51.6 0.7

1962 523.4 157.9 25.8 0.7

1963 516.2 164.7 24.2 0.9

1964 568.9 198.7 26.7 0.8

1965 624.3 211.7 31.5 1.0

1966 625.9 214.9 39.3 0.9

1967 629.2 223.7 29.6 0.8

1968 687.7 256.7 26.8 0.6

1969 720.3 278.8 43.3 0.4

1970 773.9 323.1 64.2 0.7

1971 780.5 318.0 81.5 0.8

1972 781.7 311.4 84.6 0.9

1973 845.6 377.0 91.6 0.8

1974 896.3 411.8 86.8 2.9

1975 888.8 381.0 96.9 2.7

1976 922.9 379.8 89.7 2.4

1977 880.8 358.0 93.8 2.6

1978 891.3 350.6 117.8 8.0

1979 947.6 398.1 118.8 7.2

1980 917.9 384.6 112.6 7.3

1981 894.6 372.9 104.6 3.3

1982 818.2 329.3 107.3 3.5

1983 782.1 315.1 113.4 4.4

1984 882.2 372.4 126.7 6.0

1985 909.6 375.8 137.8 10.1

1986 920.7 370.0 149.5 12.0

1987 945.5 367.8 161.4 12.1

1988 964.4 400.9 167.7 10.8

1989 991.0 424.3 171.9 12.4

1990 980.6 397.1 179.3 14.2

1991 952.1 398.9 190.6 19.0

1992 907.8 334.0 209.8 25.2

1993 944.3 354.0 226.4 33.0

1994 967.8 383.0 250.7 37.3

1995 918.6 459.8 261.9 41.2

1996 899.5 455.1 252.3 43.9

1997 920.2 481.5 268.6 55.1

1998 899.4 461.9 246.9 51.8



of continuous increase in China’s iron ore imports, the ratio of China’s

iron ore production and imports over total world iron ore production and

exports reached 20.7 per cent and 46.1 per cent in 2006, respectively.

There are three key reasons for China’s rapidly increasing demand

for iron ore in recent decades up to 2006. First, the increase of dom-

estic consumption of iron and steel products, driven and accelerated

Table 8.1 (continued)

Year World iron ore

production (Mt)

World iron ore

exports (Mt)

China’s iron ore

production (Mt)

China’s imports

of iron ore (Mt)

1999 891.0 444.8 237.2 55.3

2000 953.3 505.1 222.6 70.0

2001 934.6 501.9 217.1 92.3

2002 988.9 533.5 232.6 111.5

2003 1079.9 590.5 262.7 148.1

2004 1190.4 646.0 311.3 208.1

2005 1312.9 718.9 420.5 275.3

2006 1577.0 728.3 582.0 326.3

Sources: CISA (2004); CISA (various years); International Iron and Steel Institute (various years).

0

10

20

30

40

50

60

70

80

1960

Per

cen

t

Share of China’s iron ore Import over world output

Share of China’s import over China’s output

Share of China’s iron ore import over world export

Share of China’s output over world output

1963

1966

1969

1972

1975

1978

1981

1984

1987

1990

1993

1996

1999

2002

2005

Sources: CISA (2004); CISA (various years); International Iron and Steel Institute (various years).

Figure 8.1 Chinese shares of iron ore production and trade in world totals,

1960–2006 (per cent)



by the massive construction of infrastructure resulting from the rapid

and unprecedented pace of urbanization and the change in industrial

structure towards capital- intensive industries, was the main driver for

China’s increasing demand for iron ore during the two decades to 2006.

It is agreed in the literature that there is a strong correlation between

economic growth (represented by changes in a country’s GDP per

capita) and iron and steel consumption in the process of industrialization

(Chenery et al., 1986), and China provides no exception. Figure 8.2 shows

the relationship between GDP per capita calculated at constant 2000

US dollar prices, and the apparent consumption of crude steel in China

during the period of 1960–2006. With rapid economic growth due to

continuous industrialization, urbanization, openness to trade and other

macroeconomic economic reforms, China’s GDP per capita increased

significantly in the 1990s and 2000s with an average annual growth rate

of 8.34 per cent (between 1997 and 2006), more by far than the world

average at 2.67 per cent. This drove up the total consumption of iron and

steel products. In 2006, China’s total apparent consumption of crude steel

was 387.93 million tonnes, accounting for 34.6 per cent of the total world

consumption of crude steel. The dramatic increase in China’s demand for

steel products boosted its demand for iron ore from both domestic and

international markets, of which imports have become a more and more

important source of supply.

00 500 1000 1500 2000

50

100

150

200

250

300

350

400

450

App

aren

t con

sum

ptio

n of

cru

dest

eel (

mm

t)


Source: International Iron and Steel Institute (various years).

Figure 8.2 Relationship between GDP per capita and apparent

consumption of crude steel in China, 1960–2006



Second, the insufficient and low- quality domestic supply of iron ore has

been unable to meet the requirements of China’s rapidly expanding iron

and steel industry. Although China’s own reserves of iron ore are plentiful

in quantity and it has been ranked as one of the largest iron ore producers

in the world, domestic supplies of iron ore in terms of both quantity and

quality have been insufficient for meeting the increased demand from the

production of iron and steel products, especially those of high quality.

Figure 8.3 shows world iron ore production by region in 2005. China’s

crude iron ore production reached 420.5 million tonnes in 2005, account-

ing for 26 per cent of the total world output and making it the largest

producer of iron ore in the world. However, there was still a large gap of

275.3 million tonnes (accounting for around 57 per cent of China’s total

consumption) between demand and supply, making it necessary to import

iron ore from overseas sources to meet the shortfall.

In terms of the quality of domestic supply of iron ore, Figure 8.4 shows

the explored reserves of iron ore by richness in China in 2003. From the

total reserves of 57.7 billion tonnes, 85.8 per cent was relatively poor

quality, with iron content of less than 40 per cent, and only 1.9 per cent

was relatively rich, with iron content of more than 48 per cent. The average

ferrous content of China’s iron ore reserves was less than 33 per cent. This

suggests that the quality of China’s domestic supply of iron ore was largely

insufficient for the production of high- quality steel, so that utilizing its

Canada2%

US4% Latin America

22%

Africa4%

OECD Europe2%

Eastern Europe1%

CIS12%

Middle East1%India and SA

9%

East Asia0%

Oceania17%

Japan0%

China26%

Note: SA refers to other South Asian countries.

Source: Calculated using data from International Iron and Steel Institute (various years).

Figure 8.3 World production of iron ore by region, 2005 (per cent)



own reserves requires iron ore to be further processed, involving higher

costs in production. As a consequence of the need to supplement dom-

estic iron ore supplies in both quantity and quality owing to the increasing

demand from production, it is not surprising that China’s imports of iron

ore from the international market have been rising over time.

Third, it is the relative price advantage of iron ore in the international

market over that in the domestic market which shifts China’s demand for

iron ore from its domestic market to the international market. From the

production perspective, demand for iron ore is generally inelastic with

respect to its price for three reasons (Chang, 1994): (1) Iron ore input

accounts for only approximately 5 per cent of steel production costs (Tex

Report, 1988). Therefore, the costs of steel production are likely to be

largely unaffected by small increases in the price of iron ore. (2) There is

no substitute for iron ore in the production of steel in integrated steel mills,

and as such, steel producers face little room for adjustment to the product

mix. As a result, producers are unlikely to significantly alter quantities of

iron ore given a short- term change in its price. (3) Steel production plants

are, in general, highly specific and capital- intensive operations; since maxi-

mizing the utilization of capital can help achieve significant economies

of scale, normal operation aims to sustain high capital utilization. Given

that decreasing capital utilization due to an increase in iron ore prices

would significantly affect unit steel costs of output, demand for iron ore

may not alter significantly following a change in price. Given the inelastic

demand for iron ore, the relative prices on the domestic and international

25%25%–40%40%–48%48%Others

5%12%

2%1%

80%

Source: Cao et al. (2007).

Figure 8.4 China’s explored reserves of iron ore by richness, 2003

(per cent)



markets have thus played an important role in determining China’s iron

ore imports.

Figure 8.5 shows the nominal and the real price of iron ore on the inter-

national market during the period 1960–2007. Although the nominal price

continued to increase, the real price has been falling since the mid 1970s

owing to the depreciation of the US dollar and the reduction of interna-

tional transportation fees (except for the five years to 2007). Combined

with the increasing costs of domestic supply, this has encouraged China’s

enterprises to shift their demand for iron ore to the international market

since they can reap obvious savings in the cost of steel production by

importing the high- quality ore.1

We have now summarized the three important reasons for China’s rapid

increase in iron ore imports: domestic demand, relative price and domestic

supply. Next we examine how those factors affect China’s iron ore imports

by estimating an import demand function with the aggregate time- series

data.

THE TIME- SERIES MODEL OF INTERNATIONAL DEMAND AND DATA

The estimation of an import demand function is considered to be sufficient

to analyse China’s international demand for iron ore. Following Chang

(1994) and Tcha and Wright (1999), we specify the log–linear empiri-

0

10

20

30

40

50

60

70

80

90

100

2006

Pric

e of

iron

ore

(U

S$/

t)Real price Nominal price

1960

1962

1964

1966

1968

1970

1972

1974

1976

1978

1980

1982

1984

1986

1988

1990

1992

1994

1996

1998

2000

2002

2004

Source: ABARES (2006).

Figure 8.5 Nominal and real prices of iron ore on the international

market, 1960–2007 (US$/tonne)



cal model for examining the roles of various factors, such as domestic

demand, relative price and domestic supply, in affecting China’s iron ore

imports, as shown in Equation (8.1):

ln IMt 5 b0 1 b1 ln DACt 1 b2 ln RPt 1 b3 ln DSt 1 b4 ln IDRt

1 b5 ln EXPIt 1 b6 ln EXSTt 1 gD90 1 et (8.1)

where IMt is China’s iron ore import quantity at time period t; DACt is

the domestic apparent consumption of crude steel at time period t; RPt

is the real price of iron ore in the international market in constant 2000

US dollars; DSt is China’s domestic supply of crude iron ore; and IDRt

is the ferrous content of domestic iron ore supply at time period t. D90 is

a dummy variable equal to 1 if t $ 1990 and 0 otherwise, which is used

to capture the different trend of China’s iron ore imports after 1990.

Differing from previous studies, we also put China’s exports of pig iron

(EXPIt) and steel products (EXSTt) into the regression to identify the

impact of China’s exports of iron and steel products on its demand for

iron ore from the international market. Finally, et denotes the residual and

all variables are given as natural logarithms; bs are coefficients. Equation

(8.1) defines China’s imports of iron ore as a function of its steel consump-

tion, the real price in the international market and the domestic supply (as

a substitute). The basic logic behind the equation comes from the demand

function, which emphasizes that demand for iron ore from the interna-

tional market in China is determined by its consumption, price, domestic

substitute and exports of iron and steel products.

The data used for our estimation are the aggregate time- series data for

China during the period 1960–2005. The variables, such as China’s total

import of crude iron ore from the international market (IMt); the domestic

apparent consumption of crude steel (DACt); China’s domestic supply of

crude iron ore (DSt); and China’s exports of pig iron (EXPIt) and steel

products (EXSTt) are defined as the same as those in China Iron and Steel

Industrial Data Compression for 50 Years (CISA, 2004). The data for those

variables before 2000 are taken from China Iron and Steel Industrial Data

Compression for 50 Years and those after 2000 come from China Iron and

Steel Statistical Yearbook (CISA, various issues). IDRt is defined as the

ratio of the amount of iron extracted from domestic iron ore over the total

amount of crude iron ore produced domestically. The real prices of iron

ore in the international market (RPt) are defined as the spot market prices

of iron ore obtained from the Australian Commodity Statistics (ABARES,

2006), which are deflated using the US consumer price index (CPI) taken

from the WDI online database.



Since there might exist autocorrelation in the residual (et), the estima-

tion of Equation (8.1) with the ordinary least squares (OLS) method

may suffer from the time- series problem. Thus, the Dickey–Fuller (DF)

test for unit root of main variables and for the residual of some linear

combination of those dependent variables should be made to identify

the integration and cointegration relationship among those variables.

Table 8.2 shows the DF test results for each variable, and it suggests

that all variables are the first- order integrated. This can also be shown

in Figure 8.6, since all the first- order differences of those variables are

stable.

Table 8.3 shows the results of DF tests for the cointegration relation-

ship between the dependent and independent variables for both the first-

difference and the error correction models. The results show that the null

hypothesis that there exists no cointegration between those variables is

rejected for both cases at the 1 per cent significance level. This suggests

that both the first difference and error correction models can be used to

estimate the relationship between China’s imports of iron ore and their

determinants over time.

Thus, the two empirical models can be specified as below:

Table 8.2 Dickey- Fuller test for unit root of main variables

Variable Item No. of

observations

t- test

statistic

1% critical

value

MacKinnon

approximate p- value

ln IM 45 0.39 −3.61 0.98

d ln IM 44 −6.71 −3.62 0.00

ln DAC 45 0.81 −3.61 0.99

d ln DAC 44 −7.01 −3.62 0.00

ln RP 45 −1.83 −3.61 0.36

d ln RP 44 −4.03 −3.62 0.00

ln DS 45 −0.12 −3.61 0.95

d ln DS 44 −4.97 −3.62 0.00

ln IDR 45 −1.40 −3.61 0.58

d ln IDR 44 −7.27 −3.62 0.00

ln EXST 45 −0.51 −3.61 0.89

d ln EXST 44 −5.02 −3.62 0.00

ln EXPI 45 −1.09 −3.63 0.72

d ln EXPI 44 −5.66 −3.65 0.00




First- difference Model

D ln IMt 5 g0 1 g1D ln DACt 1 g2D ln RPt 1 g3D ln DSt 1 g4D ln IDRt

1 g5D ln EXPIt 1 g6D ln EXSTt 1 gD90 1 ut. (8.2)

Error Correction Model

ln IMt 5 g0 1 g1 ln IMt21 1 g2 ln DACt 1 g3 ln DACt21 1 g4 ln RPt

1 g5 ln RPt21 1 g6 ln DSt 1 g7 ln DSt21 1 g8 ln IDRt 1 g9 ln IDRt21

1 g10 ln EXPIt 1 g11 ln EXPIt21 1 g12 ln EXSTt 1 g13 ln EXSTt21

1 gD90 1 wt. (8.3)

Finally, all estimations are carried out using STATA 8.0 based on

Equations (8.2) and (8.3) and the aggregate time- series data. The results

show that the F- statistic tests in the first- difference and the error correc-

tion models are 2.33 and 166.41, respectively, both of which are statisti-

cally significant at the 5 per cent level (Table 8.4). This implies that both

model specifications provide good fit. Meanwhile, the results are also free

–4

–3

–2

–1

0

1

2

3

1960d

ln

d ln IM d ln DAC d ln RP d ln DS d ln IDR d ln EXST d ln EXPI

1963

1966

1969

1972

1975

1978

1981

1984

1987

1990

1993

1996

1999

2002

2005

Note: d for first differencing.


Figure 8.6 Changes in the logarithm of main variables in China,

1961–2005


156

Table

8.3

D

ick

ey–F

ull

er t

est

for

coin

tegra

tion a

mong m

ain

vari

able

s

Tes

t st

ati

stic

s1%

cri

tica

l valu

e5%

cri

tica

l valu

e10%

cri

tica

l valu

e

1st

- dif

fere

nce

: e(

t)−

6.8

1−

3.6

2−

2.9

5−

2.6

1

EC

M:

e(t)

−7.9

2−

3.6

5−

2.9

6−

2.6

1

MacK

inn

on

ap

pro

x.

p- v

alu

e 5

0.0

0

Th

e n

ull

hyp

oth

esis

of

no

co

inte

gra

tio

n i

s re

ject

ed a

t th

e 1%

lev

el.

Note

: E

CM

5 e

rro

r co

rrec

tio

n m

od

el.

Sourc

e:

Au

tho

rs’

ow

n c

alc

ula

tio

n.



from the serial correlation problem, as shown by various diagnostic tests

which were carried out.

CHINA’S IRON ORE IMPORTS AND ITS DETERMINANTS

How are China’s imports of iron ore from the international market deter-

mined? The estimated results from both the first- differencing and the error

Table 8.4 Determinants of Chinese iron ore imports, 1960–2005

1st- difference model Error correction model

Coefficients p- value Coefficients p- value

Dependent variable d ln IM ln IM

Number of observations 42 42

Constant 0.53 (0.44) −6.11*** (0.00)

D1990 – – 0.85*** (0.01)

ln IM (−1) – – 0.37* (0.06)

ln DAC – – 3.26*** (0.00)

ln DAC (−1) – – −1.40** (0.02)

d ln DAC 1.39*** (0.01) – –

ln RP – – 1.23*** (0.01)

ln RP (−1) – – −0.54 (0.21)

d ln RP 0.47 (0.21) – –

ln DS – – −2.68*** (0.00)

ln DS (−1) – – 1.79*** (0.00)

d ln DS −1.17** (0.02) – –

ln IDR – – −3.11*** (0.00)

ln IDR (−1) – – 1.50 (0.15)

d ln IDR −2.78*** (0.00) – –

ln EXST – – −0.16 (0.41)

ln EXST (−1) – – −0.23 (0.15)

d ln EXST 0.03 (0.87) – –

ln EXPI – – 0.15* (0.04)

ln EXPI (−1) – – −0.04 (0.57)

d ln EXPI 0.03 (0.70) – –

F- statistic test 2.33 (0.05) 166.41 (0.00)

Note: *, ** and *** represent the estimated coefficient statistically significant at the 10 per cent, 5 per cent and 1 per cent level respectively. ‘D1990’ refers to the dummy for the specific year of 1990.




correction models show that four key factors have played different roles in

determining the outcomes.

First, the domestic consumption of iron and steel products is the most

important determinant of China’s imports of iron ore from the interna-

tional market. As expected, Table 8.4 shows that the coefficients of the

domestic consumption of crude steel in both models are positive and

statistically significant at the 1 per cent level, implying that the domestic

consumption of crude steel is positively related to the imports of iron ore

in China. This implies that domestic consumption is a major driving force

in China’s iron ore imports, as other factors, such as price and domestic

substitution are well controlled. Moreover, the short- and long- run elas-

ticities of China’s imports of iron ore from the international market with

respect to the changes in domestic consumption of crude steel can be pro-

jected from the estimated coefficients of the error correction model, which

are 3.26 and 4.33, respectively. This suggests that a 1 per cent increase in

domestic consumption of crude steel may result in a 3.26 per cent increase

in China’s iron ore imports in the short run and a 4.33 per cent increase in

China’s iron ore imports in the long run. The difference in demand elastici-

ties of China’s iron ore imports between the short and long run suggests

that the impacts of domestic consumption of iron and steel products on

imports of iron ore are much larger in the long run. This is consistent with

the increasing trend of China’s demand for iron ore from the international

market as shown in Table 8.1.

Second, China’s imports of iron ore from the international market are

price- inelastic. Table 8.4 shows that the coefficient of the real price of iron

ore in the international market is positive, but statistically insignificant at

the 10 per cent level in the first- difference model, while that of the lagged

real price of iron ore in the international market is negative and statisti-

cally insignificant at the 10 per cent level in the error correction model.

These results may suggest that China’s iron ore imports have generally

been independent of the real price of iron ore in the international market

during the four decades up to 2007. The lack of price elasticity can also

be verified by the co- movement of China’s imports of iron ore from the

international market and the change in the real price of iron ore in the later

years of this period. A possible explanation is that the increase in China’s

demand for iron ore was so large that it also raised the domestic prices

of iron ore supply, reducing the price difference between supplies from

the domestic and international markets. Such an effect would weaken the

ability of China’s iron and steel enterprises to switch their sources of iron

ore inputs from the international market to the domestic one in response

to the surge of iron ore prices worldwide.2

Third, the domestic supply of iron ore is a substitute for importing iron



ore from the international market, but the substitutability is seriously

restricted by the poor ferrous content in domestically supplied iron ore.

As shown in Table 8.4, the coefficient of domestic iron ore production

in the first- difference model is negative and statistically significant at the

5 per cent level while that in the error correction model is negative and

significant at the 1 per cent level. This implies that the domestic supply of

iron ore in China can play an important role in substituting for imports.

However, the substitution elasticity of domestic supply retrieved from the

error correction model in the short run is more than that in the long run

(−2.68 and −1.41, respectively, as estimated from Table 8.4). This implies

that a 1 per cent increase in domestic supply may substitute 2.68 per cent

of imports in the short run compared to only 1.41 per cent of imports in

the long run. This finding also suggests that the substitutability of domes-

tic supply of iron ore for imports is weaker in the long run. A possible

explanation is that there is a significant quality difference between domes-

tic and imported iron ore. The difference in quality makes China’s iron

and steel enterprises prefer to import iron ore in the long run, all other

things being equal. This interpretation can be supported by the evidence

from the negative and significant elasticities of China’s iron ore imports to

the ferrous contents of China’s crude iron ore in both the short and long

run (−3.11 per cent and −2.56 per cent, respectively, as estimated from

Table 8.4).3

Fourth, although exports of pig iron seem to have a positive impact

on China’s iron ore imports, exports of steel products do not. Table 8.4

shows that the coefficients of exports of both pig iron and steel products

in the first- difference model are insignificant, while only the coefficient of

exports of pig iron (one out of four coefficients) in the error correction

model is positive and statistically significant at the 5 per cent level. This

may imply that no significant positive relationship exists between exports

of iron and steel products and China’s iron ore imports from an empiri-

cal perspective.4 A policy implication is that restricting exports of iron

and steel products might not be an efficient way for China to reduce its

iron and steel enterprises’ dependence on imports of iron ore, particularly

when the efficiency effects of exports of these products by Chinese firms

and the structural linkage with the steel mills in China are considered (see

Chapter 9).

CONCLUSIONS

The dramatic increase in China’s imports of iron ore since the late 1990s

has exerted considerable pressure on world supplies of the ore, resulting



in a rapid increase in prices on the international market. What are the

determinants behind such an increase in China’s demand for iron ore, and

how will this trend change in the future? These are questions which have

important implications for both users and suppliers of this key commod-

ity. To answer these questions, this chapter applies the time- series analysis

to examine China’s imports of iron ore and some of its determinants

from a demand perspective, using the industry- level data over the period

1960–2005. The results show that the rapid increase in China’s imports of

iron ore from the international market came mainly from China’s domes-

tic consumption of iron and steel products, and this trend has tended

to continue in the long run because China’s per capita consumption of

steel products is still relatively low and the country continues to be in the

middle phase of rapid industrialization. Moreover, since there are insuf-

ficient domestic supplies of iron ore, as well as significant quality differ-

ences between imported and domestic iron ore, the substitution between

the domestically supplied and imported iron ore is limited, particularly in

the long run since China’s imports of iron ore are largely price- inelastic.

This price inelasticity partly explains why iron ore prices on international

markets continued to rise strongly after 2005. It also suggests that a further

increase in the iron ore price on the international market is likely until the

large gap between supply and demand has been eased through either the

increase in supply as we observed has happened to the commodity market

in the second half of 2012, or the softening of demand. Finally, the empiri-

cal results do not reveal a significant positive link between China’s exports

of pig iron and steel products and its imports of iron ore. Therefore, it may

not be ideal in terms of efficiency for China to try to restrict the exports

of iron and steel products in order to ease the pressure of China’s iron ore

imports on the international market.

NOTES

1. Labson et al. (1995) showed that after processing China’s low- quality ore and imposing production taxes, the unit price of China’s iron ore inflated to approximately US$35 per wet tonne, which is much higher than the world trade price of US$25 per wet tonne.

2. The negotiations between the Chinese steel mills and the world suppliers of iron ore on long- term contracts since 2007 illustrate the point.

3. This finding seems to provide some assurance to the world suppliers of iron ore in making the long- term investment to meet the future demand for their products from China in the future.

4. This may be due to the fact that China had only begun to export iron and steel products since 2007 in response to the high prices of those products on international markets. It is also observed that the metal content of the exports from China have been on the rise, resulting from the shift in the export bundle from labour- intensive to capital- intensive products.



REFERENCES

Australian Bureau of Agriculture and Resource Economics and Sciences (ABARES) (2006), Australian Commodity Statistics, Canberra: ABARES.

Cao, B.Y., G.L. Wang and B. Jiang (2007), ‘China’s usage of mining resources: present, challenges and strategy’, accessed April 2009 at http:/cl.newmaker.com/art_21562.html.

Chang, H. (1994), ‘Estimating Japanese import shares of iron ore’, Resources Policy, 20 (2), 87–93.

Chenery, H., S. Robinson and M. Syrquin (1986), Industrialisation and Growth: A Comparative Study, New York: Oxford University Press for the World Bank.

China National Bureau of Statistics (2006), China Iron and Steel Yearbook, Beijing: China Statistical Press.

China Iron and Steel Association (CISA) (2004), China Iron and Steel Industrial Data Compression for 50 Years, Beijing: CISA.

CISA (various years), China Iron and Steel Statistical Yearbook, Beijing: China Iron and Steel Association.

Garnaut, R. (2012), ‘The Contemporary China resources boom’, Australian Journal of Agricultural and Resource Economics, 56 (2), 222–43.

International Iron and Steel Institute (various years), World Steel Yearbook, Brussels: World Steel Association.

Labson, S., P. Gooday and A. Manson (1995), ‘China’s emerging steel industry and its impact on the world iron ore and steel market’, ABARES research report no. 95- 4, Canberra.

Tcha, M. and D. Wright (1999), ‘Determinants of China’s import demand for Australia’s iron ore’, Resources Policy, 25 (3), 143–9.

TEX Report (1988), Iron Ore Manual, Tokyo: TEX Report.


162

9. Restructuring China’s steel industry and the implications for energy use and the environment

Guoqing Dai and Ligang Song

INTRODUCTION

The steel industry is a relatively large energy consumer and polluter in China.

For example, in 2003, the shares of the key pollutants from the steel industry

in China’s total industrial emissions were as follows: waste water accounted

for 8.4 per cent, sulphur dioxide (SO2) 3.9 per cent, smoke 5.8 per cent, indus-

trial dust 15.3 per cent and chemical oxygen demand (COD) in industrial

water pollution 17 per cent.1 Energy consumption constitutes a significant

portion of the overall costs of steel production. For example, in 2000, energy

consumption accounted for 35 per cent of total production costs of the steel

industry. This compares selected energy- intensive industries as follows: 40

per cent for petrochemical, 50 per cent for aluminum, 40–50 per cent for

construction materials and 70–75 per cent for fertilizers. Although the share

of energy consumption in the total cost of production looks relatively low

compared with these other industries, the steel industry’s level of efficiency

in utilizing energy remains far below the global technological frontier. For

example, in 2000, the steel industry’s energy consumption per unit of crude

steel produced was about 40 per cent higher than the international level

based on the best technology applied in those developed countries.2

Given the continuing importance of the steel industry in the process

of industrialization and development in China, the country faces the

challenge of how to reform the industry through pursuing structural

adjustment that prioritizes energy savings and pollution reduction, while

ensuring that production levels continue to meet the increasing demand

for steel. Were the steel industry to succeed on these fronts then, it would

represent a major contribution to the realization of the goals set by the

central government in achieving energy efficiency and emissions reduction.

In this chapter, we discuss the progress being made in energy saving and

pollution reduction in the Chinese steel industry; identify those underly-


Restructuring China’s steel industry 163

ing factors that drive improvements in energy and emissions efficiency;

and discuss how government could further implement those policies that

have proved to be successful in achieving the goals of improving energy

efficiency and emission reduction in the steel industry.

PROGRESS IN ENERGY SAVING AND POLLUTION REDUCTION IN THE STEEL INDUSTRY

Energy Saving and Conservation

The steel industry is a major energy user. In 2000, large steel mills con-

sumed 118 million tonnes of coal equivalent (Mtce), accounting for about

10 per cent of the national total energy consumption.3 As far as energy effi-

ciency is concerned, the steel industry faces the following major problems:

the proliferation of small firms with low degrees of industry concentration;

the use of backward technology and equipment, especially by those small

steel mills; and the concentration of production with low- value- added

products. Furthermore, the rapid growth of production capacity fuelled

by local governments, whose primary concerns are collection of taxation

and expansion of local employment, has not been beneficial from an envi-

ronmental perspective.

In dealing with these problems, the central government has since 2003

adopted various measures aimed at optimizing the industrial structure and

increasing industrial concentration, through the closure of a large number

of small operations; upgrading the technologies used in the existing firms;

and tightening up regulations regarding industrial pollution and emissions

control. Specific measures taken have included lifting the ratio of firms’

own capital to external finance to 40 per cent and then further to 60 per

cent for approving new projects; reducing tax rebates granted for exports

of certain steel products; directly controlling the scale of bank loans

flowing to the sector; and providing finance and technical support to firms

for emissions reduction.

The measures implemented by the government in addressing those

problems have produced some tangible outcomes. The industry achieved

substantial progress in energy saving, pollution reduction and water saving

in the few years after these policies were implemented. For example, the

unit energy consumption (energy intensity) dropped substantially. Taking

large and medium steel enterprises as a group, comprehensive energy con-

sumption per unit of output was reduced from 960 kgce/t of crude steel in

1999 to 645 kgce/t in 2006, a reduction of 33 per cent over eight years, as

illustrated in Figure 9.1.



The trend of falling energy consumption per unit of steel output can be

compared with the increasing trend of the steel production over the same

period, as shown in Figure 9.2. The 33 per cent fall in energy intensity has

been accompanied by an increase of 3.4 times in the total steel output by

those large and medium firms over this period.

The reduction in energy intensity detailed above is impressive. However,

it should be noted that during this period the Chinese steel industry

achieved significant technological upgrading and extended the value chain

of production to a higher level, so that using a volume measure of steel is

likely to underestimate the extent of decline in energy use per unit of value

added. From 1999 to 2003, the gross output value (in constant prices) of the

Chinese steel industry increased by 213 per cent, while in the same period

the tonnage of crude steel output increased by only 79 per cent.4 Based on

this value term, the energy consumption of per unit of gross output value

in 2003 was roughly 50 per cent below that of 1999. This evidence suggests

that energy saving achieved by the steel industry has been quite compre-

hensive in coverage and substantial in extent. The overall trend is reflected

in all major processes of steel- making in China (Table 9.1).

Except for the basic oxygen furnace converter (BOF), which saw a sub-

stantial increase of 26 per cent in energy consumption per tonnage of steel

production between 2000 and 2005, all other processes witnessed falls in

energy intensities, especially EAF and rolling, which saw falls of 24 and 25

per cent over this period, respectively.

500

600

700

800

900

1000kgce/t

960

1999

930

2000

870

2001

807

2002

770

2003

765

2004

741

2005

645

2006

Source: Chinese Steel Industry Development Research Institute, Beijing.

Figure 9.1 Comprehensive energy consumption of large and medium state-

owned steel mills (kgce per tonne)



Reduction of Pollution

As a result of the policy measures, the pollution emissions of the indus-

try also dropped substantially. Over the period 2000–06, SO2, industrial

smoke and dust, and COD emissions were reduced by 52 per cent, 70 per

cent, 68 per cent and 77 per cent, respectively, as shown in Table 9.2.

A similar trend can be observed with respect to the solidified waste

930870

807770

741

645

960

765

124 127151

182

222

356

423

283

0

200

400

600

800

1000

1200kg

ce/t

0

50

100

150

200

250

300

350

400

450

mill

ion

ton

1999 2000 2001 2002 2003 2004 2005 2006

Energy consumption per tonCrude steel production


Figure 9.2 Energy consumption versus crude steel production, 1999–2006

(kgce/t; Mt)

Table 9.1 Change of energy consumption in major processes for member

companies of CISA, 2000–05 (kgce/t)

Year Sintering Coking Iron- making BOF EAF Rolling

2000 68.9 160.2 466.1 28.9 265.6 117.9

2005 64.8 142.2 456.8 36.3 201.0 88.5

Change (level) −4.1 −17.9 −9.3 7.5 −64.6 −29.4

Change (%) −5.9% −11.2% −2.0% 25.8% −24.3% −25.0%

Note: BOF 5 basic oxygen furnace converter; EAF 5 electric arc furnace.




generated by the steel industry (Table 9.3). The fall in solid waste was sub-

stantial, at 126 per cent reduction over the period under review.

Freshwater Saving

Freshwater is a scarce resource and is becoming increasingly so in China.

As a major water consumer, the steel industry has strengthened its efforts

in freshwater conservation in responding to the increasing scarcity of the

resource. For example, freshwater consumption in the steel industry fell

substantially from 25.24 t/t in 2000 to 6.56 t/t in 2006. At the same time,

the water reuse (recycling) rate greatly increased, from 88 per cent to 95

per cent over the same period (Figure 9.3).

Although enormous progress has been made since 2000, the industry

remains a major contributor to aggregate energy consumption and pollu-

tion emissions in China because of the nature and relative importance of

the industry in the Chinese economy in the second half of the first decade

of the twenty- first century. In 2006, the industry consumed around 15 per

cent of the country’s total industrial energy consumption. Its emissions of

waste water accounted for 14 per cent of the total industrial waste water

Table 9.2 Reduction of pollutant emissions per tonne of crude steel output

SO2 Smoke Dust CO2

2000 5563 1696 5077 985

2006 2660 518 1618 228

Change (level) 2903 1178 3459 757

Change (%) −52.2 −69.5 −68.1 −76.9

Note: The measurement unit for SO2, smoke and dust is mg/m3, and for CO2 is mg/L.


Table 9.3 Reduction of solidified waste by large and medium enterprises,

2000–05

Unit 2000 2005 Change (%)

Solid waste kg/t 728.7 603.2 −125.5

Sludge kg/t 121.2 96.8 −24.4

Industrial waste and others kg/t 39.3 38.4 −0.9




and its solid waste represented 16 per cent of the industrial total (Wang,

2007a). Therefore, further advancements in energy conservation and

pollution reduction remain an important task for the steel industry.

INCENTIVES FOR CHANGE IN ENERGY CONSERVATION AND POLLUTION REDUCTION

The main incentive for making progress in energy saving and pollution

reduction is that all Chinese steel companies, including both large and

small firms, are facing intense competitive pressure. There are over 300

integrated steel companies which produce both hot metal and final steel

products in China. Among them, 66 companies produced over 1 million

tonnes of crude steel in 2007. The market share of the top five largest steel

companies of China in 2007 was as follows: Baosteel Group (29 million

tonnes, world ranking five), accounting for 5.8 per cent of total steel

production; Anben Group (Anshan and Benxi, 24 million tonnes, world

ranking seven), accounting for 4.8 per cent of the total; Shagang Group

(23 million tonnes, world ranking eight), accounting for 4.7 per cent of the

total output; Tangsteel Group (23 million tonnes, world ranking nine),

accounting for 4.7 per cent; and WISCO (Wuhan steel, 20 million tonnes,

world ranking 11), accounting for 4.1 per cent of total steel production.

25.24

18.81

15.5813.73

11.27

8.6

6.56

87.84

89.08

90.55 90.73

92.28

94.1594.8

m3 /

t

0

5

10

15

20

25

30

84

86

88

90

92

94

96

%

2000 2001 2002 2003 2004 2005 2006

Freshwater consumption per tonne

Industrial water reuse rate


Figure 9.3 Freshwater consumption per tonne and industrial water reuse

rate of large and medium enterprises, 2000–06



The top five steel companies collectively control just under one quarter

of the market, which is far below the levels of industrial concentration

seen in the steel industries of major producing markets such as Europe,

North America, Japan and Korea. In such a fiercely competitive market,

steel- makers need to reduce their production costs by all possible means,

and reducing energy consumption is a large part of this effort. With the

cost of energy progressively moving towards a fully market- based system,

the imperative to economize on energy inputs in the industry will remain

in place.

A second reason for the increase in environmental efficiency is related

to the ongoing reform of SOEs and the associated changes in microeco-

nomic circumstances. In the past, most large and medium steel companies

were SOEs, and therefore developments impacting on this segment of the

economy were also highly relevant for steel firms. The administration of

most large and medium steel companies has been transferred from the

central government to provincial governments and also to more local

governments under whose jurisdictions those firms are physically located.

In 2007 only four steel companies, Baosteel, Anshan Steel, WISCO and

Panzhihua Steel, were under the administration of SASAC (State- owned

Assets Supervision and Administration Commission of the State Council).

One of the Chinese government’s policy objectives with respect to its

industry restructuring has been to encourage SOEs, including steel com-

panies, to undergo reform in ownership structure and ‘go public’ (that

is, become listed on the stock market) whenever and wherever possible.

In 2006 there were about 30 steel companies listed on the Shanghai and

Shenzhen stock exchanges. Some local governments sold off part or all of

their steel equity holdings during the process of privatization which had

been ongoing on a large scale in China since the early 1990s (Garnaut et

al., 2006). This dramatic change in the ownership structure in the industry

plays an important part in impacting on corporate attitudes towards effi-

ciency and productivity.

Another important contributor to changes in the energy use of the

steel firms was generated by the emergence and growth of private steel

companies. The market share of private steel companies was traditionally

small because the steel industry is so capital- intensive that most private

companies, usually small in scale, were not able to raise sufficient funds

to participate in this kind of production. However, during the ten years to

2007, the production from private steel companies has grown extremely

quickly. The output of steel produced by private steel companies reached

more than 40 per cent of the total output in 2007. The largest private steel

company, Shagang Group, was the third- largest producer of crude steel

(23 million tonnes) in China in 2007. Another private company, Fosun



International (listed in Hong Kong under the code 0656, HK), controlled

about 20 million tonnes of steel production capacity by its own investment

and through acquiring majority stakes in existing steel companies. CITIC

Pacific (0267, HK) was the leading specialized (alloy) steel- maker in

China. The efficiency drive of the entire steel industry has been accelerated

by the dynamic participation of private companies such as these.

As for the large SOEs in the industry, the various levels of government to

whom they belong are not involved directly in the daily operations of firms’

management. Rather the focus is on the issues of large- scale investment

and the appointment of senior executives. These firms have enough opera-

tional autonomy and the profit incentives to implement energy- saving and

pollution- reduction measures as part of the industry’s restructuring.

A further incentive to achieve energy savings is that prices for coal and

freshwater rose strongly over the decade to 2007 as part of the govern-

ment’s efforts to rectify the distortions in factor markets. After the SOE

reforms, the motive for profits has become central to the concerns of man-

agement. The continual increases of the prices of some of the inputs such as

iron ore have forced the steel companies to economize on energy and water

consumption, although the degree to which this can be achieved is limited

by the pace of technical progress underway in the industry. For example,

the price of coking coal delivered to Shoudu (Capital) Steel Corporation

in 2007 was 122 per cent higher than the level in 2001. During this period,

the coking rate (kg per tonne crude steel) was reduced by 13 per cent.

Technically speaking, it is not easy to achieve such reductions in the coking

rate, but the price signal is a powerful driver for firms to do so, and enter-

prises did respond accordingly. The reduction of freshwater consumption

per tonne of steel production can be explained along similar lines.

Over time, technological upgrading and improvement in economies of

scale have both contributed positively to energy conservation. In meeting

the challenges of restructuring the industry, further measures have been

taken to promote energy saving and pollution reduction. First, efforts have

been made to reduce the iron- to- steel ratio and increase the continuous-

casting ratio and the rolling yield. In an integrated steel mill, iron- making

takes over two- thirds of total energy consumption. Therefore, reducing

the amount of hot metal used in steel- making is a significant avenue for

achieving energy savings. As a result, the iron- to- steel ratio decreased sig-

nificantly over the period 2000–06, as shown in Table 9.4.

Steel casting techniques have changed progressively from a traditional

reliance on mould casting to a predominance of continuous casting pro-

cesses which have reduced energy consumption dramatically and increased

the rolling yield. The industry has made great progress on this front, as

detailed elsewhere in this book and illustrated by Figure 9.4.



Economies of scale are positively associated with firms’ energy savings

(Table 9.5). On all the measures reported in Table 9.5 including energy

consumption, coking ratio, water use and emissions, large operations are

all superior to smaller ones. Considering the fact that many small steel

mills still operate in China, one can easily conclude that to increase the

scale of operation (equipment) by increasing the industry concentration

is an effective way of reaching the goals of energy savings and emission

reduction in the steel industry.

During the ten- year period to 2006, Chinese steel enterprises upgraded

their equipment to a larger scale. Consequently, the number of large- scale

operations in the industry increased dramatically over the period 1995–

2006, as shown in Table 9.6. Table 9.6 shows that in 2006, there were 50

blast furnaces (BF) (of a total of 275) with production capacities of 2000

Table 9.4 Pig- iron- to- crude- steel ratio, 2000–06

Total steel industry Key enterprises

2000 1.02 0.916

2006 0.96 0.90


86.8 87 88.4 89.8 91.3 92.5 94 94.2 94.9 95 95.6 95.65

0

10

20

30

40

50

60

70

80

90

100

%

Continuous casting ratio Rolling yield

46.5

1995

53.3

1996

60.7

1997

68.8

1998

77.4

1999

85.3

2000

88.2

2001

91.2

2002

93.5

200395

.92004

97.5

2005

98.5

7

2006


Figure 9.4 Increase of continuous casting ratio and rolling yield

(per cent), 1995–2006



m3 or above, with iron- making capacity of 100 million tonnes, account-

ing for 46 per cent of the total iron- making capacity in the industry. Of

the 224 converters, 110 are in the above 100- tonne bracket, representing

steel- making capacity of 154 million tonnes, accounting for 58 per cent

of total capacity. The increasing proportion of output coming from firms

using more advanced techniques helps reduce the average energy intensity

Table 9.5 Comparison of key performance indexes between small and

large steel mills

Index Unit BF , 300m3,

Converter,

EAF , 20

tonne

BF . 1000m3,

Converter , 120

tonne, EAF , 70

tonne

Difference

Energy consumption kgce/t 499 420 79

Coking rate kgce/t 542 340 202

PCI kg/t 125 180 −55

Electricity consumption kw/t 500 250 250

Smoke/dust emission kg/t 2 0.1 19 times

SO2 emission kg/t 5.42 1.23 3.4 times

Freshwater consumption m3/t 0.33 0.17 0.16

Note: BF stands for blast furnace; EAF stands for electric arc furnace; PCI stands for pulverized coal injection, which is used to reduce the coke consumption when producing hot metal; coking rate is an indicator which represents the rate between the coke consumed and hot metal produced per tonne.

Source: Wang, C. and Chi (2007).

Table 9.6 Changes in scale of steel production, 1995–2006

1995 2006

Number of

firms

Production

(10 000 t)

Number of

firms

Production

(10 000 t)

BF .3000 m3 3 874 12 3551

2000–2999 m3 11 1836 38 7009

1000–1999 m3 28 – 52 5632

Converter ≥300 t 3 648.6 3 908

100–299 t 14 1256.8 107 14 488

50–99 t 114 9969

EAF ≥100 t 4 376.5 17 1574

Source: China Steel Industry Development Research Institute, Beijing.



for the entire industry (Wang, C. and Chi, 2007). However, the industry

has remained highly fragmented by international standards. Hence, elimi-

nating small and medium- sized production is a priority if further energy

savings are to be achieved through increasing economies of scale and

upgrading the technologies used in the industry.

The industry is making progress in promoting secondary energy uti-

lization and other energy- saving technology. For example, the reuse of

energy, particularly secondary energy (residual heat and residual energy)

is a main energy- saving measure of many steel- makers. In recent years,

TRT, CDQ, BF/converter gas recovery (defined below) and its reuse have

been widely introduced by large and medium steel enterprises in China.

● Top gas pressure recovery turbine (TRT) – The total number of blast

furnaces BFs in 2007 in China was around 1200, of which 120 had

capacity of more than 1000 m3. There were about 210 sets of TRT in

operation, of which the coverage for the BFs larger than 1000m3 was

over 90 per cent. The total power generated by TRT was 2.1 billion

kilowatt hours (kW h) in 2007 (Wang, 2007c).

● Coke dry quenching (CDQ) – CDQ technology can reduce compre-

hensive energy consumption by about 15 kgce/tonne. By the end of

2006, China had set up 44 sets of CDQ with a capacity of 48 million

tonnes per year. At that time, coke ovens with the CDQ system were

used in only a small proportion of operations (Wang, 2007b). By the

end of 2008, China had nearly 80 sets of CDQ devices, with produc-

tion capacity of 70 million tonnes of steel per year.

● BF/converter gas recovery equipment – By the end of 2006, 77 per

cent of the key steel enterprises had installed BF gas recovery equip-

ment, and a total of 261 billion m3 of gas was recovered from using

that system that year; 64 per cent of key large steel enterprises had

installed converter gas recovery equipment, and 10 billion m3 of

gas was recovered; 68 per cent of key steel enterprises had installed

converter residual heat steam recovery equipment producing similar

results in terms of energy savings.

● BF/converter dry dust removal system – BF and converter dry dust

removal systems can not only save energy but also reduce dust emis-

sions. There are many Chinese steel- makers that have introduced

these two systems.

According to a report by CISA, in the period 1990–99, process optimiza-

tion, energy management enhancement, energy- saving equipment and/or

technology, and raw materials improvement contributed to energy saving

by 41, 25, 19 and 15 per cent, respectively.



The Chinese government complements the efforts of individual firms

by implementing policies that encourage energy saving and environmental

protection. On 8 July 2005, the NDRC issued Decree 35, ‘Steel Industry

Development Policy’. This document clearly indicates that (the steel

companies) ‘should save energy and decrease the energy consumption

level, improve the standard of environment protection and make full,

multiple and reasonable use of resources according the principle of sus-

tainable development and recycling economy’. They should ‘improve the

standard of making full, multiple and reasonable use of waste gas, liquid,

residue, etc. as large as possible, work hard to achieve “zero discharge”,

transform the original steel mills into recycling ones’. The NDRC also

promulgated the concrete average consumption standards of energy and

freshwater for the whole steel industry. It demanded the phasing out of

obsolete equipment and listed the standards that enterprises need to meet

in installing new equipment.

Separately, the State Environmental Protection Administration (SEPA)

issued ‘Cleaner Production Standard – Iron and Steel Industry’ (HJ/

T189- 2006) in 2006 which included the following three measures. First,

the export of energy- intensive or high- pollution products was discouraged

by cancelling export tax rebates and even imposing an additional export

tax for some steel products. Second, the State Council, NDRC, NBS and

SEPA, now the Ministry of Environmental Protection (MEP), jointly

issued a statistical index system of energy consumption per unit of GDP in

November 2007, to force local governments and enterprises to incorporate

energy consumption into the overall evaluation of economic and social

development and annual performance evaluations. The amended Act of

Energy Conservation came into effect on 1 April 2009. Third, to reduce

the pollution in Beijing, the central government had decided several years

previously to have Shoudu (Capital) Steel Corporation (with 8 million

tonnes of crude steel production) stop production in metropolitan Beijing

and relocate to Caofeidian, a coastal site near Tangshan city in Hebei

province. Some other large steel companies also plan to move out of big

cities and are choosing less populous sites for rebuilding their factories.

Furthermore, there is a huge gap between China’s most advanced steel

enterprises and the industry laggards regarding their performance in

energy saving and environmental protection. Generally speaking, large

and medium enterprises have invested more and achieved relatively more

significant outcomes with respect to energy saving and environmental pro-

tection, while small firms have made much slower progress. According to

one study by CISA, the coking rate of the large enterprises was only 58–80

per cent of that for the least- efficient small firms, as shown in Table 9.7.

Finally, mergers and acquisitions (M&As) within the steel industry are



promoting energy saving and pollution reduction. Baosteel and Shoudu

(Capital) Steel Corporation acquired Bayi Steel and Shuicheng Steel

in early 2007 and 2006, respectively. From a technical perspective, the

two acquired companies had made great progress in energy saving after

being merged with the large firms (Tables 9.8 and 9.9). In both cases, the

records for energy (water) saving were remarkable. Within two years,

energy consumption was reduced by 21 per cent for Shuicheng Steel and

11 per cent for Bayi Steel and their water consumption was reduced by 29

and 39 per cent, respectively. The policy implication is clear that China

should accelerate the pace of consolidation in the steel industry through

M&As or other means of cooperation such as taking over through share-

holdings. This would not only change the industry’s rates of energy effi-

ciency, water use and pollution, but also improve the industry’s overall

performance through increased industry concentration, expanded econo-

mies of scale, application of more advanced technologies and efficient

management. Of course, in order to achieve this goal China needs to

overcome the rampant local protectism which has prevented M&As from

happening in the past.

Table 9.7 Comparison: large and small steel- makers in energy saving,

2006

Coking rate (kg/t) Comprehensive coking rate (kg/t)

Large enterprise (A) 329.5 483.5

Small enterprise (B) 569.0 602.0

A/B 57.9% 80.3%

Note: Comprehensive coking rate means the total energy consumed including the coke, PCI and electricity in producing hot metal per tonne.

Source: The data of large steel- makers are the average figure of Baosteel, Angang, WISCO and Shoudu (Capital) Steel Corporation. The data of small steel- makers are the average figure of the last four in the 73 key steel- makers collected by CISA.

Table 9.8 Comparison: before and after M&A of Shuicheng Steel

2005 2007 Change (%)

Comprehensive energy consumption (kgce/t) 829.37 659.11 −20.5

Freshwater consumption (t/t) 4.77 3.40 −28.7

Coking rate (kg/t) 497 412 −17.1




CONCLUSIONS

The Chinese steel industry has made significant progress in achieving

energy conservation and environmental protection. However, based on

international benchmarks there is still considerable room for improve-

ment. To make further progress, the Chinese steel industry should endeav-

our to enhance technological advancement, seek economies of scale and

further improve management through deepening corporate reform. The

industry and government should allow more M&As to take place in order

to increase industrial concentration. Much needs to be done to continue

building on the achievements in the industry made thus far. The steel

industry has a major part to play in achieving China’s ambitious national

goals regarding energy saving and environmental protection.

NOTES

1. See CASS (2005), table 11- 8, p. 174.2. See Ministry of Science et al. (2007), p. 349.3. See Ministry of Science et al. (2007), p. 353.4. Since 2003, the NBS has not published the growth rate of the steel industry based on

constant prices.

REFERENCES

Chinese Academy of Social Sciences/Institute of Industrial Economics (CASS) (2004), China’s Industrial Development Report, Beijing: Economy and Management Publishing House.

Ministry of Science and Technology, China Meteorological Administration, and the Chinese Academy of Sciences (2007), China’s National Assessment Report on Climate Change, Beijing: Science Publisher.

Garnaut, R., L. Song and Y. Yao (2006), ‘Impact and significance of SOE restruc-turing in China’, China Journal, 55, 35–63.

Table 9.9 Comparison: before and after M&A of Bayi Steel

2005 2007 Change (%)

Comprehensive energy consumption (kgce/t) 675.2 601.43 −10.9

Freshwater consumption (t/t) 9.45 5.78 −38.8

Coking rate (kg/t) 466 449 −3.6




Ministry of Environmental Protection (MEP), (2007), A Statistical Index System of Energy Consumption Per Unit of GDP, November, Beijing: MEP.

National Development and Reform Commission (NDRC) (2005), Decree 35: Steel Industry Development Policy, July, Beijing: NDRC.

State Environmental Protection Administration (SEPA) (2006), Cleaner Production Standard – Iron and Steel Industry, HJ/T189- 2006, Beijing: SEPA.

Wang, C. and J. Chi (2007), ‘Some analyses and suggestions to current energy saving of the Chinese steel industry’, China Steel Focus, 3, 37–40.

Wang, T. and J. Chi (2007), ‘Advanced equipment and reasonable standard of energy use’, China Metallurgical News, 26 April, p. 5.

Wang, W. (2007a), ‘How to reduce energy consumption in the steel industry?’, China Metallurgical News, 18 January, p. 5.

Wang, W. (2007b), ‘CDQ, a worthwhile popularized technology for energy- saving and pollution reduction’, China Metallurgical News, 19 May, p. 6.

Wang, W. (2007c), ‘TRT, a notable technology for energy- saving’, China Metallurgical News, 28 June, p. 6.


177

Glossary

ABARE Australian Bureau of Agricultural and Resource Economics

ACF Ackerberg et al. (2008)

AFC Asian financial crisis

BF blast furnace

BOF basic oxygen furnace

CDQ coke dry quenching

CICC China Industry Classification Code

CIS Commonwealth of Independent States

CISA China Iron and Steel Association

COD chemical oxygen demand

DF Dickey–Fuller

EAF electric arc furnace

EUROFER European Confederation of Iron and Steel Industries

FD first- difference regression technique

FDI foreign direct investment

GFC global financial crisis

GFCF gross fixed capital formation

GMM Generalized Method of Moment

IRTS increasing return to scale

ISIC International Standard Industry Code

IU intensity of use

IVA industrial value added

KCS Kuznets curve for steel

kgce kilograms of coal equivalent

LMEs large and medium enterprises

LP Levinsohn and Petrin (2003)

LR log- likelihood test

M&As mergers and acquisitions

MEP Ministry of Environmental Protection

NBS National Bureau of Statistics

NDRC National Development and Reform Commission

OHF open- hearth furnace

OLS ordinary least squares

OP Olley and Pakes (1996)



PCI pulverized coal injection

PPP purchasing power parity

PRC People’s Republic of China

R&D research and development

SASAC State- owned Assets Supervision and Administration

Commission of the State Council

SEPA State Environmental Protection Administration

SEs small and private enterprises

SOEs state- owned enterprises

tce tonnes of coal equivalent

TFP total factor productivity

TRT top gas pressure recovery turbine

VAT value added tax

WSA World Steel Association

WTO World Trade Organization


179

Index

Bold text for graphs and tables

aluminium 30Anben Group

annual steel production levels of 167Anshan Steel

under administration of SASAC 168

Argentinaauto penetration in 24GDP per capita 24

Asian fi nancial crisis (1997–9) 13impact on production in steel

industry 5Association of Southeast Asian

Nations (ASEAN)members of 131steel imports of 132, 133

Australia 32automobile penetration in 24GDP per capita 24iron ore production in 5

Baoshan Iron and Steel Corporation 7–8

launch of (1978) 4operational (1985) 4

Baosteel Groupacquisition of Bayi Steel (2007) 174,

175annual steel production levels of 167under administration of SASAC 168

Bayi Steelacquired by Baosteel Group (2007)

174, 175Belgium

automobile penetration in 24GDP per capita 24

Brazilautomobile penetration in 24economy of 30GDP per capita 24

stage of industrialization in 47steel use per capita 51

British Petroleum (BP)proven Chinese coal reserves 139

Bureau of Metallurgical Industry 6

Canadaautomobile penetration in 24GDP per capita 24steel export ratio of 131

China 13, 17, 25, 175automobile penetration in 24, 34, 37Beijing 33, 93, 136, 173Cultural Revolution 2economy of 1–3, 5–6, 14, 17, 33, 36,

45–6, 55, 65, 71, 124, 139, 146, 166

entry into WTO (2001) 5, 138founding of People’s Republic of

(PRC) (1949) 2GDP per capita 47, 49, 63, 65, 149government of 5, 10–12, 14, 138,

142, 163, 168, 173Guizhou 51Hong Kong 23, 132, 133, 168hukou system of 47income per capita 37industrial emissions from 10, 162iron consumption rate of 158, 160iron imports of 145–6, 148, 154, 157,

158–60iron exports of 145, 152–3iron ore reserves of 14, 145, 150,

151, 153iron production in 1–2, 71, 72, 77–8,

80, 103, 100, 104, 105, 110, 112–13, 147–8, 150, 152, 159, 171

KCS of 33Macao 132, 133



manufacturing sector of 106–8, 110, 120–22, 124

metal intensity of 11–13, 17–18, 25–8, 30, 32–5, 37–9

net export of steel by 11Northern Song dynasty (920–1126

CE) 1openness to FDI 31population of 23, 49PPP rates of 137proportion of electric furnaces using

scrap for steel production 10proven coal reserves of 139provinces of 49–50, 51, 52, 53–4, 55,

57, 58–60, 62, 173rate of iron ore consumption 6, 38,

145–6ratio of continuous casting in 2Shanghai 33, 51, 93share of world exports 31, 36stage of industrialization in 47,

50state-owned enterprises (SOEs) in

4, 8, 13, 72–3, 84–5, 87, 89–92, 100–103, 126, 168–9

steel consumption rate of 3, 5, 7, 20, 37–9, 45–6, 49, 52–3, 55–7, 58–9, 60, 62–3, 64, 65, 66, 129, 149, 160

steel exports of 129, 130–31, 132–3, 137, 140, 141, 142, 145, 153

steel imports of 129, 130–31, 133steel production in 1–3, 5–7, 9,

43–4, 53, 55–7, 58, 72, 74, 77–8, 80–81, 100, 103, 104, 105, 110, 112–13, 129, 134, 138, 151, 162–3, 164–6, 167, 171, 172

urbanization rate in 5, 35, 47China Industry Classifi cation Code

(CICC) 111, 115, 127, 128level in Chinese manufacturing

sector, 108–9China Iron and Steel Association

(CISA) 6, 165, 172member of WSA 6members of 135, 136

Chinese National Bureau of Statistics (CNBS) 97

Annual Manufacturing Enterprise Census 106, 108, 124

estimate of crude steel production levels (2007) 107

Chinese Statistical Yearbookdata provided by 98

CITIC Group 169Commonwealth of Independent States

(CIS)steel imports of 131, 133

copper 30consumption of 25, 26–8

European Confederation of Iron and Steel Industries (EUROFER) 138

European Union (EU)net export of steel by 11proportion of electric furnaces using

scrap for steel production 10steel imports of 133

fi rst-diff erence (FD) regression technique 59–60, 61

foreign direct investment (FDI) 31, 98, 115

linkage of Chinese manufacturing sector 108

Fosun Internationalannual steel production levels of

168–9France

automobile penetration in 24economy of 1GDP per capita 24

general least squares (GLS) estimation 74

uses of 74German Steel Federation 138Germany

auto penetration in 24economy of 1GDP per capita 24percentage of long steel products

produced by 81Global Financial Crisis (GFC)

(2008–9)impact on steel production 131

gross fi xed capital formation (GFCF) 26–9

as percentage of GDP 25


Index 181

Haitistage of industrialization in 47

Hanyang Iron Worksestablishment of (1890) 2

Indiasteel imports of 131, 133

Indonesiaauto penetration in 24GDP per capita 24stage of industrialization in 47

industrial value added (IVA) 26–8, 30

as percentage of GDP 25decline due to services activity 29

industrialization 1, 7, 11, 13–14, 17–18, 22–3, 33, 45–6, 53–4, 59, 66, 89, 149

stages of 10, 21, 30, 46–7, 48, 50–51, 129, 139, 160, 162

strategies for 34International Monetary Fund (IMF)

25project for automobile ownership

per thousand persons 37International Standard Industry Code

(ISIC)level in Chinese manufacturing

sector 108iron 6, 38, 69

agricultural use of 1consumption rate of 158, 160mining of 90ore production 5, 14, 71, 103, 105,

108, 145–6, 150pig iron 71, 90, 153, 170

Italyautomobile penetration in 24GDP per capita 24

Japan 13, 23, 35, 168automobile penetration in 24economy of 1, 45–6GDP per capita 35metal intensity of 25–8, 30, 39net export of steel by 11, 35–6openness to FDI 31percentage of long steel products

produced by 81population of 23

ratio of continuous casting in 2share of world exports 36steel exports of 130–31steel use per capita 51

Kuznets, Simon 46Kuznets relationship

concept of 18intensity of use (IU) analysis 19–21Kuznets curve for steel (KCS) 19,

21–2, 33, 35, 58

Luxembourgstage of industrialization in 47

Malaysiaautomobile penetration in 24GDP per capita 24

Mexicoautomobile penetration in 24GDP per capita 24

National Bureau of Statistics (NBS) 78, 173

census conducted by (1998–2007) 77National Development and Reform

Commission (NDRC) 173Decree 35: China’s Steel Industry

Development Policy (2005) 140, 173

implementation of ‘About Restricting Iron and Steel Firms’ Rush Investment’ and ‘Iron and Steel Industry Development Strategy’ (2003) 91

implementation of ‘Accelerating Structural Change in Iron and Steel Industry’ (2006) 91

ordinary least squares (OLS) regression technique 77, 109, 118

Organisation for Economic Co-operation and Development (OECD)

members of 131

Panzhihua Steelunder administration of SASAC

168



Peruiron ore production in 5

Plaza Accord (1985) 35Poland

stage of industrialization in 47production theory

concept of 93–4models of 94–6

purchasing power parity (PPP)Chinese rates of 137US rates of 137

Russian Federationautomobile penetration in 24GDP per capita 24net export of steel by 11steel export ratio of 131

State-owned Assets Supervision and Administration Commission of the State Council (SASAC)

companies under administration of 168

Second World War (1939–45) 21Shagang Group

annual steel production levels of 167–8

Shuicheng Steelacquired by Shoudu (Capital) Steel

Corporation (2006) 174Shoudu (Capital) Steel Corporation

acquisition of Shuicheng Steel (2006) 174

implementation of contracting (1981) 3

pricing of coking coal delivered to 169

Singaporeautomobile penetration in 24economy of 46population of 23

South Africaautomobile penetration in 24GDP per capita 24

South Korea 17, 25, 38, 168economy of 1, 22, 30, 45–6metal intensity of 12–13, 25–8, 30,

34, 39net export of steel by 11openness to FDI 31–2

percentage of long steel products produced by 81

population of 23share of world exports 36stage of industrialization in 47steel export ratio of 131steel imports of 131steel use per capita 51

Spainautomobile penetration in 24GDP per capita 24

State Environmental Protection Administration (SEPA)

‘Cleaner Production Standard – Iron and Steel Industry’ (2006) 173

steel 10, 38–9, 69, 135casting techniques for 169, 170consumption rates of 3, 5, 7, 20, 37–

9, 45–6, 49, 52–3, 55–7, 58–9, 60, 62–3, 64, 65, 66, 129, 149, 160

crude 2–3, 7, 22, 45–6, 51–4, 55–7, 58–9, 60–63, 64, 65, 66, 71, 73–4, 90, 107, 130, 135, 149, 153, 158, 162–4, 165–6, 168

products constructed using 7, 9, 80–81, 163

use of basic oxygen furnaces (BOF) in production of 73, 164

use of blast furnaces (BF) in production of 170–71

use of coke dry quenching (CDQ) in production of 172

use of electric arc furnaces (EAF) in production of 73

use of open-hearth furnaces (OHF) in production of 4, 73

use of top gas pressure recovery turbine (TRT) in production of 172

Swedenautomobile penetration in 24GDP per capita 24

Taiwaneconomy of 22steel export ratio of 131steel imports of 131, 133steel use per capita 51

Tangsteel Groupannual steel production levels of 167


Index 183

Thailandautomobile penetration in 24GDP per capita 24net export of steel by 11

Tianjin Seamless Steel Tube Corporation 7–8

launch of (1989) 4operational (1996),4

total factor productivity (TFP) 82–6, 87, 111, 114

concept of 109estimations of 69–71, 74, 76–7, 81,

87, 118, 124regression of 116–17shock 75

Ukrainenet export of steel by 11steel export ratio of 131

United Arab Emirates (UAE)net export of steel by 11

United Kingdomautomobile penetration in 24economy of 1GDP per capita 24Industrial Revolution 1

United Nations 35United States of America 13, 17, 23,

32, 38–9anti-dumping measures undertaken

by 132automobile penetration in 24–5, 34economy of 1, 21, 45GDP per capita 24, 35–6, 47, 51KCS of 21, 35percentage of long steel products

produced by 81PPP rates of 137proportion of electric furnaces using

scrap for steel production 10share of world exports 36steel intensity of 24steel use per capita 18, 19, 51

Word Steel Association (WSA)members of 6, 31

World Trade Organization (WTO)Chinese entry into (2001) 5, 138founding of (1995) 132

Wuhan Iron and Steel (WISCO)annual steel production levels of 167under administration of SASAC 168


the chinese steel industryâ€™s transformation : structural change, performance and demand on...

Documents