3.Sources of Structural Changes

3.1 Theory and Methodology

As Tab.1 shows, a country produces intermediate inputs domestically for output. When it is not enough, the country imports inputs to satisfy internal and external demands, including production, consumption（C）, investment（I）, government expenditures（G）and export.Meanwhile, the produced output is also used as domestically produced inputs in next period.The process circulates back and force.There are two ways to transform export:altering domestic production and altering inputs'import.If the transformation into sophisticated sectors was based on increase of imported inputs, local content of these sectors'export was reduced and china gained less from high qualified export.The structural change was therefore not satisfactory.Indeed, domestic value added does not exactly equal profit of export but it is one of the most important factors affecting the gain. Hummels, Ishii, Yi（HIY 1999）propose the concept of Vertical Specialization（VS）, noting imported intermediate inputs in the production for export.The share of VS is also the share of foreign value added.This study aims at analysing variations of VS share of Chinese export and find out the main source of china's structural changes.

From Tab.2, we make series of equations as what have done by HIY1999.The vertical calculations are showed as follows:

Where denotes domestically produced inputs of sector i for producing sector j, denotes imported inputs of sectori for producing sector j, VA j denotes value added in production of sector j.Dividing them separately by Total Input of sector j（Xj）obtains domestic input coefficient（aij）, imported input coefficient（bij）and value added ratio（cj）. AD is n×n direct input matrix of domestic production of inputs, AM is n×n imported input matrix, AV is 1 × n value added matrix, u is unit matrix（1 × n diagonal matrix）.Eq.（4）illustrates vertical balance, which are used for testing if matrix'calculation is correct or not.The vertical specialization（VS）is computed by HIY1999 definition in eq.（5）. EXPi denotes export of sector i.EXP is n×1 export matrix.

Total imported inputs of sector i are used to produce sector j（）, at the same time, sector j's output will be used for export, consumption, investment or for other domestic production in the second, the third...period.Besides of direct imported input, indirect one should be also taken into account.Before being exported, imported inputs circulate in several stages of production, writing as:

The final methodology of computing vertical specialization share is defined in eq.（7）and eq.（8）.VS S i is imported inputs in production for export of sector i.VS S t is share of vertical specialization in Chinese total export.（I-AD）-1 is the well-known Leontief inverse matrix.EXPt is the overall export of China.When considering horizontal balance, we refer to Koopman（2008）study and obtain the domestic and foreign value added share.

FM n ×1 matrix denotes final use of imported inputs of sector i and FD n ×1 matrix denotes final use of domestic inputs of sector i, including consumption expenditures, investment and export.matrix denotes intermediate use of domestic inputs of sector i, matrix denotes intermediate use of domestic inputs of sector i. I MP is n ×1 import matrix.X denotes n ×1 output matrix.For each sector, gross output equals gross input.Thus we obtain eq.（11 a）then eq.（11 b）.DVA denotes domestic value added.F VA denotes foreign value added.According to eq.（13）, F VA equals share of VS in eq.（7）.

3.2 Data and Preliminary Results

This paper utilises two datasets:OECD and China National Bureau of Statistics（NBS）.Tab.3 displays their differences.Firstly, available years of OECD I/O table are more than that of NBS.However, the latest year of NBS is 2007, more recently than the former（2005）.Indeed, NBS also publishes I/O table for 1995,2000 and 2005.These tables only contain 33 or 42 sectors according to China custom classification.Thus there is no need to employ them and if employ, we will meet a problem of consistence with OECD classification.Secondly, NBS I/O table embodies 123 sectors for 1997,2002 and 135 sectors for 2007, while OECD identifies 41sectors for 1997 and 48 sectors for other years.As for manufacturing, NBS contains 82sectors, while OECD only 22 sectors.Thirdly, OECD constitutes imported I/O table but NBS does not publish any officially.We then distinguish imported matrix of NBS table, calculate VS share of 1997,2002 and compare them with that of OECD.If they are compatible, we will extend the period to 2007 by using NBS table.

At outset, we reconstruct NBS 123 and 135 sectors'table into 48 sectors, which are the same as OECD sectors.Then, we assume that the share of imported intermediate inputs in total intermediate inputs of sector i equals the ratio of import relative to the sum of import and gross product of this sector.

Ping（2005）has proved this assumption.Although OECD computes imported inputs' share by custom data, it also utilizes this assumption when the data are not available.As a result, for more than 1/4 sectors, difference of VS share between OECD and NBS is less than 0.002.For the rest, variation over 97-02 is similar.For some sectors such as machinery & equipment, motor vehicles, results by two dataset are equivalent.We have reason to believe the robustness of the assumption and the methodology and it is feasible to extend OECD table to 2007 with adjustment by NBS 2007 table.Fukasaku（2011）computes the VS share with OECD data and IDE-JETRO tables.His main findings are showed in column（1）of Tab.4.We average the gap between OECD and NBS for each sector at 1997 and 2002, then plus NBS VS share for 2007（eq.15）.As for other years, we esti mate VS share according to eq.（16）.VSS i, oecd, t is VS share of sector i at year of t measured by OECD.The adjusted VS share is showed in column（4）.Generally, the VS share of Chinese export increased by 11% during 1995-2005.That of manufacturing increased nearly 13% over this decade.It means that domestic value added and the gain from export decreased.Fortunately, it seems to be up in 2007 like what has happened in 1997.

By technological level

VS share by technological level is showed in Tab.5 and Tab.6.Tab.5 highlights the results'comparison, which separates manufacturing into two tech levels.For high and medium high tech level, NBS underestimates 2% VS share relative to OECD.As for low and medium low tech level, there appears no large difference.The sophisticated sectors' export needs more imported inputs than low skilled one.For instance, in 20051000$Chinese exports of sophisticated sectors need 372$ imports of intermediate input, while that of low skilled sectors need 213$.Meanwhile, the needs of the former went up faster than the later with growth rate 16% than 6%.Also with adjustment in eq.（15）for 2007, it infers that the foreign value added of export for both levels diminished during 20052007.Tab.6 illustrates variation of VS at four tech levels.Chinese export depends more on import than before.Especially for high-tech export, the VS share has rose by 30% during the decade of 1995—2005.Besides, the higher tech level, the more need of inputs'import.In this sense, the clear technology scissors difference over 1994-2004 was under the background of the decreasing domestic value added, in other word, of decreasing gain from export.China's structural changes were not satisfactory.

By sector of manufacture

Fig.5 illustrates VS share of manufacturing sectors during 1997,2002 and 2007 with measurement by NBS IO table.Fig.6 illustrates manufacturing VS share calculated by OECD for 1995,2000 and 2005.Indeed, measurement by OECD table engenders the same variation tendency over 1997-2002-2007 as NBS, thus we do not display this variation in graph 6, Tab.7 and Tab.8 illustrate growth rate of these VS share over five years.In Fig.5, the first six largest VS share belong to sophisticated sectors.Nearly all sectors'VS share increased during 1997-2007 except for sector 5 Textile, which decreased by 7% from 2002 to 2007（Tab.8）.Fig.6 shows the same phenomena, Except for sector 13 Iron and steel, all the sectors needing large VS share are sophisticated sectors.

The VS share of high-tech sectors, including sector 10,17,19,20,23, continued increasing without stop during 1995-2007.Among them, sector 17（office accounting computing machinery）and sector 19（radio, television and communication equipment）were the sectors that need the most imported inputs, up to 40% in production for export. The VS share of medium high-tech sectors, including sector 9,16,18,21,22,24, was generally 20%-30%.It went down over 1995-2000, i.e.China made progress in local content and succeeded in structural reform during these years.Unfortunately, it has increased again since then.Sector 18（electrical machinery and apparatus）needed the most imported inputs among them.Its VS share also diminished from 1995 to 2000 and re-increased to 30% in recent years.Besides, sector 21（motor, vehicle and trailers）increased local content during 1995-1997,2000-2002, but decreased in the other years.The rest sectors in low-tech and medium low-tech level did not change so much, except for some sectors, such as sector 7（paper products）, sector 8（coke, petroleum, nuclear）and sector 11（rubber, plastics）.Its VS share increased the most among all the manufacturing sectors over 1995-2000 and decreased the most over 2000-2005.That signifies China enhanced local content of export in recent years only for low qualified sectors.The foreign value added of sector 13（iron and steel）jumped 73% over 2000-2005 and 36% over 2002-2007.Sector 5（Textile）was the only sector that decreased foreign content of export during 2002-2007.

Fig.7 measures VS share by NBS 135 sectors table.Nearly all the high and mid hightech sectors increased VS share without stop, except for computer manufacture, electronic equipment（VS share decreased a little from 2002-2007）and electronic components（VS share decreased from 1997-2002, after then re-increased）.As for low and medium lowtech level, sectors of coking, rubber, fire-resistant materials, steel and tobacco decreased foreign value added（increased domestic value added）during 1997-2002.Sectors of textile, Knitted fabric, textile wearing apparel, leather, fur and feather increased local content of export during 2002-2007.Totally, the sectors that increased local content of export during recent years are the sectors at low and medium low tech level.As for sophisticated sectors, only the medium high-tech sector of electronic component increased domestic value added.China's pattern of structural changes is not durable.If it wants to continue development of sophisticated sectors, it must enhance local content of export.

3.3 Combine I/O Table with Trade Data and Further Findings

By source country and technological level

The trade data in this part derive from CHELEM-CIN under ISIC Rev.3 classification.VS share supposes to be the same among different partner countries, from which china imports.Ping（2005）and Miroudot（2009）also use this hypothesis in their studies. The share of imported inputs from country k in the production for Chinese export is written in eq.（17）.IMP ik is Chinese import of sector i from country k.IMP i is Chinese total import of sector i.

Fig.8 shows the share of intermediate inputs imported from other partners.Totally speaking, manufacturing production for export needed more imported inputs than overall economic production.At outset, the United States, Japan and European industrialised countries were the main source of Chinese import.They occupied more than half of Chinese vertical specialization.Yet, their variations were relatively stable.US's share did not vary over times.Japan and European export of intermediate inputs made a better showing than the former but also did not much progress.Their share of Chinese inputs'import became less than the half in 2007.On contrary, ASEAN economies and Asian new industrialized countries emerged quickly.During 1995-2005, imported inputs from Korea rose by 3% of Chinese manufacturing export.That from Taiwan rose by 2.4%.These two countries contributed 43% to Chinese overall increase of VS share（Tab.9）.When considering technological level showed in Fig.9, ASEAN countries'export of high qualified intermediate inputs increased quickly, so did Korea and Taiwan.These three contributed 90% to China's high-tech VS growth as showed in Tab.10.While US, Europe and Japan did not vary so much.For export of medium high-tech and medium low-tech, Korea, Taiwan and Europe increased faster than others and contributed the most to Chinese VS increase.It is interesting that for low-tech export, Europe increased the most rapidly and contributed the most, while Korea and Taiwan went down.Besides, the biggest contributor appears to be also the rest countries of the world（ROW）like Latin America.

Fig.10 illustrates the evolution of each country's share in Chinese overall import. There are four types of evolution.（1）The share decreased for all technological level products, such as in Japan.（2）The share increased for all technological level, such as in ASEAN countries.（3）The share decreased for high technique level while increased for other lower one, such as in Europe and the United States.（4）The share increased for sophisticated sectors（high and medium high tech level）while decreased for other lower one, such as in Korea and Taiwan.This is compatible with the conclusion from Fig.9.

In total, Asian new industrialized economies（NIE）and ASEAN countries'contribution increased rapidly, especially for high-tech products.The structure of China's vertical specialisation with Korea and Taiwan shifted from low-tech to high-tech.The traditional three sources of inputs'import US, Japan and Europe performed relatively stable.Asia is the largest contributor of Chinese VS of manufacture, especially for high-tech（92%）and medium high-tech（73%）.For VS of low-tech, they are Europe and Latin America who contributed the most.