sophistication of china’s exports and foreign...

50
Sophistication of China’s exports and foreign spillovers Joachim Jarreau * and Sandra Poncet September 17, 2009 Abstract Recently a number of studies have focused on the upgrading of China’s exports, a process which has started in the early 1990s and has accompanied the impressive increase in volume of China’s trade. It appears that the area of specialization of China’s economy has shifted from labor-intensive sectors to more skill and capital-intensive sectors, as well as medium and high-technology sectors. Several authors have recently tried to quantify this fact, and to iden- tify the determinants of Chinese export sophistication. We used a database of China’s exports over the period 1997-2007 to investigate these questions. First, we relied on a measure of export sophistication proposed by Rodrik et al. (2006), and find in line with these authors that China exhibits a level of sophistication strikingly high for its overall level of development. We are able to distinguish between exporters by type of firm and between processing and ordinary trade, and thus to identify their respective contributions to the overall sophistication : not surprisingly, we find that the processing sector in China is significantly more sophisticated; however, the ordinary sector has also upgraded importantly over the recent years. Finally, we test the hypothesis of spillovers from foreign invested firms to domestic exporters in China. We find evidence of a positive impact of foreign firms’ export structure on that of domestic firms, a result that is robust to the use of alternative measures of sophistication, as well as to instrumentation. Keywords: FDI, Spillovers, export sophistication, China. * Paris School of Economics Master Student and CEPII; Email: [email protected] Centre d’économie de la Sorbonne (Université Paris 1) and CEPII. Email: sandra.poncet@univ- paris1.fr.

Upload: tranquynh

Post on 06-Feb-2018

216 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

Sophistication of China’s exports and foreignspillovers

Joachim Jarreau ∗ and Sandra Poncet†

September 17, 2009

Abstract

Recently a number of studies have focused on the upgrading of China’sexports, a process which has started in the early 1990s and has accompaniedthe impressive increase in volume of China’s trade. It appears that the area ofspecialization of China’s economy has shifted from labor-intensive sectors tomore skill and capital-intensive sectors, as well as medium and high-technologysectors. Several authors have recently tried to quantify this fact, and to iden-tify the determinants of Chinese export sophistication. We used a databaseof China’s exports over the period 1997-2007 to investigate these questions.First, we relied on a measure of export sophistication proposed by Rodrik etal. (2006), and find in line with these authors that China exhibits a levelof sophistication strikingly high for its overall level of development. We areable to distinguish between exporters by type of firm and between processingand ordinary trade, and thus to identify their respective contributions to theoverall sophistication : not surprisingly, we find that the processing sector inChina is significantly more sophisticated; however, the ordinary sector has alsoupgraded importantly over the recent years. Finally, we test the hypothesisof spillovers from foreign invested firms to domestic exporters in China. Wefind evidence of a positive impact of foreign firms’ export structure on thatof domestic firms, a result that is robust to the use of alternative measures ofsophistication, as well as to instrumentation.

Keywords: FDI, Spillovers, export sophistication, China.

∗Paris School of Economics Master Student and CEPII; Email: [email protected]†Centre d’économie de la Sorbonne (Université Paris 1) and CEPII. Email: sandra.poncet@univ-

paris1.fr.

Page 2: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

Introduction

Since the early nineties, China has been integrating into world trade at an astound-ing pace. Chinese exports have increased by more than 500% between 1992 and2007, faster than the growth rate of the economy. This has radically transformedthe functioning of China’s economy, going from an isolated position with exportsrepresenting less than 10% in 1980 to a highly integrated one, with an export ratioof more than 30% in 2005. This process has been accompanied by a not less impres-sive diversification of China’s trade, as its manufactured exports pervaded all sectorsof world trade, from low-technology textile to high-tech electronics and computers.

This success of China’s trade integration has elicited several debates amongeconomists. Much of the attention has focused on the upgrading of China’s ex-ports: economists (but also world consumers) have noticed since the mid-ninetiesthe impressive broad range of China’s export products and in particular the abilityof chinese producers to export capital and skill-intensive products, high-technologyproducts, and in general products which are usually thought to belong to the area ofspecialization of more developed countries, those relatively more endowed with thecorresponding factors. The success on world markets of chinese firms in the sectorsof ITs, mobile phones and electronics, such as Lenovo and Founder, TCL, Skyworth,have been symptomatic of this evolution. From a theoretical point of view, this fea-ture of China’s trade raises the question of whether this pattern of specialization ofChina is at odds with standard trade theory. Given that China is relatively more en-dowed in low-skill labour, why should it export skill and technology-intensive goods,and perhaps more importantly, why should it gain from it? Some authors haveargued that the Chinese example provided a case for moving beyond the standardframework of trade theory, and to emphasize the importance, for developing coun-tries, to aquire the capabilities to produce and export goods with a high level of tech-nology. In this view, China’s policy of technological development, and in particularits efforts to import foreign technology through opening to foreign investment, hasenabled Chinese exporters to enter markets of ‘sophisticated’ manufactured goods,and this has been key to the success of China in world trade.

Opposing to this view, several authors have argued that the apparent upgradingof China’s exports has been misleading: it would only be attributable to foreignproducers operating in the assembly trade sector, that is, the production of high-

2

Page 3: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

technology goods using labor-intensive processes and imported inputs. The factthat the final assembly of these goods take place in China would be simply theresult of higher fragmentability of production processes and of production sharing,and should not indicate a higher domestic capability to manage technology.

Identifying the right diagnosis is crucial to anticipate the future evolution ofChina’s economy and its position in world production patterns: in the words of B.Naughton, the question is to know if “China [will] remain the world’s factory, churn-ing out massive quantitities of laboriously produced goods? Or will China insteadvault into the front ranks of world economies by contributing new products and pro-cedures, innovative standards, and breakthrough ideas? Will China become a globalcenter of innovation, joining the ranks of developed countries and the emerging EastAsian centers of technological creativity?”. Effective adoption of technology is nec-essary for China to increase its productivity and sustain high growth rates. It is tothis purpose that China has adopted, starting in the early 1980s, a policy of openingto foreign investment, in the hope that technological capabilities and managementpractices would spill over to domestic producers. Evaluating the efficiency of thispolicy is the central question of our study.

Using a database of Chinese customs trade data covering the 1997-2007 period,we first tried to quantify the upgrading of China’s trade. We were able to breakdown China’s export flows by type of firm ownership and type of trade (assemblytrade, or ordinary). We found that the recent upgrading of China’s trade has alsoconcerned domestic producers operating in the ordinary trade sector: their exportstructure has significantly shifted from low-technology manufactures, such as textileand garment, to medium and high-tech products such as machinery and electronics.

Next, we tried to test if the export performance of domestic producers had bene-fited from the presence of foreign producers. We find that the level of sophisticationof foreign producers’ exports had a significant impact on that of domestic produc-ers. Using alternative measures of export sophistication, we find that this impact ismost important in the range of medium-technology products, and that it operatedat least partly through the extensive margin of trade (the opening of new productlines). Finally, we adress the potential endogeneity of our variable of interest, the

3

Page 4: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

sophistication of foreign firms, by using instrumental variables.

The rest of this article is organized as follows. In the first section, we present thetheoretical literature relative to the concept of export sophistication, as well as themeasures which have been proposed of it. Among these, we will be using the mea-sure of Rodrik (2006), which we present in greater detail. We then present empiricalevidence about the recent evolution of China’s export structure. In the second sec-tion, we review the existing literature on the determinants of export sophistication,and the empirical literature on the importance of foreign spillovers on the economicdevelopment and export perfomance of host countries. Finally in the third section,we present our statistics on the export structure of China’s producers by firm typeand trade type; then the results of our regressions, estimating the impact of foreignspillovers.

1 Sophistication of China’s exports : theoretical lit-

erature and empirical evidence

1.1 Export sophistication: definition and measures

The development of China’s export sector, and the apparent correlation between itsupgrading process, the increasing share of its exports in China’s GDP and in worldtrade, and the simultaneous growth, have brought up the question of whether, andwhy, this technological upgrading is beneficial to a developing economy such asChina. Indeed, according to standard trade theory, China should benefit from spe-cializing in labour, low-skill intensive production processes, for which its endowmentsprovide it with a comparative advantage. An important theoretical literature hasproposed models going beyond this framework to explain why developing countriesshould gain from importing and adopting technology from more advanced economies.These models focus on the process of learning and adoption of new technologies as acostly process: the ability to produce a good efficiently and export it will be deter-mined not only by the capacity of the country (embedded in its endowments), butalso by the existing capabilities, that is, the ability to manage a given technology inan efficient manner. This represents a significant departure from the neo-classical

4

Page 5: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

framework, in which technology is assumed to be easily codified, priced, and im-ported by producers in developing countries in a readily exploitable form. Here, itis on the contrary argued that much of the new technology is “tacit” (Rodrik); and,moreover, that the process of learning new technologies is subject to externalities,such that public policies designed to encourage this process are needed and have animpact on the export pattern of countries. In this framework, patterns of compara-tive advantage across countries will depend not only on their respective endowments,but also on the national policies encouraging technological learning and technologyimport.

Rodrik and Hausman (2003) is one example of such a model, in which the em-phasis is put on the costly process of “self-discovery”, that is, of learning where thecomparative advantage of a country lies. These authors argue that the process oflearning what can be produced at low cost in the country, requires an investment byentrepreneurs, in the form of trial and errors; the knowledge thus produced generatesan important social value in the form of externalities, for which these “discoverers”are, in general, not fully rewarded. Hausman, Hwang and Rodrik(2006) extend thismodel to examine the consequences of externalities in local cost discovery in terms ofexport patterns: the export structure of a country will depend not only on its intrin-sic comparative advantage, but also on the resources allocated to the cost discoveryprocess. More precisely, the range of goods that a country can potentially produceis determined by its endowments in skills and human capital; but the actual basketof goods which will eventually be exported by the country, and the associated levelof productivity, will be the result of the intrisically uncertain process of discovery ofthe underlying cost structure of the economy. In turn, countries which aquire thecapability to export a basket of products of higher productivity (given their initialfactor endowments) will gain more from trade and achieve higher growth rates.

Similarly, Lall(2000a) also puts emphasis on the externalities associated to theprocess of aquiring technological capabilities, insisting on the importance, for de-veloping countries, to adapt technologies rather than to genuinely innovate; he con-cludes on the benefits, for developing countries in particular, to design policies aim-ing at supporting import and adoption of foreign-developed technologies by domesticproducers.

These models provide a case for moving beyond the neo-classical theory of

5

Page 6: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

growth, according to which openess to trade and institutions of good quality arethe only necessary ingredients to growth and the process of convergence of devel-oping countries economies towards advanced economies. They argue that “what acountry exports matters”. But how to characterize the potential for growth in agiven country’s export basket? Hausman, Hwang and Rodrik (2006) propose to usean indirect measure of the productivity of a country’s export basket, obtained bycomparing it with the income level of countries with similar export structures. In thefollowing section, we present this measure of export ‘sophistication’ in more detail,as we will be using it extensively in our study. We then discuss alternative measuresof export sophistication and compare them.

Measures of export sophistication

In the model of Hausman, Hwang and Rodrik (2006), each good which a country canpotentially produce and export is assumed to have an intrinsic level of productivityassociated to it. The overall level of productivity of a given export basket willbe the outcome of the uncertain cost discovery process. As an empirical measureof export ‘sophistication’, these authors propose to infer the revealed productivitylevel associated to the production of a given good by observing the level of incomeof countries which export that particular good :

Prodyk =∑

j

xjk/Xj∑j(xjk/Xj)

Yj, (1)

where Prodyk is the level of sophistication of the good k, xjk is the value ofexports of that good by country j and Xj is the total values of country j’s exports;Yj is the per capita level of income of country j. Thus prodyk is a weighted averageof the income levels of this good’s exporters, with the weights being the share of thisgood in each country’s total exports. The bigger share a given good weighs in theexports of rich countries, the more ‘sophisticated’ it will be.

Next, the measure of a the sophistication level associated with a country’s exportbundle is obtained as :

EXPYj =∑

l

xjl

Xj

Prodyl (2)

6

Page 7: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

Here, the sum is over ll goods exported by the country. Thus, EXPYj is aweighted sum of the productivity levels associated to each exported good, weightedby the share of each good in total exports of the country.

Having defined this measure, one can examine the relationship between the levelof GDP per capita and the export sophistication across countries. One finds that thelevel of income is a good predictor of sophistication, which is to be expected givenits definition; but it appears that China is an outlier in this relationship, with itssophistication measure well above what would be predicted from its level of income,as will be seen below.

The purpose of this measure is thus not to determine directly the intrinsic fea-tures which make a product sophisticated, e.g. the technology embedded in it, thelevels of specialized skills required to produce it, R&D investments, etc.; but ratherto infer, from observed patterns of trade, which products require a high level ofdevelopment to export. Hence, “sophistication” in this sense should be taken asmeasuring, as a whole, a mix of the sources of competitiveness of high-wage coun-tries. This would include technology, in a broad sense : not only product/processinnovation, but also the capabilities needed to handle technologies efficiently, valuechain organization, infrastructure (for example ICTs). But, as noted by Lall, Weissand Zhang (2005), who proposed a very similar measure of product sophistication,this measure can also capture other features of high-income countries’ exports, suchas differentiation, low fragmentability (as high fragmentability can cause productionto be relocated in low-wage countries) availability of natural resources of which richcountries are exporters, and so on. Moreover, trade distortions caused by barriersand subsidies also influence the measure of sophistication, as it modifies the patternfor production across countries. For these reasons, one should bear in mind thatthis measure is an amalgam of all these factors, more than a specific technology, orskill-intensity measure.

Schott (2006) proposes to use another measure of export sophistication, based onthe similarity of export bundles (using the export similarity index from Finger andKreinin 1979), to assess the extent of the overlap between China’s export bundle

7

Page 8: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

with that of OECD countries. Using US import data, he finds China’s sophisticationto be increasing, and above the level of countries with similar endowments. Thisresult is mainly attributable to the high product penetration of China’s exports inmanufacturing sectors, that is, the fact that the product range of chinese exports isextremely broad.

Use of product classifications

Another approach to characterize the structure of exports consists in using prod-uct categories to examine in which sectors did China’s share of exports increase,using disaggregated trade data. Several classifications can be used to this purpose.First, the HS or SITC classifications, at the 4, 6 or 8-digit level of disaggregationdistinguish between industries or sectors. This presents the advantage of follow-ing which industries are gaining weight in a country’s export bundle, but on theother hand, it does not provide us with a measure of sophistication, or of technologyor factor content : for instance, a sector such as “electronics” would include somehigh skill-content, differentiated products such as integrated circuits, but also someelectrical parts produced through labor-intensive assembly. For this reason, someauthors have proposed alternative classifications of products, directly based on thetechnology content. Lall (2000a) has proposed such a classification of products bytechnological level, which combines the OECD classification based on technologicalactivity and an earlier method (based on Pavitt, 1984); it distinguishes betweenresource-based products, primary products, and low-, medium and high-technologymanufactures. Some examples of this classification are shown on table 1; the com-plete list of products can be found in the original paper by S. Lall (2000).

As we will be using this classification as an alternative way to characterize exportpatterns, an important point for our study is to examine how the levels of technol-ogy defined by Lall are consistent with the measure of sophistication proposed byRodrik. As we mentioned earlier, the definition of the sophistication cannot be holdas strictly measuring technology : it possibly takes into account other aspects whichmake a given product likely to be exported by advanced, technology-rich countries.Therefore it is useful for us to compare it with alternative measures of technology atthe level of products, before using one or the other. Another reason for doing this isthat the sophistication measure, and in particular the ranking of products by sophis-

8

Page 9: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

Figure 1: Technological classification of exports, examples (from Lall(2000a))

9

Page 10: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

tication level, is changing over time, following the changes in the pattern of exportsacross countries. On the contrary, the classification of products is time-invariant. Asa result, comparing the two enables us to check whether the sophistication measureis consistent throughout the period of interest : if some products are to be consid-ered as of low sophistication at the beginning of the period, and highly sophisticatedat the end, only because of changes in the export pattern of that product across theworld, then it would raise doubts about the reliability of this measure.

Therefore, we compute for each year in the period considered, average levels ofsophistication of products in each technology category of Lall : there are six cate-gories, “High technology” (HT), “Medium-tech” (MT) and “Low-tech” (LT), “primaryproducts” (pp), “resource-based” (RB) and “others” (excluded). The correspondinggraph is shown on figure 2.

6000

8000

1000

012

000

1400

0av

erag

e pr

ody,

US

con

stan

t 200

0 $

1995 2000 2005 2010year

pp RB LT MTHT

average prody by Lall categories

®

Figure 2: Average levels of sophistication (prody) by product categories

This graph shows that the measure of sophistication of products (“prody”) com-puted as in equation 1, and the techonology-based classification are consistent witheach other, in the sense that products classified as “High-tech” exhibit a significantlyhigher average level of sophistication than those in the “Medium-tech” category,which themselves are more sophisticated than Low-tech products, and so on. Con-sequently, if the share of High-tech products in a given country’s exports increases,

10

Page 11: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

this leads necessarily to an increase in the measured sophistication of these exports.

1.2 The structure and evolution of China’s trade

The impressive growth of China’s exports in volume, of over 500 percent between1992 and 2007, and the increasing share of China in world’s trade (from less than 1%in 1990 to 3.5% in 2001), has understandably elicited a not less impressive increasein the interest of economists in this topic and the number of studies focusing on thestructure and evolution of China’s exports. Among this literature, a number of stud-ies have focused on assessing whether the increase in export volume was accompaniedby a shift in the structure of exports, that is, whether the “sophistication” of Chineseexports had been increasing, and by how much. Rodrik (2006) has contributed im-portantly to bringing attention to what he considered as a “puzzle” in China’s exportstructure. In this paper, the author examined the evolution of China’s export bun-dle from the early nineties, when the process of China’s integration into world tradetook of, with the exports ratio to GDP increasing from 12% in 1990 to almost 30%in 2002. He noticed that China’s export bundle increasingly resembled that of moredeveloped countries, such as electronics, high-technology products and, in general,capital- and skill-intensive products. This particular feature of China appears to re-main when controlling for human capital and other determinants of export structure.Rodrik(2006) interprets this as evidence that China has managed to export a bundleof goods much similar to that exported by countries three times richer in per capitaterms; and that this ability to specialize into skill- and technology- intensive goodswith high value added, has been a key to China’s successful export performance.Beyond the theoretical question of whether China’s pattern of trade was at oddswith standard trade theory, the interest in this question has also been undoubtedlyfueled by concerns arising in the public debate, in particular in the US, that Chinawould be competing more and more with OECD countries in producing the samegoods. The concern is that the fast process of China’s integration into world tradehas not only had de-industrializing effects on countries with similar endowments (cf.for example Wood and Mayer 2009), but that it would also be increasingly com-peting with much more developed countries in the sectors where their comparativeadvantage lies, so that the effects of this increased competition in these countrieswould possibly not be restricted to labour-intensive industries and low-skill workers.

11

Page 12: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

Is China’s technological upgrading only due to processing exports? Is itselling high-tech products at a discount?

An important and unresolved question has been to know whether the observed highlevel of sophistication of China’s export bundle was attributable to processing (orassembly) trade, that is, to sophisticated products being assembled in China, us-ing imported sophisticated inputs and labour-intensive production processes (thistype of trade benefiting from tax exemptions on imported inputs in China); or ifthe ordinary trade sector had also benefited from this product upgrading, in whichcase a larger share of the value added is domestically produced. Lemoine and Unal-Kesenci (2004) find that the development of China’s assembly operations, along withproduction sharing with Asian countries in particular, is mainly responsible for theapparent diversification and technology upgrading of its trade. Analysing trade databy stages of production and by technology content, they find a general picture of‘dualism’ : on one side, medium and high-technology products (such as machinery,electrical parts) are being mainly exported by foreign affiliates after being processedin China, while on the other, domestic firms operate integrated production chains inlow-technology manufacturing sectors (garments, textile, chemistry). More recently,Amiti and Freund (2008) confirm that a large part of the growth of exports in ma-chinery and hard manufacture is attributable to growth in processing trade. Theylook at the evolution of the skill content of exports over the 1992-2005 period, andfind that the increase in skill content is also due to processing trade only. Dean, Fungand Wang (2007) measure the extent of vertical specialization in Chinese trade; us-ing trade data and I/O tables, they find that in average 35% of the value of China’sexports are attributable to imported inputs, up to 50% in some sectors.

Another part of the debate has emphasized the importance of examining not onlywhich products are being exported by China, but also at which price they are sold.Schott (2006) goes on to examine unit values of chinese exports, and finds themto be below the price at which similar products are sold by OECD countries. Thisindicates the need to take into account the process of vertical specialization withinsectors : China could be exporting the same products as richer countries, but notof the same quality. Thus, these findings seem to indicate that if the export bundleof China is at odds with what comparative advantage theory predicts, the relatively

12

Page 13: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

low price it receives from it is consistent with it. Schott also remarks that the use offactor endowments figures only at the national level could mask important disparitiesat the sub-national level, which if taken into account could help to reconcile China’sexport structure with its endowments. Xu (2007) is another example of studieswhich compare unit values of similar products exported by China and by developedcountries, and find that chinese producers sell their products at a discount, whencompeting with more technologically advanced countries, thus partially reconcilingthe observation with trade theory predictions.

First statistics on China’s export sophistication

In this section we make use of our data on chinese trade to assess the level of evolu-tion of China’s export sophistication. Using the definition of export sophisticationintroduced above, we compared sophistication levels across countries, and comparedthem to GDP per capita levels (in real terms, constant dollars of 2000); we find asimilar pattern as in Rodrik (2006), as shown on figures 3 and 4: the level of GDPper capita appears as a significant predictor of the sophistication of world countries’exports, but China is an ‘outlier’ in this relationship. In 1995, the export bundleof China is at about the same level of sophistication as that of countries such asPortugal, Greece or, interestingly, Singapore, although all these countries have upto three times (for Singapore) higher levels of income than China.

We then move on to look at the pattern of export sophistication across provincesin China. Chinese provinces are known to exhibit important disparities in industrialstructure and in overall development; their export performance is also likely tovary with geographical position. Therefore, one can expect disparities in exportpatterns across provinces, such that the observed high level of sophistication ofChinese exports could be imputable to some particular provinces’ trade.

Map 5 displays the levels of export sophistication across provinces in 1997. Weobserve important disparities, with overall, the provinces in the eastern and southernparts of China exporting the most sophisticated products. It is noticeable that theareas where specific policies of liberalization and of openess to trade and foreigninvestment were put in place, such as the special economic zones (SEZs, createdin 1980) and the Economic and Technological development zones (ETDZs, 1984),exhibit some of the highest levels of export sophistication : the SEZs are situated in

13

Page 14: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

AUS

BGDBOL

BRA

CAN

CHE

CHL

CHN

COL

CYP

DEUDNK

DZA

ECU

ESP

FIN

GRCHRV

HTI

HUN

IDNIND

IRL ISL

JAM

KEN

KOR

LCA

LKA

MAC

MDG

MEX

MYS

NLD

NZL

OMN

PER

PRT

PRY

ROMSAU

SGP

SWE

THA

TTOTUR

USA

7.5

88.

59

9.5

5 6 7 8 9 10lgdp

lexpy Fitted values

®

Figure 3: Sophistication levels (expy) and real GDP per cap., in logs, 1995

AUS

BGD BOL

BRA

CAN

CHE

CHL

CHN

COL

CYP

DEUDNK

DZAECU

ESP

FIN

GRCHRV

HTI

HUN

IDN

IND

IRL

ISL

JAM

KEN

KOR

LCA

LKA

MAC

MDG

MEXMYS

NLDNZL

PER

PRT

PRY

ROM

SAU

SGP

SWE

THA

TTOTUR

USA

88.

59

9.5

10

6 8 10 12lgdp

lexpy Fitted values

®

Figure 4: Sophistication levels and real GDP per cap., in logs, 2007

14

Page 15: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

the Guangdong and Shangai provinces, while the Jiangsu province and the city ofTianjin host several open cities with ETDZs.

Figures 6 and 7 show the relationship between levels of income per capita andof export sophistication across provinces, in 1997 and 2007. As for countries above,the GDP per capita is a significant predictor of sophistication, but the correlationbetween the two is less significant; the elasticity of sophistication with respect toGDP per capita is also lower (about 13%, when it is about 31% across countries).

Shangai

Figure 5: Export sophistication levels (in constant 2000 dollars) across China’sprovinces, 1997

15

Page 16: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

Beijing

Tianjin

Hebe

Shan

Inne

Liaoning

Jilin Heil

Shangai

Jiangsu

Zhej

Anhu

Fuji

Jian

ShanHena

Hube

Huna

Guangdong

GuanHain

Chon

Sich

Guiz

Yunn

ShaaGans

Qing

Ning

Xinj

4000

5000

6000

7000

sop

hist

icat

ion

0 .2 .4 .6 .8 1real GDP per capita, 1997

®

Figure 6: sophistication and GDP levels by province, 1997

Figure 7: sophistication and GDP levels by province, 2007

BeijingTianjin

Hebe

Shan

Inne

LiaoningJilin

Heil

Shangai

Jiangsu

ZhejAnhu

Fuji

Jian

Shan

Hena

HubeHuna Guangdong

Guan

Hain

Chon

Sich

Guiz

Yunn

Shaa

GansQing

Ning

Xinj6000

7000

8000

9000

1000

011

000

soph

istic

atio

n

.5 1 1.5 2 2.5real GDP per capita, 2007

®

16

Page 17: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

020

4060

8010

0pe

rcen

t of m

ean

of s

hare

_shi

ptyp

e

1997 1998 1999 2000 2002 2003 2004 2005 2006 2007

Ordinary trade OthersProcessing

®

Figure 8: shares of trade types in total exportsNote: ‘Others’ group special types of flows such as border trade, barter trade, goods on lease,

Aid, etc.

The ordinary trade / processing trade gap

Is the high level of export sophistication of China only due to its processing industry?To assess this question, we start by looking at the respective shares of each type oftrade in total exports. This is shown on figure 8. Exports are shared in proportionsof about 40% / 60% between ordinary and processing trade; the proportion remainsfairly stable over the period.

Next, the export sophistication by type of trade is diplayed on graph 9. Weobserve an overall steep increase in the sophistication level of exports, consistentlywith other studies. A significantly higher level of sophistication is associated withassembly trade operations, and the gap with the ordinary type of trade seems fairlystable over the period. To give an example, the export bundle in the ordinary tradesector in 1997 had a sophistication level of about 5900$, which was similar to thatof Algeria; while the export structure in Assembly trade in China was more similarto that of Australia, at about 6800$. This partly confirms the picture of a ‘dual-ism’ between the two types of trade, which has been highlighted in the literature.However, we observe that the increase in sophistication level of exports is not at allrestricted to processing trade; on the contrary, this upgrading process has been more

17

Page 18: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

important in ordinary trade activities, as this level has been up by 39.5% between1997 and 2007. In assembly trade the relative increase is of 35%.

6000

7000

8000

9000

1000

0S

ophi

stic

atio

n, U

S 2

000

$

1995 2000 2005 2010year

Ordinary trade Processing trade

®

Figure 9: soph by shiptype

1.3 Does sophistication matter for growth?

In Hausman, Hwang and Rodrik(2006), the measure of sophistication is taken as aproxy for the level of productivity of the goods that a country is capable of export-ing; the argument for focusing on this measure is that a higher export sophisticationeventually enables a country to achieve higher growth. These authors test this pre-diction by running a classical growth regression on a panel of countries, adding thelevel of sophistication of exports (defined as detailed above) as a predictor of subse-quent growth. They find a significant positive effect of the initial log sophisticationof countries on growth of GDP per capita.

In order to assess the direction of causality, they then instrument countries’sophistication on their population size and land area (the argument for this choiceof instruments is to be found in the theoretical model, where the productivity level ofa country’s exports is determined by the number of investors starting new projects).Their IV regressions yield significant and larger coefficients on the sophisticationvariable.

18

Page 19: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

We started by testing the same specification as in Hausman et. al. on China’sprovinces (results are shown in table 1): regressing current real GDP per capita onthe previous year’s value, the stock of human capital and the sophistication level ofthe province’s exports, we find a positive and significant effect of the sophisticationlevel on subsequent growth rates.

We use two alternative variables as proxies for the stock of human capital : oneis the ratio of tertiary education graduates over the population (computed usingdata from China’s Statistical Yearbook, see data section below). The other is ameasure of the average number of years of education in the population, computedby following the method proposed by Barro and Lee (1996), which uses the stocks ofgraduates from several academic levels, and the average number of years correspond-ing to each of these levels in the country. This latter measure has the advantage ofaggregating the different levels of educational attainment in the poulation into onenumber; while the former focuses more on the highest recorded level, which is ofmore direct interest to us. In our growth regressions, both measures have positiveand significant coefficients.

Next, we included additional regressors in order to control for potential determi-nants of growth at the provincial level in China, and to reduce the risk of an omittedvariables bias on the measured effect of the sophistication.

The investment rate is found to have no significant effect on growth (or a negativeeffect in the last specification); which could indicate a tendency to over-investmentin physical capital in chinese provinces; the FDI ratio (ratio of FDI inflow to GDP)is also found to have no significant effect.

In order to control for differences in the internal reform process, we includethe share of investment in fixed assets done by state-owned enterprises in totalinvestment value of provinces. We use this variable to proxy for the weight of stateenterprises in the economy of the province: it is often argued that these enterpriseswere usually less dynamic, and benefited less from technical progress and from theopeness to market mechanisms. More broadly, it is taken as an indicator of the macroenvironment of the local economy (see for example Boyreau-Debray (2004)). It seemsimportant to control for differences in the advancement of market liberalization inthe provinces, as this process can be expected to influence growth both directly, and

19

Page 20: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

through the changing production and export structure; in this case omitting thisvariable would lead to a bias on the measured effect of export sophistication.

As expected we find a negative and significant coefficient measured on this vari-able. We also include a proxy for the development of infrastucture equipment in theprovinces: we use the ratio of railway and road network length to the surface areaof the province; we follow in this choice Demurger(2001) .

Finally in the last specification, we include year fixed-effects, in order to controlfor time-varying factors common to provinces. In the case of growth regressions, it isimportant to control for these factors, as it is likely that many factors at the nationallevel can impact the growth of provinces and the export structure simultaneously.Policy reforms implemented at the national level (e.g. the WTO accession in 2001),or changes in the pattern of world trade, are such processes which the inclusion oftime fixed-effects enables us to control for.

Including year fixed-effects significantly affects the coefficients on all variables,but the coefficient on export sophistication remains positive and significant. Itslower value indicates that the inclusion of time fixed effects is likely to control forfactors affecting both growth and export sophistication positively.

It may be doubtful at this stage to give the observed correlation between so-phistication and growth a causal interpretation, as it seems plausible that reversecausation from growth to the sophistication of exports can occur. We did not finda convincing way to instrument the sophistication variable: contrary to the cross-country regressions in Hausman et. al. (2006), the population size and land areavariables are not predictors of export sophistication at the provincial level in China,possibly because labour mobility and the transfer of technology across provinces aremuch less restricted than across countries. As a result, at this stage we have to seeour results only as evidence of correlation between the two variables, rather than ofa truly causal link.

20

Page 21: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

Table1:

Pan

elgrow

thregression

s,19

97-200

7.Dep

endent

variab

le:logreal

gdppe

rcap.

(1)

(2)

(3)

(4)

logreal

gdppc

logreal

gdppc

logreal

gdppc

logreal

gdppc

logof

lagged

real

gdppe

rcap.

0.97

6a0.94

2a0.88

4a0.59

4a(0.024

)(0.027

)(0.033

)(0.051

)

logsoph

istication

0.14

7a0.18

1a0.20

8a0.10

4b(0.051

)(0.050

)(0.051

)(0.049

)

logstockof

tertiary

educationgrad

uates,

0.049a

ratioto

popu

lation

over

15(0.017

)

logaverag

eyearsof

education

0.39

8a0.33

1a0.09

0(0.107

)(0.106

)(0.143

)

investmentrate

-0.012

-0.058

b

(0.027

)(0.026

)

FDI,ratioto

GDP

-0.010

0.00

2(0.009

)(0.009

)

shareof

Stateenterprisesinvestment

-0.259

a-0.083

tototalinv

estm

ent

(0.062

)(0.060

)

loginfrastructure

-0.001

-0.006

(0.022

)(0.025

)

Con

stan

t-1.073

b-2.279

a-2.367

a-1.110

b

(0.469

)(0.523

)(0.519

)(0.537

)prov

ince

fixed-effe

tsyes

yes

yes

yes

year

fixed-effe

tsno

nono

yes

Observation

s33

033

033

033

0R

20.96

10.96

20.96

40.97

2Stan

dard

errors

inpa

rentheses

cp<

0.1,

bp<

0.05,a

p<0.01

Note:

Allregression

sinclud

eprovince

fixed-effe

ctsforChina

’s30

provinces(T

ibet

exclud

edforlack

ofda

ta).

RealGDP

percapita

isob

tained

bydefla

ting

nominal

prov

incial

GDP

bythena

tion

alconsum

erpriceindex

(CPI),then

dividing

byprovincial

popu

lation

.Infrastructure:

density

oftran

sportation

(highw

ay+railw

ay)netw

ork

21

Page 22: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

2 Determinants of sophistication; the importance of

importing foreign technology

What should be the basic determinants of the sophistication level of a country’sexports? In the model of Hausman, Hwang and Rodrik(2006), one expects the pro-ductivity level, or sophistication, of a country’s exports to be determined both by‘fundamentals’, that is, by the endowments of a country in physical and humancapital, which determine the range of goods which it can potentially produce; andby the resources invested in innovation and cost-discovery. In the empirical part ofthis paper, the authors go on testing this prediction by regressing, across countries,the export sophistication of countries on GDP levels, stocks of human capital, andthe size of the population, which in their view proxies for the number of ‘innovators’starting new projects. They find positive and significant effects of all these variableson their measure of sophistication.

However, one plausibly important determinant of export sophistication that hasnot been empirically tested in the existing literature is operating through technologyand information spillovers from foreign producers. Evidence on the Chinese policy oftechnological development suggests that this determinant is particularly importantin the case of China: for example, Rodrik(2006) suggests that the remarkable level ofsophistication of chinese exports may be imputed to the policies of technology importimplemented in the country since the early 1990s, which were in particular aimingat attracting foreign investors and giving them incentives for sharing informationwith chinese partners :

“China’s openness to foreign investment and its willingness to createspecial economic zones where foreign producers could operate with goodinfrastructure and minimum hassle must therefore receive considerablecredit. But if China has welcomed foreign companies, it has always doneso with the objective of fostering domestic capabilities. To that end,China has used a number of policies to ensure that technology transferwould take place and strong domestic players would emerge. (...) For-eign investors were required to enter into joint ventures with domesticfirms (in mobile phones and in computers). Weak enforcement of intel-

22

Page 23: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

lectual protection laws enables domestic producers to reverse engineerand imitate foreign technologies with little fear of prosecution.”

The idea that opening to foreign investment is a suitable policy to foster thedomestic technological development of a country has been one of the rationale fornumerous developing countries to adopt such policies, as well as for multilateralagencies to encourage them to do so. In the case of China, it is generally admittedthat the policy toward foreign investment implemented since the early nineties hasbeen successful in upgrading the country’s technological capability. However, ingeneral evidence on the beneficial effects of incoming FDI on the host economy hasbeen subject to discussion. Moreover, most of this literature has focused either onan effect on growth, at a macro level; or, more specifically at a micro level, on aneffect on the productivity of firms. Few studies have proposed to identify precisely aparticular channel of the impact of FDI on growth, as we intend to do by examiningif the level of sophistication of exporting foreign firms has an influence on that ofdomestic firms.

In the following section, we briefly review some of the literature on the effectsof foreign investment on domestic firms. Next, we review the existing evidence forChina.

2.1 Empirical evidence on the impact of foreign investment

and spillover effects

At the cross-country level, numerous empirical studies have found positive effects ofFDI opening on growth (for example Agosin and Mayer 2000, Carkovic and Levine2002, Edison et al 2002). But the reliability of these results has been put in doubt(see Rodriguez and Rodrik 2001) for ignoring a possible reverse causation fromgrowth to openness, and also for possibly being driven by omitted variables havingan effect both on growth and on openess, such as institutional features of countries.

Studies adopting a sub-national approach have been more able to overcome someof the shortcomings mentioned above; in particular, this approach enables to bettercontrol for institutional features and their evolution, which would be expected to bemore similar within a country (see De Mello (1997) for a survey).

A number of studies have focused on the effects of FDI at the micro level, to

23

Page 24: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

identify an impact on the productivity of domestic firms or on their export perfor-mance. This approach has the advantage of being able to pin down more preciselythe spillover effects, apart from the effect of capital inflows accompanying foreigninvestment. Among these, evidence is mitigated. Aitken and Harrison (1999), usingplant-level panel data of Venezualian firms, find that foreign investment negativelyaffects the productivity of domestic-owned plants, which they interpret as a marketcrowding-out effect. Aitken Hanson and Harrison (1997) study the export behaviorof domestic firms in Mexico, and find evidence that the proximity of multinationalfirms increases the probability for domestic firms to access export markets; their viewis that foreign firms provide neighbouring firms information about foreign marketsand thus reduce the entry costs for firms willing to start exporting. But Bernard andJensen (2004) find no evidence of export spillovers on a panel of US manufacturingfirms, be they region-specific, industry-specific or both.

Javorcik (2004) has argued that the failure of some of the literature to findpositive intraindustry spillover effects could be due to the fact that these effectscould rather occur across industries, specifically through backward linkages (that is,local producers benefiting from spillovers from the presence of foreign affiliates indownstream sectors). Testing this hypothesis on a firm-level data from Lituania,she finds evidence of such spillovers but from shared domestic and foreign ownershipplants (joint ventures) only.

In the case of China, the coincidence of impressive growth rates in the countryand of a surge in foreign direct investment since the early 1990s (China becomes thesecond recipient of FDI in the world in 1994) has strongly pushed to conclude thatthis attractiveness to foreign investment had played a role in fostering growth. Sev-eral studies have attempted to pin down a causal link. Demurger (2001) uses paneldata on chinese provinces over the 1985-1998 period to estimate an extended growthmodel, and finds that the amount of foreign investment received by a province has apositive and significant effect on subsequent growth of GDP per capita, controllingfor the investment rates, the stock of human capital, urbanization and transportinfrastructure. Berthelemy and Demurger (2000) build an expanded model of en-dogenous growth to account for the endogeneity of FDI and the possible reversecausation from growth to foreign investment; then they use a simultaneous equationapproach to estimate the model on a panel of China’s provinces; they thus confirm

24

Page 25: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

the important role played by foreign investment in provincial growth in China.

Thus, among the abundant literature on spillover effects from FDI one can findtwo general approaches, one focusing on the effect on growth at the macro level, theother investigating effects at the micro level. There is little empirical work assessingthe specific effect of foreign invested enterprises on the export sophistication ofthe host economy, although the literature on export sophistication has highlightedthe important role that such spillovers can play in determining a country’s exportspecialization. Two recent exceptions need to be mentioned : a recent article byWang and Wei(2008) studies the impact of the share of foreign invested enterprises,and of processing activities, on export sophistication at the city level. They findno significant impact, and that human capital and export-promoting policies areinstead, key determinants of sophistication. Using a somewhat similar approach,Xu and Lu (2008) also look at the impact of foreign firms on export sophisticationacross industries. However, these two articles do not specifically focus on spillovers,as we try to do in the following.

3 Empirical estimation : the importance of foreign

spillovers as a determinant of domestic sophisti-

cation

3.1 Data

Our primary source for this study is the Chinese customs’ export and import data(we essentially used export data), which report all trade flows in and out of China,by value, quantity (and unit), trade partner, and province of origin or of destina-tion. The products are initially classified in HS-8 codes, which we aggregated toHS-6 (5017 product lines in the 1996 classification). Since we are not working atthe product level but using aggregate measures, we found that this level of disag-gregation was enough to distinguish product properties such as technology level,framgmentability, differentiability, etc., whithout keeping a too high level of detail.

Reported are also the type of ownership of the exporting firm; the type of Ship-

25

Page 26: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

ment : ordinary or processing; and the transport mode (by air, sea, rail..), which wedid not use in our study.

• The ownership type distinguishes between Private firms, State-owned, Collec-tive, Foreign, Joint Ventures and Others. Among domestic firms, State-ownedentities were the most common until the early 1990s, while Privately-ownedfirms became more common as market reforms were progressing. Among non-domestic firms, the “joint venture” type includes several contractual forms,including equity and non-equity joint ventures, and joint cooperatives. Equityjoint venture was the most current form of foreign investment in China untilthe mid-1990s, as it was strongly encouraged by Chinese authorities, aimingat facilitating information and technology sharing between partners. As the“marriages” between foreign and Chinese managers were not always successful,Chinese regulations became more accomodating and wholly foreign-owned en-terprises became the first form of foreign investment, accounting for two thirdsof total fdi inflows in 2004. This evolution can be seen in the growing share ofexports by Foreign firms relative to Joint ventures in our data.

• The type of Shipment : Processing trade includes all trade flows by firmsoperating in the assembly sector, that is, importing inputs to process themin China and re-export the finished products (these producers benefit froma preferential tax regime on imported inputs). Ordinary trade is for normalexports, while a third (“Others”) category groups other flows such as aid, bordertrade and consignment.

The data report two entries for the province of origin of exports. One is theproduction location, the other is the trade location. The two are equivalent for mostflows, but they differ in the cases where a producer located in one province partnerswith a exporting firm elsewhere, whose purpose is to handle the shipment of theproducts to their destination. When comparing the aggregated flows by productionand by “trade” location, we observe that a small share of exports transit throughanother province (often Beijing or Shangai) before being exported.

For our purposes, we would prefer to use the location of production, as we want torelate these flows to determinants of specialization at the provincial level; however,only the trade location data are reported for the whole period; therefore in order

26

Page 27: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

to have a consistent panel, we used the trade location. In average, the distinctioncreates a difference of less than 3% in provinces’ export flows.

In order to compute the sophistication levels associated to export baskets, weneed world trade data for the period; we use the BACI data to this purpose, a CEPIIworld database for international trade analysis at the product-level, constructedusing COMTRADE original data, as detailed in Gaulier and Zignago (2008).

Merging these trade data with the Chinese customs data raised one issue withproduct classifications: the Harmonized system (HS) classification updates regu-larly, and did so in 1996, 2002 and 2006; when it happens, some product lines arebeing added by splitting of former lines into two or more new lines. Chinese customshave been adopting these updates; but the BACI data, which we use for world tradeflows, are using the 1992 classification for all years. As we needed to merge these twosources, we had to convert Chinese customs data into 1992 product codes; thus someflows registered under different product codes are merged into one same 1992 code,causing the loss of some detail. However this additional level of detail in the productcategories is likely to have a marginal effect on our measures of sophistication.

Once our world trade data made consistent with our Chinese customs data, weuse the former to compute the average income level associated to each export prod-uct at the HS-6 level, according to the formula shown in the first section (equation1); we obtain the data on world countries’ levels of real GDP per capita from theWorld Development Indicators database (World Bank). We are thus able to com-pute these average income levels for 5017 products (in 1997), using trade flows of192 countries. Then, the sophistication measure associated to a given export basketis obtained as the weighted average of each exported product’s associated incomelevel (equation 2).

In the first section, we explained that the measure of sophistication we use issubject to possible biases, in particular if some resource-based or primary productsare an important component of some rich countries’ exports : in this case theseproducts would wrongly appear as ‘sophisticated’. To adress this issue we tried tocompute sophistication levels on the subset of products excluding resource-basedand primary products; we found that it did not affect importantly the results of our

27

Page 28: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

regressions.

For data on economic indicators at the provincial level, which we use in our re-gressions, we relied on the China’s Statistical Yearbooks and Provincial Yearbooks(computed by the National Bureau of Statistics) as main sources. Real GDP wasdeflated using the consumer price index (CPI). Other variables expressed in realterms were deflated using the index of GDP (base 1990) given in the StatisticalYearbook.

Finally, to construct instrumental variables we used data on patents granted bycountry of origin, as an indicator of innovative activity; these data are computed bythe World Intellectual Property organization (WIPO) and available online. Thesedata refer to the country of origin of the first-named applicant or assignee, whichenable us to capture the innovative activity of firms of given nationality, includingforeign invested enterprises.

3.2 Facts and figures on China’s domestic and foreign ex-

porters

Before testing the hypothesis of spillovers from foreign to domestic enterprises, westart by examining the export characteristics of the different types of firms, in termsof their relative importance in total export activity of China, and in terms of so-phistication levels.

How are total exports shared between the six types of firms ownership we iden-tify? The evolution of these relative shares over the period 1997-2007 are shown ongraph 10 and table 2.

We observe a remarkable growth of the private sector over the period, its shareof export activity moving from zero in 1997 to 20.5% in 2007. This is accompaniedby a comparable decrease in the share of State-owned enterprises, from 56.2% to18.5%. Part of this evolution is attributable to the progress of market reform andof privatization (rather named “restructuring”), which accellerated in the mid-1990s(following the 15th Communist Party Congress in Sept. 1997 in particular; see B.

28

Page 29: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

Naughton 4.6 for details).Another feature is the growth of the share of fully-foreign firms relatively to joint

ventures; the former account for more than two-thirds of total non-domestic exportactivity in 2007. This is consistent with the evolution of Chinese policy towardcontractual forms of foreign investment, as explained in the data section above.

020

4060

8010

0pe

rcen

tage

of t

otal

exp

orts

1997 1998 1999 2000 2002 2003 2004 2005 2006 2007

Collective ForeignJV OtherPrivate State

®

Figure 10: shares of firm types in exports

Table 2: Shares of total exports, by firm ownership statusCollective Foreign JV Other Private State

year1997 2.49% 17.13% 23.95% 0.23% 0.04% 56.16%1998 2.94% 19.97% 24.13% 0.25% 0.08% 52.62%1999 3.49% 22.26% 23.25% 0.18% 0.33% 50.49%2000 4.25% 23.79% 24.21% 0.15% 0.96% 46.65%2002 5.80% 29.60% 22.70% 0.03% 4.24% 37.63%2003 5.75% 33.38% 21.55% 0.02% 7.95% 31.35%2004 5.36% 36.15% 20.99% 0.02% 11.67% 25.81%2005 4.80% 38.53% 19.91% 0.01% 14.77% 21.98%2006 4.25% 39.50% 18.74% 0.01% 17.84% 19.66%2007 3.86% 39.32% 17.83% 0.03% 20.50% 18.46%

These figures are likely to mask differences in the type of trade: ordinary orprocessing trade. To examine the structure of trade by type of firms more closely,

29

Page 30: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

we look at the shares of each ownership type within each type of trade. This isshown in figures 11 and 12. We observe that the trade system remains highly“dualistic” : foreign firms and joint ventures are mainly specialized in assemblytrade (with a significant, but fast decreasing share accruing to State enterprises).In “ordinary” trade, one observes large movements in the distribution across firmtypes : in the beginning of the period, the major part of this type of export isdone by domestic firms, and mainly by State enterprises. Over the 10-year periodconsidered, foreign firms and joint ventures have gained a significant share of thistype of trade, reaching more than 20% in 2007; while among domestic firms, the“boom” of private enterprises is particularly obvious in this category of exports,moving from 0 to more than 40% of total exports, and thus reducing the shareaccruing to State-owned enterprises.

020

4060

8010

0pe

rcen

tage

of p

roce

ssin

g ex

port

s

1997 1998 1999 2000 2002 2003 2004 2005 2006 2007

Collective ForeignJV OtherPrivate State

®

Figure 11: shares of company types in assembly trade

We now turn to the export structure of each type of firms, starting by lookingat the sophistication levels of exports. On figures 13and 14are displayed the sophis-tication levels of the different firm types, in Ordinary trade and in Processing traderespectively.

These graphs point to a greater homogeneity across firm types in ordinary trade,than in assembly trade. The gap between non-domestic firms (fully foreign-ownedand joint venture firms) and domestic firms, in term of export sophistication, is

30

Page 31: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

020

4060

8010

0pe

rcen

tage

of o

rdin

ary

trad

e

1997 1998 1999 2000 2002 2003 2004 2005 2006 2007

Collective ForeignJV OtherPrivate State

®

Figure 12: shares of company types in ordinary trade

5000

6000

7000

8000

9000

soph

istic

atio

n, U

S 2

000

$

1995 2000 2005 2010year

Collective ForeignPrivate StateJV

Ordinary trade

®

Figure 13: sophistication levels by firm types, ordinary trade

31

Page 32: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

6000

7000

8000

9000

1000

0so

phis

ticat

ion,

US

200

0 $

1995 2000 2005 2010year

Collective ForeignPrivate StateJV

Processing trade

®

Figure 14: sophistication levels by firm types, processing trade

more marked in assembly trade; morevover, in the ordinary trade sector, foreignfirms and joint ventures do not always exhibit higher sophistication levels than do-mestic firms. This seems to indicate that the higher technology level of processedexports is brought in by foreign managers and/or engineers, not just embedded insophisticated imported inputs. Foreign capital keeps an advantage in producing so-phisticated products in China, which domestic firms can not really achieve, even ifthey have access to the same imported inputs, and benefit from the same incentivepolicies toward assembly trade.

This is confirmed when looking more closely at the export structures acrossthe different types of firms and of trade, using the product categories defined byLall(2000a) and introduced in section 1. To do this we look at the shares of eachcategory of exports - High-tech, Medium-tech, etc. - in the total value of exports of agiven type of firm. This alternative measure has the advantage of pinning down moreprecisely the specialization patterns of each type of exporter, while the sophisticationgive an aggregated measure. It also enables us to measure the intrinsic evolution ofan export structure independently of the evolution of world trade patterns, contraryto the sophistication measure, which is referring to world trade and GDP data.

The corresponding graphs are shown in figures 15, 16, 17, 18, 19 and 20.

32

Page 33: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

.05

.1.1

5.2

shar

e of

tota

l exp

ort v

alue

1995 2000 2005 2010year

Collective ForeignPrivate StateJV

Share of HT products, Ordinary trade

®

Figure 15: Shares of High-Tech products in exports across firm types, Ordinarytrade

.1.1

5.2

.25

.3sh

are

of to

tal e

xpor

t val

ue

1995 2000 2005 2010year

Collective ForeignPrivate StateJV

share of MT products, Ordinary trade

®

Figure 16: Shares of Medium-Tech products in exports across firm types, Ordinarytrade

33

Page 34: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

.35

.4.4

5.5

.55

.6sh

are

of to

tal e

xpor

t val

ue

1995 2000 2005 2010year

Collective ForeignPrivate StateJV

Share of LT products, Ordinary trade

®

Figure 17: Shares of Low-Tech products in exports across firm types, Ordinary trade

0.2

.4.6

.8sh

are

of to

tal e

xpor

t val

ue

1995 2000 2005 2010year

Collective ForeignPrivate StateJV

Share of HT products, Processing trade

®

Figure 18: Shares of High-Tech products in exports across firm types, Processingtrade

34

Page 35: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

.15

.2.2

5.3

.35

.4sh

are

of to

tal e

xpor

t val

ue

1995 2000 2005 2010year

Collective ForeignPrivate StateJV

Share of MT products, Processing trade

®

Figure 19: Shares of Medium-Tech products in exports across firm types, Processingtrade

.1.2

.3.4

.5.6

shar

e of

tota

l exp

ort v

alue

1995 2000 2005 2010year

Collective ForeignPrivate StateJV

Share of LT products, Processing trade

®

Figure 20: Shares of Low-Tech products in exports across firm types, Processingtrade

35

Page 36: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

Looking at these graphs confirms the fact observed above : there is greater ho-mogeneity in export structure across firm types, in ordinary trade, than in assemblytrade. For example, in the ordinary trade sector, the shares of Medium-Tech prod-ucts in the export baskets of each type of firm are all included between 8% and 15%in 1997, and between 23% and 28% at the end of the period, in 2007. Similarly,High-tech products represent in 1997 between 3 and 6% of the exports of each typeof firm, and between 7% and 21% in 2007. We also notice that foreign firms are notthe ones most specialized in High-tech or Medium-tech products.

On the contrary, the gap between foreign capital firms and domestic firms is muchmore marked in the processing sector, as these types of firms exhibit completelydifferent export patterns : in 2007, the share of High-tech products is over 40 % forJoint ventures and over 60% for fully foreign-owned firms; it lies between 16 and25% among domestic firms. Conversely, Low-tech products represent less than 20%of the exports of non-domestic firms, while it counts for 23 to 41% of the exports ofdomestic firms.

These figures confirm that the processing and ordinary trade activities representtwo very different sectors in China. In the first, it seems that domestic and non-domestic firms operate in different markets, export different kinds of products, andthe evolution of their export structure seems rather disconnected. On the other hand,the ordinary trade sector appears as a more homogenous entity, with foreign anddomestic firms exhibiting similar patterns of trade, and the evolution of this exportstructure over time - a moderate increase in the share of High-tech products, a moremarked increase of the share of medium-tech products and a decrease of the weightof low-tech exports, appears to be common to all types of firms operating in thissector. This seems to point to the hypothesis that potential spillovers from foreignto domestic firms would more likely occur in the ordinary trade, where these firmsexport similar products. In practice, our results confirmed this, and we essentiallyfocused on ordinary trade in our specifications.

One important point in our estimation strategy is to decide which type of firmswe want to focus on. When testing the hypothesis of spillovers from foreign capitalfirms to domestic firms, there are several possibilities ; it can be that only a giventype of firms with foreign capital generate these spillovers. For instance, one couldthink that only Joint ventures would do so, because of their “mixed” capital struc-

36

Page 37: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

ture, which necessarily brings chinese and non-chinese to work closely together inthe firm and thus, facilitates information and technology sharing. On the domesticside, the same question arises of whether some type of firms would be more exposedto spillovers from foreign enterprises than others : as we already noted, State-ownedfirms are generally considered to have been less productive and less open to tech-nology transfers, because they respond in general less to market incentives. Privatefirms could be thought to be the best candidate for benefiting from spillovers, be-cause conversely they should be more reactive, and more eager to access foreigntechnology. Alternatively, one could think that the type of ownership is not an im-portant determinant of the potential to generate, or to benefit from spillovers. Therelative homogeneity of sophistication levels across domestic firms of different typeson one side, and across non-domestic firms on the other, brings some support to thisview. Moreover, one can think that there is significant churning across firm typesover the period, in the sense that some firms could have changed their declared type,without significant changes in their activity (e.g., State firms becoming private, asthe restrictions on private enterprise loosened). This brings support to the choiceof aggregating the firms into two types, domestic and non-domestic. In practice,this latter approach proved more effective, therefore we use aggregated firm typesin most specifications.

3.3 Results

3.3.1 Does the sophistication of foreign firms impact that of domesticfirms?

In order to identify an impact of foreign sophistication on domestic sophistication,we need to control for possible determinants of these two variables. The use ofprovince fixed effects in the panel regressions enables us to control for any time-invariant specificities of the provinces, such as differences in factor endowments, andin institutional environment. In addition, since we want to examine the correlationbetween our two variables of interest across time, we need to control for factorsvarying in the time window considered, which have a potential impact on bothvariables.

Thus our baseline specification is regressing the sophistication of domestic firms

37

Page 38: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

in ordinary trade, on the physical and human capital stocks, and the sophisticationof nondomestic firms in ordinary trade. The results are shown in table 3. We thencontrol for the advancement of market reform at the provincial level, since this factoris likely to influence the export performance of both domestic and incoming foreignfirms; to this purpose we use the share of State investment in total investment,similarly as in the growth regressions. We also control for the cumulated inflow offdi; as discussed for instance in Demurger(2000), fdi attraction has been found to be amore important determinant of the economic performance of chinese provinces thanexport promotion, measured by the openess ratio. Moreover, including this variableenables us to control for the relative size of foreign capital in the economy of theprovince, while our variable of interest, the sophistication of non-domestic firms,focuses on the export structure of these firms, independently of their size. Finallywe control for the development of infrastructure equipment, proxied by transportnetwork density as in the former regressions.

In the third specification we include year fixed-effects to control for any factorexhibiting a trend common to chinese provinces. The spillover effect remains signif-icant and the estimated coefficient is slightly lower than in the former regressions.

Finally, we investigate whether the impact of nondomestic firms’ sophisticationon domestic firms also appears in the processing trade sector. We find no significanteffect in this sector, which seems consistent with the previous indications of fewerspillover potential in these export activities.

3.3.2 Alternative measures of export sophistication

In order to get a more precise idea of the channels through which foreign firms andjoint ventures influence the export activity of chinese domestic firms, we now turnto using alternative measures of the export structure. We use the categorizationof products into technology levels to identify the shifts in the export specializationpatterns. Thus we will use the share of a product category (e.g., “High-tech”) in thetotal exports of a given group of firms, similarly as what we did in the former section.As noted previously, the ranking of these product categories by sophistication levelfollow the expected order, so that the shares can be taken as an alternative measureof sophistication, in the sense that an increase in the share of High-tech products inthe export structure mechanically corresponds to an increase in sophistication. The

38

Page 39: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

corresponding regressions are shown on table 4.We start by looking at the ordinary trade sector. Regressing the share of High-

tech products in the exports of domestic firms, on the same indicator for non-domestic firms, and using the same controls as in the former regressions, we finda positive and significant effect of non-domestic specialization in high technologysectors on the pattern of specialization of domestic firms. We also include an ad-ditional control for the R&D effort of firms, using the registered number of patentsgranted as proxy. We find that this variable is also positively correlated with aspecialization of the domestic sector into high-tech products. The same exerciseconducted on the category of Medium-tech products indicates a more importantspillover effect in this range of products; it also shows that the specilization intothis sector is less dependent on the human capital stock and on the R&D effort, andmore on the physical capital stock.

These results seem consistent with previous evidence on the evolution of spe-cialization patterns in China. The Medium-tech products are the category whichincreased the most in the overall export pattern of ordinary trade in China. Thisincrease has also been documented in the literature (see Amiti and Freund (2008)).It seems plausible that this range of products is where domestic producers can ben-efit the most from the observing and partnering with foreign producers, as thesetypes of products require less genuine innovation (as aimed at by R&D efforts) butmore adaptation of existing technology through ‘incremental innovation’, using theterminology of Puga and Trefler (2007).

In the next two specifications, we used an alternative measure of specializationinto medium-tech and high-tech production: the density of products, defined as thenumber of used lines (that is, of products for which a non-zero export flow is ob-served) in a given category of products, over the total number of existing productlines (observed in world trade flows of the same year) in that category. Using thismeasure enables us to identify whether the spillover effect is impacting the intensiveor the extensive margin of trade: in the first case, the increased specialization into,say, high-tech goods is occuring by increasing the exported volumes of the existingproducts; while in the second, it is through the export of new products. The ob-served significant correlation between the density of products of non-domestic and

39

Page 40: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

domestic producers seems to indicate that at least part of the movement is occuringat the extensive margin, suggesting that the introduction of new products by foreignproducers facilitates the entry of domestic producers into their export. However, afiner analysis, using firm-level data, would be necessary in order to properly identifysuch an effect.

Finally, in the last two specifications, we check whether similar spillovers areoccuring in the processing trade sector. We use the same specifications as in thefirst two columns, using the measured shares of exports computed for assembly tradeproducers, instead of ordinary trade. We find no significant spillover effect from non-domestic to domestic producers in this case, confirming again that the potential forspillovers in this sector is less important.

3.3.3 Instrumental variables

The specifications we have used so far, to identify the existence of spillovers fromnon-domestic to domestic producers, are subject to a potential risk of omitted vari-ables bias. We do control for a number of potential determinants of both domesticand foreign sophistication, such as physical and human capital stocks, advancementof market reform, availability of infrastructure, and R&D effort, as well as time-varying variables common to all provinces (such as policy reforms at the nationalscale, e.g. WTO accession), which year fixed-effects are accounting for, and time-fixed variables specific to provinces, which province fixed-effects are accounting for.Nevertheless, it can be argued that some important determinants are being omit-ted, the time variation of which could impact the export structure of domestic andforeign exporters simultaneously; one could think for example of industrial policiesbeing implemented at the provincial level, aiming at encouraging the technologicalupgrading of both domestic and foreign producers, through incentive policies suchas fiscal instruments. If such variables play an important role and are not accountedfor by the set of controls we are using, then this would lead us to wrongly attributethe observed correlation between foreign and domestic sophistication to a causal linkfrom one to the other. In order to assess the causality of this relationship, we thusused instrumental variables in our main specification .

40

Page 41: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

To instrument our potentially endogenous variable, the sophistication of non-domestic exporters, we need to find a variable impacting the export structure offoreign producers in China, without possibly having any direct effect on that ofdomestic producers. One can think that variables capturing the technological levelof the countries of origin of foreign producers in China could play this role: this ideais based on the assumption that foreign investors in China are bringing with thempart of the technological know-how available in their place of origin.

We find two measures which can play this role: one is precisely the export so-phistication of countries other than China: using world trade data, we can computethe average income level (sophistication) associated to the export baskets of everyexporter, in the same way as we did for China. The second measure of the technolog-ical level of countries we used is more closely related to the innovation and researchand development effort of firms of a given origin: using the WIPO data base, wecomputed the total number of registered patents by nationals of each country, foreach year in our period of interest.

Then, we need to construct variables accounting for the variation, across provinces,of the composition of total FDI by nationality: the idea being that the export so-phistication of foreign producers in a province will reflect the sophistication level ofthe countries which account for the major part of FDI in this province. Because welacked information on the origin of FDI by province for each year of the period, weused instead the distance from provinces to countries of origin of FDI as proxy; webase this solution on the hypothesis that the relative importance of a given countryof origin in the total FDI inflow into a province is inversely related to the distancebetween them, similarly as in gravity equations for trade flows.

Thus, we constructed our instruments as the following :

IV 1it =∑

j

EXPYjt

distij(3)

IV 2it =∑

j

patentsjt

distij(4)

In the above formulas, the sum is over countries of origin of foreign investmentin Chinese provinces. As previously, EXPYjt denotes the sophistication level ofcountry j’s exports, at year t. distij denotes the geodesic distance from country

41

Page 42: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

j to province i; patentsjt is the total number of registered patents by nationals ofcountry j at year t.

Finally, we had to choose on which set of countries to compute these weightedsums. We first tried to use all countries for which data are available; but with thiswe obtained weak instruments, possibly because a number of countries contribute avery small part to foreign investment, or do not contribute at all. Thus, we restrictedinstead to a small list of Asian countries which contribute importantly to FDI inChina: these are Hong-Kong, Singapore, Japan, South Korea. According to the dataavailable to us on the composition of FDI inflows by province (for the year 1997),these four countries regularly appear among the five most important contributorsto total foreign investment in provinces. An additional advantage of using asiancountries is that they offer greater variation in distances across provinces, so thatthe resulting instrumental variable is more likely to capture differences in the timevariation of foreign sophistication between the provinces.

Results of the two-stage least squares estimations are shown on table 5. Inthe first specification, the two instruments used are computed as in equation 3,summing on Honk-Kong and Singapore for the first instrument, Japan and Koreafor the second. In the second specification (column 2), we use the instruments basedon the number of patents by country (equation 4), computed on the same pairs ofcountries. We use the same controls as in our baseline OLS specification displayedon table 3. In the first stage, the excluded instruments are found jointly significant.In the second stage, the Hansen J-test indicates that the hypothesis of exogeneity ofthe instruments cannot be rejected. The F-statistics obtained in the Cragg-Donaldtests of weak instruments are above the 15% critical value only, which does not ruleout the possibility of weak instruments.

On the bottom rows, we report the results of the Davidson-McKinnon tests ofthe endogeneity of our non-domestic sophistication variable, which indicate thatthe null hypothesis of exogeneity cannot be rejected. Thus, we rely on our OLS es-timates as being more efficient than the estimates from the instrumented regressions.

42

Page 43: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

Conclusion

In this study we have examined the recent evolution of the sophistication and tech-nology levels of Chinese exports, using a chinese customs database detailing exportflows on the 1997-2007 period. These data enable us to break down export flows byfirm ownership type and by trade type, and thus to identify the recent evolutions inexport activity in China. Using a measure of sophistication introduced by Rodrik etal, as well as a classification of manufactured products by technology levels, proposedby S. Lall, we observed that the recent upgrading of China’s trade, which has beendocumented and discussed in numerous studies, cannot be only attributed to foreignproducers operating in China, nor to the assembly trade activities which representabout 60% of total exports. We confirm that partially or totally foreign-owned en-terprises are concentrated in the assembly trade sector; and that, by all measures,these producers are exporting goods of higher sophistication than domestic ones.Nonetheless, foreign producers also represent a non negligible, and growing, shareof ordinary (non-assembly) trade activity in China; and domestic producers havealso been able to significantly upgrade the technology level of their exports, enteringnew markets such as those of machinery and electronics. These observations tend toindicate that the “dualistic” picture of China’s trade, accroding to which foreign anddomestic producers would be operating in distant and unrelated export markets, theformer assembling high-tech goods, the latter being confined to low-tech production,is not, or no longer, realistic.

We then tried to test the hypothesis of export spillovers from foreign to do-mestic producers. The idea that the export performance of domestic firms couldbe positively impacted by the presence of foreign producers, and in particular,that the knowledge and know-how necessary to efficiently produce and export high-technology products, could spill over to domestic enterprises, has been at the heartof China’s opening policy toward foreign investment since the early 1980s. Thesophistication measure seems appropriate to study this effect, as it is designed toassociate to a given export bundle, an corresponding average level of development.Thus, the surprisingly high level of this measure for China’s exports seem to point

43

Page 44: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

to the ability of its producers to aquire and manage efficiently a set of technologicalcapabilities which are in general found in much more developed countries. The hy-pothesis that these capabilities have been ‘imported’ along with FDI originating inmore advance countries thus appears worth testing.

Our results confirm this pattern. Using fixed-effects panel regressions on China’sprovinces, and after controlling for a number of determinants of export sophistica-tion at the province level, we find a significant impact of the export sophisticationof non-domestic producers on that of domestic exporters. The use of Lall’s productcategories as alternative measures indicates that the effect is more important in thecategory of medium-technology goods, the sector in which the domestic producers’share of exports has been increasing the most.

Finally, in order to adress the risk of endogeneity in our regressions, we instru-mented our non-domestic sophistication variable, using weighted averages of thesophistication levels of countries of origin of FDI in China. Our tests of endogeneitythen indicate that the hypothesis of exogeneity of the non-domestic sophisticationvariable cannot be rejected. We thus conclude that our estimated coefficients fromthe OLS specification are more efficient, and rely primarily on these as estimates ofthe spillover effect.

Several extensions to this study could be proposed in order to identify moreprecisely the channels through which spillover effects are occuring. One could be touse alternative measures of sophistication, such as those which have been reviewedin the first section; or alternative product classifications, for instance focusing onthe factor content of exports. Another direction would be to assess more preciselywhether the spillovers are operating at the extensive or intensive margin of trade;but a finer level of analysis, at the product or firm-level, would be necessary toanswer this question.

44

Page 45: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

References

[1] Agosin, M. R. and R. Mayer (2000), ‘Foreign investment in Developing coun-tries : Does it grow in Domestic Investment?’, UN conference on trade anddevelopment, Discussion paper No. 146.

[2] Aitken, B. J. and A. Harrison (1999), ‘Do Domestic firms Benefit from DirectForeign Investment? Evidence from Venezuela’, American Economic Review,89, 3, 605-618.

[3] Aitken, B. J., G. Hanson and A. Harrison (1997), ‘Spillovers, foreign investmentand export behavior’, Journal of International Economics, 43, 103-132.

[4] Amiti, M. and C. Freund (2008),‘An Anatomy of China’s Export Growth’,NBER Policy Research working paper 4628.

[5] Barro, R. J. and J. W. Lee (1996), ‘International Measures of Schooling Yearsand Schooling Quality’, American Economic Review, 86, 218-23.

[6] Bernard, A. and J. B. Jensen (2004), ‘Why Do some Firms Export?’, Reviewof Economics and Statistics, 86(2) 561-569.

[7] Carkovic, M. and R. Levine (2002), ‘Does Foreign Direct Investment AccelerateEconomic Growth?’, Mimeo, University of Minnesota.

[8] Dean J., K. C. Fung and Z. Wang (2007): ‘Measuring Vertical Specializationin Chinese trade’, US International Trade Commision working paper.

[9] Demurger S. (2001): ‘Infrastructure Development and Economic Growth: anExplanation for Regional Disparities in China?’, Journal of Comparative Eco-nomics 29, 95-117.

[10] Edison, H. J., R. Levine, L. Ricci and T. Slok (2002), ‘International Finan-cial Integration and Economic Growth’, Journal of International Money andFinance, 21, 6, 749-76.

[11] Gaulier, G. and S. Zignago (2008), ‘BACI: A World Database of InternationalTrade at the Product level, 1995-2004 version’ , CEPII working paper.

45

Page 46: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

[12] Hering L. and S. Poncet (2009), ‘The Impact of Economic Geography on Wages:disentangling the Channels of Influence’, China Economic Review, 20, 1-14.

[13] Koenig, P., F. Mayneris and S. Poncet (2008), ‘Local Export Spillovers inFrance’, PSE Working Paper No. 2009-03.

[14] Lall S. (2000a), ‘The Technological Structure and Performance od DevelopingCountry Manufactured Exports, 1985-1998’, Oxford Development Studies, 28,3.

[15] Lall S. (2000b), ‘Skills, Competitiveness and Policy in Developing Countries’(2000), QEH Working Paper QEHWPS 46, University of Oxford.

[16] Lemoine, F. and D. Unal-Kesenci (2004), Assembly Trade and TechnologyTransfer : the case of China’, World Development, 32, 5, 829-850.

[17] Manova, K. and Zhang Z. (2008), ‘China’s Exporters and Importers: Firms,Products, and Trade Partners, Federal reserve Bank of San Francisco workingpaper 2008-28.

[18] de Mello L. R. (1997), ‘Foreign Direct Investment in Developing Countries: ASelective Survey’, Journal of Development Studies, 34, 1, 1-34.

[19] Naughton B. (2007), The Chinese Economy: Transitions and Growth. Cam-bridge, MA: MIT Press.

[20] Puga, D. and D. Trefler (2007), ‘Wake Up and Smell the Ginseng: InternationalTrade and the Rise of Incremental Innovation in Low-Wage Countries’, CentroStudi Luca d’Aglaino Development Studies Working Paper No. 222.

[21] Rodriguez, F. and D. Rodrik (2001), ‘Trade Policy and Economic Growth : aSceptic’s Guide to Cross-National Literature’, in B. Bernanke and K. Rogoff(eds.), National Bureau for Economic Research Macro Annual 2000, (Cam-bridge, MA: MIT Press).

[22] Rodrik D. (2006), ‘What is so special about China’s Exports?’, unpublishedpaper, Harvard University.

46

Page 47: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

[23] Schott P. (2006), ‘The Relative Sophistication of Chinese Exports’, NBERWorking paper, WP12173.

[24] Wang Z., Wei S.-J. (2008), ‘What accounts for the rising sophistication ofChina’s exports’, NBER Working paper 13771.

[25] Xu B. (2007), ‘Measuring China’s Export Sophistication’, China Europe Inter-national Business School (CEIBS) Working paper.

[26] Xu B., Lu J. (2008), ‘Foreign Direct Investment, Processing Trade, and theSophistication of China’s Exports’, China Europe International Business School(CEIBS) Working paper.

47

Page 48: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

Table3:

Pan

elregression

s-depe

ndentvariab

le:do

mesticsoph

istication

,1997-20

07(1)

(2)

(3)

(4)

domesticsoph

.,do

mesticsoph

.,do

mesticsoph

.,do

mesticsoph

.,Ordinarytrad

eOrdinarytrad

eOrdinarytrad

eProcessingtrad

elogcapitals

tock

percap.

0.18

8a0.08

9b-0.053

0.24

9(0.024

)(0.038

)(0.065

)(0.198

)

stockun

iversity

grad

uatesover

pop1

5+1.15

6a1.16

8a0.70

9c0.38

4(0.330

)(0.355

)(0.399

)(1.485

)

non-do

mesticsoph

istication

Ord.trad

e0.13

3b0.16

8a0.11

9b(0.062

)(0.042

)(0.043

)

ratioof

fdis

tock

tocapitals

tock

0.28

80.73

4b1.67

8(0.366

)(0.341

)(1.027

)

shareof

Stateinvestmentin

totalinv

estm

ent(-1)

-0.168

-0.052

0.21

5(0.121

)(0.102

)(0.410

)

infrastructure

(ratio

topo

p.)(-1)

0.00

6a0.00

5b-0.004

(0.002

)(0.002

)(0.005

)

non-do

mesticsoph

istication

,Processingtrad

e0.102

(0.155

)

Con

stan

t7.94

3a7.35

7a7.32

3a8.04

8a(0.575

)(0.431

)(0.510

)(1.757

)prov

ince

fixed

effects

yes

yes

yes

yes

year

fixed

effects

nono

yes

yes

Observation

s33

033

033

032

1R

20.58

00.63

70.68

80.35

4Heteroskeda

stic-con

sistentstan

dard

errors

inpa

rentheses

cp<

0.1,

bp<

0.05,a

p<0.01

Stan

dard

errors

arecorrectedforclustering

attheprovince

level

Note:

allvariab

lesin

logs.Allregression

sinclud

eprovince

fixed-effe

cts.

"‘Dom

esticsoph

istication

"’:soph

istication

compu

tedon

allexpo

rtsby

domestic

firms,

includ

ingState-ow

ned,

privatean

dcollectivefirms.

Non

-dom

esticsoph

istication

=samethingon

foreignfirms+

jointventures."‘Infrastruc-

ture"’

=leng

thof

railw

ayan

dhigh

way

netw

ork,

ratioto

popu

lation

.Cap

ital

stock:

compu

tedas

thecumulated

sum

ofinvestmentflo

wsin

previous

period

s,usingade

preciation

rate

of5%

.

Page 49: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

Table4:

Pan

elregression

s-shares

ofHightech,M

edium

tech

inexpo

rts,

1997

-200

7shareof

HT,

shareof

MT,

density

ofHT

density

ofMT

shareof

HT,

shareof

MT

,do

mestic

domestic

domestic

domestic

domestic

domestic

Ordinarytrad

eOrdinarytrad

eOrdinarytrad

eOrdinarytrad

ePCStrad

ePCStrad

elogcapitals

tock

percap.

-0.017

c0.11

8a0.13

5a0.06

4b-0.016

0.02

7(0.009

)(0.024

)(0.026

)(0.028

)(0.034

)(0.069

)

stocksecond

aryeduc

0.09

2b0.07

50.10

80.00

9-0.194

-0.239

(0.042

)(0.111

)(0.109

)(0.113

)(0.162

)(0.328

)

shareof

HTe

ch,

0.05

2ano

ndo

mesticOrd.trad

e(0.019

)

shareof

medium

tech,

0.15

4ano

ndo

mesticOrd.trad

e(0.044

)

prod

uctdsty

0.16

7bHtech

nondo

mesticordina

rytrad

e(0.071

)

prod

uctdsty

0.17

2aMtech

nondo

mesticOrdinarytrad

e(0.063

)

shareof

High-tech,

-0.024

nondo

mesticprocessing

trad

e(0.029

)

shareof

Medium-tech

-0.177

b

nondo

mesticprocessing

trad

e(0.082

)

logratioof

fdis

tock

-0.002

0.00

10.01

1-0.005

0.02

3-0.030

tocapitals

tock

(0.004

)(0.011

)(0.014

)(0.012

)(0.022

)(0.044

)

logpa

tentsgran

ted

0.01

4a-0.000

-0.012

-0.008

0.05

4a0.09

1b(0.005

)(0.013

)(0.013)

(0.015

)(0.018

)(0.036

)year

fixed

effects

yes

yes

yes

yes

yes

yes

Observation

s32

832

832

132

431

931

9R

20.17

40.63

80.68

00.73

80.09

40.11

2Stan

dard

errors

inpa

rentheses

cp<

0.1,

bp<

0.05,a

p<0.01

Note:

High-tech/m

edium-techprod

ucts

:follo

wingtheclassificationof

Lall(2000).

Wecompu

tetheshares

ofhigh

-tech/

medium-techprod

ucts

intotal

expo

rtsof

domestic/

non-do

mesticfirms,

inordina

rytrad

e/in

processing

trad

e.Patents:totalnu

mbe

rof

patentsgran

tedby

year

andprovince,

includ

ingthreetype

sof

patents:inventions,utility

mod

elsan

ddesign

s.So

urce

:China

statisticalyearbo

ok.Regressions

also

includ

econtrols

for

shareof

Stateinvestment,an

dinfrastructure,a

sin

previous

specification

s

49

Page 50: Sophistication of China’s exports and foreign spilloverscerdi.org/uploads/sfCmsContent/html/317/Jarreau.pdf · Sophistication of China’s exports and foreign spillovers Joachim

Table5:

IVregression

s-Instrumented:

soph

istication

ofno

n-do

mesticexpo

rts,

ordina

rytrad

e(1)

(2)

domesticsoph

istication

domesticsoph

istication

ordina

rytrad

eordina

rytrad

esoph

istication

nondo

mestic,

0.39

4a0.41

9aordina

rytrad

e(0.128

)(0.143

)

logcapitals

tock

percap.

0.05

00.04

4(0.043

)(0.046

)

stockun

iversity

educates

over

pop1

5+1.02

5a1.02

6a(0.318

)(0.321

)

infrastructure

(ratio

topo

p.)(-3)

0.01

3a0.01

3a(0.003

)(0.003

)

shareof

Stateinvestmentin

totalinv

estm

ent(-1)

-0.131

-0.135

(0.092

)(0.093

)Observation

s32

832

8R

20.57

10.55

6Partial

R2of

exclud

edinstruments

0.03

70.03

6First-stage

test

ofexclud

edinstruments

(F,(Prob>

F))

14.00(0.000

)12

.47(0.000)

Kleibergen-Paa

pun

deridentification

test

(rkLM

stat)

11.807

15.291

Chi

2p−

valu

e0.00

20.00

0Han

senJ-statisticof

overidentific

ationof

allinstrum

ents

(Chi

2)

1.09

30.10

3p-value

(0.295

)(0.74)

Weakidentificationtest

(Cragg

-Don

aldtest

F-statistic)

14.00

12.46

Stock-Yog

oweakID

test

critical

values

(10%

)19.93

19.93

Criticalv

alues(15%

)11

.59

11.59

Dav

idson-McK

inno

ntest

ofexog

eneity

F-stat

2.20

2.61

p-value

0.13

80.10

7Heteroskeda

stic-con

sistentstan

dard

errors

repo

rted

cp<

0.1,

bp<

0.05,a

p<0.01

Note:

All

regression

sinclud

eprov

ince

fixed-effe

cts.

Non

-dom

esticsoph

istication

isinstrumented

using2instrumental

variab

les.

Colum

n1:

instruments

arethesumsof

coun

tries’sexpo

rtsoph

istication

levels,weigh

ted

byinverse

distan

ceto

provinces.

Cou

ntries

are(H

onk-Kon

g+

Sing

apore)

and(Jap

an+

S.Korea).

Colum

n2:

instruments

are

sumsof

coun

tries’

registered

patentsweigh

tedby

inversedistan

ce;s

amecoun

tries.

50