Sunteți pe pagina 1din 24

Sectoral Systems and International Technological and Trade Specialisation

Franco Malerba and Fabio Montobbio


CESPRI, Universita L.Bocconi, Via Sarfatti 25, 20136, Milano, Italy Tel. ++390258363391, Fax. ++390258363399 E-mail:
franco.malerba@uni-bocconi.it fabio.montobbio@uni-bocconi.it

Presented at DRUID Summer 2000 Conference Rebild, June 15-17, 2000

This paper is part of the Research Project Sectoral Systems in Europe - Innovation, Competitiveness and Growth (ESSY) under the Third Research and Technological Framework Programme, Targeted Socio-Economic Research, TSER [Contract no: SOE1-CT 98-1116 (DG 12-SOLS)]. We are grateful to Lorenzo Cassi, Andrea Porcu, Giorgio Rivero, Riccardo Torchio for valuable research assistance. We are also indebted with Stefano Breschi for insightful suggestions. The usual disclaimer applies.

1. Introduction
This paper aims at examining the role of key sectoral systems variables in affecting the international technological and commercial specialization of countries. It takes as a starting point the conceptual analysis of sectoral systems done in a related paper (Malerba, 2000). There the claim is advanced that any examination of the working and dynamics of sectors has to take into full account the knowledge base and links and the structural features that characterize a sectors. Following that line of inquiry, this paper explores first whether the knowledge base, knowledge linkages and the structural features of a sectoral system affect countries international specialization at the technological levels and second whether international technological specialization affects international commercial specilization. The paper is organized as follows. In section 2 we will lay down the basic concepts that are going to be used in the analysis, and in section 3 the main hypotheses to be tested are presented. In section 4 the empirical model, the data and the econometric specification are presented. In particular we test the relationship between a the set of variable related to sectoral systems (market structure, knowledge links and technological co-operation) and the technological specialisation of countries. Moreover we test whether there is a relationship between the technological specialisation and the trade specialisation of countries. In section 5 we present the conclusions and prospects of future reasearch.

2. The conceptual framework.


The sectoral system approach has emphasized a set of variables that affect the working, dynamics and performance of a sectoral system. These variables can be related to the structure of knowledge, agents and artifacts. In this paper we examine the role of some of these variables in affecting countries international technological and commercial specialization. In particular, we take into consideration some basic dimensions of knowledge such as the differences in the features of knowledge of broad macro sectors, the knowledge connections among sectors, the knowledge links within sectors, some key structural features of sectoral systems related to Schumpeterian competition (such as the concentration and asymmetries in innovative activities and the emergence of new innovators) and the relevance of networks in terms of technological cooperation. In what follows we briefly discuss the concepts related to these variables.

2.1

Knowledge

Differences across macro-sectors in knowledge features An enormous literature on technology and technological change has clearly shown how sectors differ in their basic technologies and how these technologies change over time affecting the nature, boundaries and organziations of sectors. The specific basic technologies of a sectoral system constitute major constraints on the full range of diversity in behavior and organization for firms active in a sector compared to other sectors. A given technological environment defines the nature of the problems firms have to solve in their innovative and production activities and the types of incentives and constraints to particular behavior and organizations. This similarity however refers only to a broad set of firms innovative and productive behavior and organization. Within this broad set, great and persistent firms heterogeneity is present.

Knowledge links among and within sectoral systems Links and complementarities affect the knowledge base and technologies of sectoral systems. These links and complementarities may be static or dynamic type. Input-output links are one of these quantitative links. Dynamic complementarities take into account knowledge and technological interdependencies and feed-backs. Links and complementarities greatly affect firms strategies, organization and performance, the rate and direction of technological change, the type of competition and the networks among firms and among firms and non-firms organizations. This statement is based upon a set of contributions that have stressed that the boundaries of sectors should include interdependencies and links among related industries and that these boundaries are not fixed, but change over time. These contributions could go back to the French tradition of filiere which has highlighted the role of major vertical links among sectors in production activities. Also the Swedish tradition of development blocks introduced by Eric Dahmen (1989) has stressed the idea that investments often are closely interrelated and span over different technologies or activities, and that they may originate tensions and virtuous cycles among related products in the process of economic development. Dynamic complementarities among artefacts and activities thus provide force and trigger mechanisms of growth and innovation.

2.2

Schumpeterian variables

Concentration of innovative activities at the sectoral level Within a sectoral system, innovative activities may be concentrated in few innovators or may be dispersed among a large group of firms. This difference in the organization of innovative activities at the sectoral level may be related to a fundamental distinction between Schumpeter Mark
3

I and Schumpeter Mark II models. (Malerba-Orsenigo,1997; Breschi-Malerba-Orsenigo,2000). Schumpeter Mark I is characterized by "creative destruction" with technological ease of entry and a major role played by entrepreneurs and new firms in innovative activities. Schumpeter Mark II is characterized by "creative accumulation" with the prevalence of large established firms and a stable core of innovators and the presence of relevant barriers to entry for new innovators. High

technological opportunities, low appropriability and low cumulativeness (at the firm level) conditions and a limited role of generic knowledge lead a Schumpeter Mark I pattern. On the contrary, high appropriability and high cumulativeness (at the firm level) conditions and a generic knowledge base lead to a Schumpeter Mark II pattern.

New innovative actors and innovative turbulence A related aspect refers to the role of new innovators. An important role in sectoral systems is played by new agents- both new firms and non-firms organizations-. For example new firms bring in the innovation and production processes a variety of approaches, specialization and knowledge, and contribute to major changes in the population of agents and in the transformation of technologies and products in a sector. The role of new firms differ drastically from sector to sector both in terms of rate of entry and in terms of composition and origin, and thus has quite different effects on the features of the sectoral systems and their degree of change. In particular, sectoral differences in the level and type of entry seem to be closely related to differences in the knowledge base, the level, diffusion and distribution of competencies within a sectoral system, the presence of non-firms organizations such as universities and venture capital and the working of sectoral institutions (such as regulations or labor markets) (Audretch,1996; Geroski,1995, MalerbaOrsenigo,1999).

2.3 Networks and technological collaborations

Within sectoral systems, heterogeneous agents are connected in various ways through market and non-market relationships. The evolutionary approach and the innovation systems literature have stressed the diversity in knowledge and capabilities among agents the relevance of trust and the range of informal interactions and relationships among agents (Lundvall,1993, Nelson,1995, Edquist 1997; Dosi,1997; Montobbio, 2000). In addition, they have emphasized that the relationships between firms and non-firm organizations such as universities and public research centres are very important sources of innovation and change in several sectors (NelsonRosenberg,1993). In uncertain and changing environments technological collaborations emerge not
4

because agents are similar, but because they are different. In this way networks may integrate complementarities in knowledge, capabilities and specialization.

3. The hypotheses to be tested


The main focus of this paper is an enquiry on the nature of the relation between the international specialization of countries in a specific sectional system and the sectoral variables discussed above. First we examine some key determinants of international technological specialization and then we analyse the relation between international technological specialization and international commercial specialization. In doing that, it will use a highly disaggregated data set on technologies-sectors. Specifically, in the first part of the paper we are going to test whether countries international technological specialization is affected by variables related to knowledge, Schumpeterian competition and technological cooperation.

3.1 Knowledge Variables

Differences in macro sectoral knowledge features Starting from the claim that there are differences across sectors in knowledge, patterns of innovative and production activities, and linkages, we will test whether the role of the variables discussed above differs among three major sectoral systems: electronics, chemicals and machinery. Although these broad sectoral systems are composed by a large set of different micro sectors and technologies (which we capture in our disaggregated analysis) a major difference exist among them in the knowledge, technologies and scientific fields that are at the base of firms innovative and production activities. As previously said, in a future paper we will test whether differences across sectors could be linked to the patterns of innovative activities and technological regimes (and not like here to differences in basic knowledge, technologies and scientific fields), and whether these differences affect countries international technological specialization. Thus we would expect that the role of knowledge links, Schumpeterian variables and technological cooperation differs across broad sectors such as chemicals, electronics and machinery, as a consequence of differences in the basic knowledge, technologies and scientific fields of various sectional systems.

Inter-sectoral knowledge connection advantages

Here the hypothesis is that the international specialization of a country in a specific sector or technology is positively affected by its knowledge links with sectors or technologies in which the country is internationally specialized. In a sense, through knowledge links, a country will be able to benefit from its specialization in other sectors or technologies which are knowledge related.

Within-sector knowledge links Here the basic hypothesis is that firms may profit from the presence of knowledge links within the country in a specific sector or technology, and that would positively affect the international specialization of that country in that sector or technology.

3.2 Schumpeterian variables

Concentration of innovative activities at the sectoral level Previous research (Pavitt, 1984, Malerba-Orsenigo,1997, Breschi-Malerba-Orsenigo,2000) has found that the organization of innovative activities at the sectoral level could be diffused among several new firms or concentrated in few large innovators. This difference is related to a Schumpeter mark I (low concentration) and Schumpeter Mark II (high concentration) sectors and depends on the specific features of the learning (technological) regimes. In this paper we are going to test the hypothesis that, irrespective of the type of Schumpeterian patterns, countries international specialization is positively affected by a higher degree of innovative concentration in a sector. This would mean that for both Schumpeter mark I and Schumpeter Mark II sectors, countries would benefit from a large oligopolistic core of innovators (oligopolistic core hypothesis) even in Schumpeter Mark I types of sectors.

Technological entry and innovative turbulence at the sectoral level Here the hypothesis is that countries would benefit from some degree of technological entry which usually brings in new ideas, new technologies and new products. Thus some turbulence would be beneficial even to countries in highly concentrated sectors such as the Schumpeter Mark II ones. In a sense, hypotheses 3.a and 3.b are based on the statement that in each sector one may find both core and fringe innovators. While the relative balance between the two differs across sectors (according to the Schumpeter Mark I or II distinction), on average the presence of a stable oligopolistic core and of a fringe of dynamic innovative entrants favours innovation and international specialisation.
6

It should be clear that in this paper we are going to test variables affecting international technological specialization (in line with the contributions of Dosi-Pavitt-Soete,1990), and not differences in the patterns of innovative activities. In a future paper, we are going to explore whether, when grouping sectors according to the Schumpeter Mark I and Mark II distinction, structural factors patterns. affect international specialization differently in each of the two innovative

3.3 Technological cooperation

Technological cooperation A related hypothesis refers to the role of cooperation in innovative activities. Again here the hypothesis to be tested is whether a greater amount of technological cooperation is conducive to greater technological specialization at the country level because it increases the diffusion of knowledge, it gives access to complementarities and it reduces the amount of uncertainty that each firms is facing in innovative activities.

4. The empirical model


4.1 The Data and the Econometric Specification

The data set is composed by patent applications (EPO/CESPRI) and data on export values (OECD, ITC: International trade by commodities). Four countries are considered: US, Italy, France, Germany. The period covered is from 1989 to 1994 and the values are averages on three subperiods, 1989-1990, 1991-1992 and 1993-19941. We consider 135 products in three industrial sectors, in particular we have 58 chemical products, 38 electronic products and 36 machinery products2. The concordance between IPC classification and SITC classification is provided by Grupp, Munt, (1997). The products are listed in Appendix 1.

This choice is due to export data availability with SITC Rev. 3 sectoral classification which is used in the FHG_ISI concordance table with IPC. 2 CESPRI/EPO data-set has 919451 patent applications for the period 78-96. Firms that are part of business groups have been treated as individual companies. In case of co-patenting each co-patentee has been credited the patent. In the period 89-94 in Germany, France, Italy and US, CESPRI/EPO has 225987 patent applications. Our 135 classes covers 152913 patent applications (68% of EPO). The total number of patent applications for each technological class, for the 89-94 period for the four countries, ranges from 105 (chem41) to 7731 (elek31). This is because we have excluded from the analysis technological classes with a total number of patents in all countries is lower than one hundred. The average number of patent applications per technological class is 1368. The median is 973. Citations are considered for the whole period (78-96). The total amount of citations is 799038. The citations between our 135 classes are 437752. 7

In order to test the hypothesis addressed in Section 3, our benchmark specification is:

RTANij=i+HERFij+ENTRYij+COPATFij+CITij+LINKCITij+ij j=1,....,135. Products, i=1,...,4. Countries

eq.1

RTANij is the normalised index of revealed technological advantage. Knowledge variables: LINKCITij = u k ij RTAN ikt
k j

where ukij = CITkij/CITij. Note that ujij is not included in the calculation of the index. ACITij=CITjij/CITij where CITjij is the share of citations in country i from class j to class j on the total number of citations in class j and country i (period: 1978-1996).

Schumpeterian variables: HERFij: the Herfindahl index in class j, country i. ENTRYij: the share of firms, which innovate for the first time, in class j and country i. Technological co-operation: COPATFij: the share of co-patenting firms in class j, country i. The first group of variables is related to the knowledge links in a sectoral system. They tests the relevance of knowledge connections within technological classes (ACITij) and of the strength of the knowledge linkages between sectors (LINKCITij). LINKCITij values belong to the (-1, +1) interval. Positive values (for country i and technological class j) point to the existence of strong knowledge linkages in terms of citations with sectors in which country i tends to be specialized. Negative values emphasize that technological class j in country i is connected, in country i, with sectors with technological revealed disadvantages. Since this variable points at the knowledge connections of class j with technological classes different from j, in the computation of the index the elements of the main diagonal of the matrix U={ ukj } (for each country i) have been excluded. We expect a positive relationship between LINKCITij and RTANij.
8

ACITij belongs to the unit interval and represents the amount of citations into technological class j in country i, on the total number of citations from technological class j country i3. This index is a proxy for the intensity of the knowledge links within a country and within a technological class. We build a data-set in which technological classes are considered at a fairly dis-aggregated level. Therefore we assume that a high value of ACITij can represent the existence of a pool of connected firms within a country which share the same technology specific knowledge base. We expect this variable to be positively associated with the construction of technological advantages. The second group of variables HERFij and ENTRYij (Schumpeterian variables) describes the effects of innovative concentration and entry on technological specialization. We expect a positive relationship between HERFij and ENTRYij and RTANij because as explained in Section 3, technological advantages of countries could benefit from the presence of both a core of stable and persistent innovators and from the presence of a pool of new innovators exploiting new technological opportunities. Technological co-operation is proxied by (COPATFij). COPATFij is the propensity to copatent of firms in a specific technological class and country. It ranges between 0 and 1. This proxy clearly covers a small part of the possible collaborations between firms innovative activities. Nevertheless we believe that is worthwhile to ask whether even this small aspect of technological co-operation has statistical significance in its association with international technological specialisation.

4.2 Evidence on the relation between sectoral systems and technological specialisation

We first estimate (Tab.1, column 1) Equation 1 in a simple static framework using the entire pooled sample in a specific sub-period4 (a dynamic element is introduced in equation 2). We control for macro-sector group-wise heteroscedasticity using feasible generalized least squares and sectoral fixed effects and differences across countries using country fixed effects. We controlled also for collinearity among the explanatory variables. Correlation matrices (Tab 2.a in the Appendix) show that there is no severe collinearity between independent variables. Tab 1 (column 1) shows that, consistently with our expectations, technological specialization is positively associated with our knowledge variables. The high positive value of the estimated coefficient of intensity of knowledge linkages with other technological classes in which countries
3

This indicator does not take into account the absolute level of citations within a technological class. It simply emphasizes the share of patents for which citing and cited technological classes are the same on the total amount of citation in that class. It might be that ACITij is high even if the amount of citation is low. 4 Estimates in Tab.1 refer to the first sub-period in our dataset: 1989-1991. The results for the other sub periods are consistent with these results. 9

have advantages (LINKCITij) is particularly noteworthy. This also implies the negative effect of strong knowledge linkages with unspecialized technological classes. Moreover RTANij is positively associated with the strength of knowledge links within a technological class (ACITij). With regard to the Schumpeterian variables, we observe that technological advantages are positively associated with the index of innovative concentration (HERFij). The estimated relation between concentration and technological advantages might not be independent from the rate of entry of innovative firms. In particular it is plausible that technological specialisation is affected by a specific combination of concentration and rates of entry. Therefore we added an interaction term (ENTRYij*HERFij) between concentration and entry which tests whether the marginal effect of higher innovative concentration on technological advantages is decreased when entry is higher. The estimated coefficient does not reject this hypothesis: the positive relation between the Herfindahl index and RTANij is lower in presence of entrants competing with incumbent firms. Finally the level of technological co-operation among firms, as expressed by the share of copatenting firms on the total amount of patenting firms (COPATFij), is positively associated with technological specialization. Even if this index catches only a very limited portion of the many possible forms of firms collaboration, our evidence suggests that it has a positive impact on the construction of technological advantages. As we argued in section 3 the existence of macro-sectoral specific knowledge features suggests that the relationships between the knowledge, schumpeterian and technological collaboration variables and the patterns of technological specialization of country can be sector specific. In this framework we can assess sectoral diversity from three points of view: coefficients significance and magnitude in different sectors and non-linearities. Therefore we run a separate set of regressions for the three different macro sectors: Chemicals, Electronics, Mechanics (Tab. 1, column (2), (3) and (4)). We tested for non linearity in the relation between the schumpeterian variables and RTANij. We included the squared terms for ENTRYij and HERFij. It turns out that just in the Electronic sector we have a squared term (ENTRYij* ENTRYij) significantly different from zero and the result is presented in Tab.1 column (3a) (we comment upon this below). The portion of RTANij variance explained ranges from 0.43 in Chemicals to 0.23 in Electronics. For what concern the knowledge variables, strong and consistent results for all macro sectors concern the variable reflecting the inter-sectoral knowledge connection advantages (LINKCITij). Estimated coefficients are significantly different from zero in all sectors. The value of the estimate is particularly high in the Electronic sector followed by the Machinery sector. Knowledge links within sectors (ACITij) have a positive role only in the Chemicals and estimated coefficients do not present values significantly different from zero in the two other macro-sectors.
10

With regard to the Schumpeterian variables, the evidence of the regression on the entire sample is confirmed in Chemicals and Machinery. In particular in the Machinery sector we observe a high value of the estimated effect of innovative concentration (HERFij) on RTANij. At the same time it is particularly strong the interaction effect of HERFij with ENTRYij suggesting that in these technological classes concentration affects positively technological advantages only if it is not associated with a high share of new innovative firms. The same result, in a weaker form, holds for the technological classes in the Chemical sector. In Electronics, if we take the linear specification in column 3 (Tab.1), among the Scumpeterian variables, only the coefficient of the interaction term (ENTRYij*HERFij) shows an estimated value significantly different from zero. This probably stresses that countries are not specialised in those classes where there are few innovative firms and innovation is sporadic. At the same time, results reported in Tab.1 (column 3a) show a highly non-linear relation between ENTRYij and RTANij and, in particular, they suggest the existence of an invertedU relationship. Both the linear and the squared terms for ENTRYij are statistically significant at the 0.05 level and they presents opposite signs. We suggest, therefore, that in Electronics, technological specialisation is most likely to be associated to higher level of entry, but only up to a certain threshold. In sum it turns out that the probability to observe that country i is specialised in a technological class j (in Electronics) is high for medium level of ENTRYij jointly with small levels of innovative concentration (HERFij).

11

Tab.1: Estimation of eq. 1. TOT (1) Linkcit Acit Herf Entry Entry*Entry Herf*Entry Copatf 2.85**
(9.27)

CH (2) 1.7**
(3.4)

EL (3) 3.50**
(4.52)

EL (3a) 3.32**
(4.39)

MECC (4) 2.24**


(3.5)

0.26**
(3.16)

0.29*
(2.4)

0.7**
(3.06)

0.81*
(2.15)

1.8*
(1.9)

-1.34**
(-4.38)

-1.24**
(-2.8)

-0.98*
(-2.4)

2.03* (2.1) -1.6* (-2.01) -1.27*


(-1.7)

-3.44**
(-2.7)

0.19*
(2.005)

0.26*
(1.6)

N 409 152 137 137 120 R2 (a) 0.30 0.43 0.23 0.26 0.39 X2(Wald) 205.74** Specification (1) uses FGLS estimator with groupwise heteroscedasticity. Specification (2), (3) and (4) use the robust White/Huber/sandwich estimator of variance (t in parenthesis) (a) for the FGLS regression in column (1), R2 refers to the OLS estimation. ** 99% significativity level * 90% significativity level + 85% significativity level

To explore further we now turn to the analysis of the whole data set, which is composed by three sub-periods. We extend the econometric work and include an important characteristic of the dynamics of technological specialization: its degree of persistence. The estimation of the determinants of technological specialization is now based on the following specification: RTANtij=i+RTANt-1ij +HERFtij+ENTRYtij+COPATFtij+CITtij+LINKCITtij+ij

eq.2

First consider the pooling of all sectors. Results are displayed in Tab. 2 (column 1). As expected, the auto-regressive component is positive and confirms a well-known result on the persistence of specialisation patterns. We turn now to the main focus of our paper concerning the impact of knowledge, Schumpeterian variables and technological co-operation on revealed technological advantages. With regard to the knowledge variables, the estimated coefficients of the inter-sectoral knowledge connection advantages are confirmed. The estimated coefficient of LINKCITij is again positive and significantly different from zero. At the same time the index of knowledge links within sectors (ACITij) does not have a statistically significant impact on RTANij.

12

Turning to the Schumpeterian variables, technological specialisation is positively associated with innovative concentration (HERFij) and with the share of new innovative firms (ENTRYij). The results of a negative impact of the interaction term between HERFij and ENTRYij remains strong. On the pooled sample there is also evidence of a positive relationship between technological cooperation (COPATFij) and patterns of technological advantages. Taken together these results confirm the ones presented in Tab.1 with two exceptions. First we observe a positive value of the estimated coefficient for ENTRYij (whose marginal effect decreases with concentration). Second, we cannot confirm the evidence suggested in Tab. 1 of a positive association between ACITij and international technological specialisation. Again the existence of macro-sectoral specific knowledge features suggests that the relationships between knowledge, Schumpeterian variables and the technological co-operation and the patterns of technological specialization of country can differ across macro-sectors. Results for the regressions on the macro-sectoral sub-sample (Tab. 2, column (2), (3) and (4)) confirm the importance of the auto-regressive component in all sectors. Moreover in all sectors the coefficients of LINKCITij are positive and statistically significant. Consistently with the findings displayed in Tab. 1, the positive signs indicate that higher knowledge connections (in terms of citations) with technological classes with high specialisation, are associated with higher technological advantages in the citing class. Again the estimated coefficient is larger in the Electronic sector. With regard to coefficient of the other knowledge variable (ACITij) the significant positive sign is confirmed for Chemicals. At the same time a significant negative relation appears for Electronics. If we turn the attention to the Schumpeterian variables, the positive association between innovative concentration (HERFij) and technological specialisation is confirmed in all sectors. This relation is particularly strong in the Mechanical sector as in the first static specification. Now evidence reports also a positive sign of the estimated coefficient of HERFij in the Electronic sector. The negative interaction between HERFij and ENTRYij is statistically significant in only two sectors out of three: Mechanicals and Chemicals. These are the sectors where the effect of the specific combination of the market structure variables is particularly strong as emphasised also in Tab. 1. In contrast with the evidence presented in Tab. 1, we do not find any significant relationship between the interaction term and technological specialisation for Electronics. Moreover the estimation of equation 2 displays a significantly positive relation between ENTRYij and technological advantages. This suggests that the existence of a pool of new innovators exploiting latent technological opportunities may be very important in the construction of technological advantages.

13

Again this effect is stronger the lower is innovative concentration and particularly in the Chemicals and Mechanical sectors5. The role of technological co-operation, appears to be positive in the Electronic and Mechanical sector while it is not significantly different from zero in the Chemical technological classes. This result is not in line with the previous specification where the estimated coefficient of COPATFij was significantly positive only in the Chemical sector.

Tab.2: Estimation of eq. 2. TOT (1) RTA(-1) Linkcit Acit Herf Entry Herf*Entry Copatf 0.78**
(37.3)

CH (2) 0.81**
(22.6)

EL (3) 0.86**
(27.1)

MECC (4) 0.86**


(26.1)

0.6**
(4.1)

0.46**
(2.6)

0.91**
(3.5)

0.41**
(1.7)

0.07+
(1.5)

-0.11
(-1.9)

0.38**
(3.6)

0.37*
(2.4)

0.38*
(2.001)

0.8**
(3.5)

0.14**
(3.2)

0.15*
(2.2)

0.13+
(1.6)

0.22**
(3.2)

-0.67**
(-4.6)

-0.73**
(-3.006)

-1.5**
(-4.7)

0.15**
(3.7)

0.12*
(1.7)

0.22**
(2.8)

N 824 307 277 240 X2(Wald) 2157** 827** 1024** 1136** Specification (1) (2), (3) and (4) use FGLS estimator with groupwise heteroscedasticity consistent variance estimator (z in parenthesis). ** 99% significativity level * 90% significativity level + 85% significativity level

4.3 Technological specialisation and trade specialisation

In this section we explore with a very simple set of specifications the relation between revealed competitive advantages in terms of export and revealed technological advantages (Amendola et al., 1998; Dalum et al., 1998). The first step is to investigate the relation between RCA and RTA for all the available observations and to enquire whether the estimated parameter displays different values according to different lag structures. Specifications (3), (4), (5) are
Also in this case we explored the significativity of squared terms for ENTRYij and HERFij. Only in the Mechanical sector, results (not reported in Tab.3) suggest the existence of an invertedU relationship between ENTRYij and RTANij. Both the linear and the squared terms for ENTRYij are statistically significant at the 0.05 level and they presents opposite signs. 14
5

designed to perform this task (We include in the data-set also the data for the UK that we have not used in the previous section due to the lack of availability of firm-level data).

RCAtkij=ij+RTAtkij+tkij RCAtkij=ij+RTAtkij +RTAt-1kij+tkij RCAtkij=ij+RTAtkij +RTAt-1kij +RTAt-2kij+tkij

(3) (4) (5)

k=1,...., 132. Products i=1,..., 5. Countries j=1,..., 3. Industrial Sectors t=1,..., 3. Sub-periods The result of estimation of specification (3), (4), (5) are shown in Table 3. Coefficients , , and are always positive and significantly different from zero. Moreover their values increase if the impact of RTA on RCA is analysed with one and two years lag. In the second step the restriction that , , and are constant across sector is considered not acceptable. Therefore a second set of specifications is proposed which tests the hypothesis of sectoral specificity in the relation between technological and export specialisation.

RCAtkij=ij+jRTAtkij+tkij RCAtkij=ij+jRTAtkij +jRTAt-1kij+tkij RCAtkij=ij+jRTAtkij +jRTAt-1kij +jRTAt-2kij+tkij

(6) (7) (8)

k=1,...., 132. Products i=1,..., 5. Countries j=1,..., 3. Sectors t=1,..., 3. Sub-periods First, almost all the estimated parameters are positive and significantly different from zero for all specification (6), (7) and (8). Note that multicollinearity affects the results in the regressions which include the contemporary variable and the distributed lags. Therefore to control for collinearity regressions on one single lagged variable are performed. Again the estimates, displayed
15

in Table 4, tend to increase with the length of the lag as in specification (3), (4), (5) with the exception of the mechanical sector for which the contemporary regression displays a higher value than the one lag regression (lower than the two lags if taken alone, though). Importantly sectoral differences emerge. The relation between revealed comparative advantages and technological specialisation tends to be higher in the electronic sector where also a higher portion of RCA variance is accounted for by RTA variance. Finally also the restriction that is constant across countries is rejected and specification (9) is proposed. RCAtkij=ij+ijRTAtkij+tkij k=1,...., 132. Products, i=1,..., 5. Countries j=1,..., 3. Sectors t=1,..., 3. Sub-periods Results are shown in Tab. 5 and confirm the previous findings. Coefficients are, in general positive and statistically significant. Estimates are bigger in the case of the electronic sector. For this industrial sector we have also higher R2. Coefficients for chemicals are always significantly different from zero (a part from Italy) and they are smaller then the estimates for electronics but bigger than the estimates for the mechanical sectors. In this last sector coefficients cannot be considered different from zero in three case out of five. Results for Italy show as expected that the relation between RTA and RCA is particularly strong in the mechanical sector and absent in the chemical sector. (9)

5. Conclusions
In this paper we have explored whether the knowledge base, the linkages and the structural features of a sectoral system affect countries international technological specialisation and whether there is a link between technological specialisation and commercial specialisation. In particular we have examined a set of variables that affect the working, dynamics and performance of a sectoral system and which have been grouped into knowledge, Schumpeterian and technological cooperation variables. In particular, we have taken into consideration some basic dimensions of knowledge such as the differences in the features of knowledge of broad macro sectors, the knowledge connections among sectors, the knowledge links within sectors, some key structural features of sectoral systems related to Schumpeterian competition (such as the concentration and asymmetries in innovative activities and the emergence of new innovators) and the relevance of
16

networks in terms of technological cooperation. We have found that these variables have in general positive effects on the international technological specialization of countries. In addition, we have examined the relationship between international technological specialization and commercial specialization. Results confirm a positive and statistically significant relationship. Also in this case the relationship show differences between sectors. These results lead to a better understanding of the effects of sectional systems variables, such as knowledge, Schumpeterian competition and technological cooperation, on international specialization. The next steps in our research agenda are the following. First, we want to test whether the differences across macro sectors that we have found are indeed relevant from a statistical point of view. Second, we would like to have a better understanding of the relationship among Schumpeterian variables. In particular we have shown that in some cases the effects on technological specialisation are non linear and that the specific combination of concentration and entry affect importantly RTANij. Finally, we aim to test whether differences across sectors could be linked to Schumpeterian patterns of innovative activities and technological regimes (and not to broad macro-sectors such as chemicals, electronics and mechanicals) and whether these differences affect countries international technological specialization.

17

Table 3: Estimation results of specifications (3) (4) (5) RTAtij RTAt-1ij RTAt-2 ij R2 F n 0.30** 0,18 59.15** 1900 (16.2) (2) 0.14** 0.32** 0.26 53.96** 1257 (6.18) (11.58) 0.30** 0,17 38.90** 1260 (13.3) (3) 0.09* 0.20** 0.26** 0.30 29.6** 627 (2.02) (4.5) (5.05) 0.42** 0.23 27.29** 630 (11.65) ** coefficients statistically different from zero at a 99% level of confidence. Sectoral and country fixed effects are included. Dep. Var. RCAtij (1)

Table 4: Estimation results of specifications (6) (7) (8) Dep. Var. RCAtij RTAtij RTAt-1ij Specification 4 Chemicals 0.21** (7.39) Electronics 0.46** (17.86) Mechanics 0.26** (4.87) Specification 5 Chemicals 0.12** 0.34** (3.78) (7.36) 0.20** (5.62) Electronics 0.24** 0.31** (4.78) (6.36) 0.49** (15.51) Mechanics 0.18** 0.12** (2.78) (1.81) 0.21** (3.64) Specification 6 Chemicals 0.08 0.29** (1.33) (3.9) Electronics 0.04 0.2** (0.43) (2.46) Mechanics 0.30** -0.01 (2.5) (-0.1) ** coefficients statistically different from zero at a 99% level of confidence. * coefficients statistically different from zero at a 95% level of confidence. Country fixed effects are included.

RTAt-2ij 0.20** (2.32) 0.34** (5.23) 0.37** (4.5) 0.54** (12.6) 0.14 (1.15) 0.29** (3.03)

R2 0,08 0,39 0.06

F 14.19** 72.24** 6.86**

n 822 568 510

0.17 0.08 0.45 0.42 0.07 0.05

17.98** 9.1** 51.3 53.86** 4.587** 3.88**

539 542 378 378 340 340

0.30 0.12 0.51 0.49 0.11 0.07

10.51** 7.18** 27.03** 34.83** 2.85** 2.52*

268 271 189 189 170 171

18

Table 5: Estimation results of specification (9) Dep. Var. RCAtij RTAtij R2 F n USA Chemicals 0.47** 0.05 9.23** 168 (3.04) Electronics 0.54** 0.16 21** 114 (4.68) Mechanics 0.34 0.02 2.5 102 (1.6) Japan Chemicals 0.87** 0.34 83.56** 164 (9.14) Electronics 0.49** 0.20 28.43** 114 (5.3) Mechanics 0.18+ 0.02 2.77 102 (1.7) Italy Chemicals 0.08 0.01 2 154 (1.4) Electronics 0.55** 0.56 140.42** 112 (11.85) Mechanics 0.32** 0.23 29.61** 102 (5.44) Germany Chemicals 0.15* 0.04 6.36 168 (2.52) Electronics 0.32** 0.21 29.53** 114 (5.43) Mechanics 0.04 0.001 0.09 102 (0.31) France Chemicals 0.13** 0.06 11.31** 168 (3.34) Electronics 0.25** 0.15 19.30** 114 (4.39) Mechanics 0.08 0.002 0.226 102 (0.5) ** coefficients statistically different from zero at a 99% level of confidence. * coefficients statistically different from zero at a 95% level of confidence. + coefficients statistically different from zero at a 90% level of confidence.

19

Bibliography

Amendola, G., Guerrieri, P., Padoan, P.C. (1998). International Patterns of Technological Accumulation and Trade, in Archibugi, D., Michie J. (eds.) Trade, Growth and Technical Change, Cambridge University Press. Ch. 7. Audretsch D. (1996), Innovation and Industry Evolution, MIT Press, Cambridge Breschi, S., Malerba, F., Orsenigo, L. (2000). Technological Regimes and Schumpeterian Patterns of Innovation, The Economic Journal Dahmen E. (1989), Development blocks in industrial economics, in B. Carlsson Industrial dynamics, Kluwer Dalum, B., Laursen, K., Villumsen, G. (1998). Structural Change in OECD Export Specialisation Patterns: De-specialisation and Stickiness. International Review of Applied Economics, v. 12, n. 3, pp. 423-443. Dosi G. (1997), Opportunities, incentives and the collective patterns of technological change, Economic Journal Dosi, G., Pavitt, K., Soete, L. (1990). The Economics of Technical Change and International Trade, Harvester Wheatsheaf, Hertfordshire. Edquist C. (ed.) (1997), Systems of innovation, Frances Pinter, London Geroski P. (1995), What do we know about entry?, International Journal of Industrial Organization, 4, 421-440 Lundvall B.A. (1993), National Systems of Innovation, F. Pinter, London Malerba F. (2000); Sectoral System of Innovation and Production. ESSY Working Paper n. 1. Malerba F. Orsenigo L. (1997), Technological regimes and sectoral patterns of innovative activities, Industrial and Corporate Change, v.6, p.83-117 Malerba F. Orsenigo L. (1999), Technological entry, exit and survival: an empirical analysis of patent data, Research Policy 28, p. 643-660 Malerba F. Orsenigo L. (2000), Knowledge, innovative activities and industry evolution, Industrial and Corporate Change Montobbio F. (2000); National System of Innovation. A Critical Survey. ESSY Working Paper n. 2. Nelson R. (1995), Recent evolutionary theorizing about economic change, Journal of Economic Literature, v.33 Nelson R. Rosenberg N. (1993), Technical innovation and national systems in Nelson R. (ed), National Innovation Systems Oxford University Press, Oxford Pavitt K. (1984), Sectoral patterns of technical change: towards a taxonomy and a theory, Research Policy, 13, p.343-373

20

Appendix 1: The List of Products.


CHEMICALS: 1) chem101 2) chem102 3) chem111 4) chem112 5) chem113 6) chem114 7) chem121 8) chem122 9) chem123 10) chem131 11) chem132 12) chem133 13) chem134 14) chem135 15) chem136 16) chem137 17) chem14 18) chem15 19) chem16 20) chem17 21) chem21 22) chem22 23) chem23 24) chem24 25) chem25 26) chem26 27) chem31 28) chem32 29) chem33 30) chem41 31) chem42 32) chem43 33) chem44 34) chem51 35) chem52 36) chem53 37) chem54 38) chem55 39) chem56 40) chem57 41) chem58 42) chem61 43) chem62 44) chem63 45) chem64 46) chem71 47) chem72 48) chem73 49) chem74 50) chem75 51) chem76 52) chem77 53) chem81 = Starch (missing for Germany, France, Italy, Japan and USA in period 1 and 2 = Proteins = Explosives, gunpowder = Fuses, ignition chemicals = Pyrotechnique articles, fireworks = Matches = Additives for lubricating oil, corrosion inhibitors = Liquids for hydraulic brakes, anti-freezing compounds = Lubricants, emulsions for grease, petroleum sulfonate, artificial graphite emulsion = Gas cleaning = Catalysts = Additives for metals = Benzoles, naphts = Electronic and electrotechnical chemical compounds = Chemical substances for constructions = Chemicals for fire extinguishers, liquid polychlor dipheniles = Artificial and natural caoutchouc = Natural polymers = Plastic trash = Plastic products = Anorganic chemical compounds = Anorganic oxygen compounds = Anorganic sulphur compounds = Other metal salts = Other anorganic chemical products = Radioactive substances = Synthetic textile fibres = Artificial textile fibres = Trash = Organic oils and fats = Wax = Artificial wax = Chemical products of woodor resins = Hydrocarbons = Alcohol = Carbon acid = Compounds with nitrogen function = Organic-anorganic compounds = Lactame, andere heterocyclische Verbindungen = Sulfamide = Ether, alcohol peroxide = Synthetic organic colours and varnishes = Tanning agents and paint extracts = Colours, varnishes, pigments = Glazes, sealing compounds = Vitamins, provitamins, antibiotics = Hormons and derivatives = Micro-organisms, vaccines = Reagents and diagnostics = Other special medicines = Other pharmaceutical products = Cosmetics (no soaps) = Etheric oils and perfumes 21

54) chem82 55) chem83 56) chem84 57) chem91 58) chem92 ELECTRONICS: 1) elek10 2) elek11 3) elek11b 4) elek12 5) elek12b 6) elek13 7) elek13b 8) elek14 9) elek14b 10) elek15 11) elek15b 12) elek21 13) elek22 14) elek23 15) elek24 16) elek31 17) elek32 18) elek33 19) elek34 20) elek41 21) elek42 22) elek43 23) elek44 24) elek511 25) elek512 26) elek513 27) elek514 28) elek521 29) elek522 30) elek53 31) elek54 32) elek61 33) elek62 34) elek63 35) elek71 36) elek72 37) elek8 38) elek9

= Soaps = Detergents = Ski-wax, furniture polishes = Fertilisers = Insecticides

= Ignition cables, electrical cars = Small electrical engines, electrodes = Portable electrical tools = Motors, electrical engines and electrodes = Magnetic tapes = Choke coils, converters, transformers = Traffic lights, etc. = Generators and equipment = Particles accelerator = Transformers = Lasers = Fridges (for home and industry), air conditioning = Washing machines, dryers, dish washers = Electrical shavers, hair-cutting machines, hoovers = Electric heating = Computers and equipments = Computer chips and equipments = Photocopying machines and equipments = Type-writers and other office devices = TV, radio, TV-cameras, video-cameras, antennas, oscilloscopes = Microphones, loud-speakers, recorders = Telephones (no mobile phones) = Radio engineering devices = Circuits = Resistors = Switches, fuses = Control panels = Cables (without ignition) = Insulators = Capacitors = Electro-magnets = Electrical diagnostic devices (no X-rays) = X-rays = Instruments to show ionic beams = Diodes, transistors = Integrated circuits = Batteries, accumulators = Portable electrical lamps

MECHANICALS: 1) masch10 2) masch11 3) masch11b 4) masch121 5) masch122 6) masch12b 7) masch131 8) masch132 = Printing machines = Steam-boiler = Machines for food processing = Steam-turbines for ships = Steam-turbines for steam power plants = Machines to process rocks, etc. = Gas-turbines for aeroplanes = Gas-turbines for power stations 22

9) masch13b 10) masch14 11) masch15 12) masch16 13) masch17 14) masch18 15) masch19 16) masch20 17) masch21 18) masch21b 19) masch22 20) masch22b 21) masch23 22) masch23b 23) masch24 24) masch24b 25) masch25 26) masch26 27) masch27 28) masch28 29) masch3 30) masch4 31) masch5 32) masch61 33) masch62 34) masch7 35) masch8 36) masch9

= Wood processing machines = Plastic processing = Saws, etc. = Spanlos arbeitende Werkzeugmaschinen = Metal-working rolling mills = Soldering irons, blow lamps, welders = Torches, furnaces = Ovens, distilling apparatuses, gas distilling = Piston-drive engines for aeroplanes = Pumps, centrifuges, filters = Engines for cars = Conveyors = Engines for ships = Anti-friction bearing = Engines for trains = Valves = Packaging machines = Scates = Fire extinguisher, spray guns = Other machines = Water-turbines = Nuclear power reactors = Other engines = Agricultural machines (without tractors) = Tractors = Constructions and mining machines = Textile machines = Paper production machines

23

Tab A2: Correlation matrices: TOTAL (n=409) Herf Entry Copatf Acit Linkcit Herf Entry Copatf Acit Linkcit 1,0000 0,1465 1,0000 -0,0046 -0,2657 1,0000 0,0372 0,1755 0,0535 1,0000 -0,1971 0,0436 -0,0579 0,1673 1,0000

CHEMICAL (n=152) Herf Entry Copatf Acit Linkcit Herf Entry Copatf Acit Linkcit 1,0000 0,3118 1,0000 -0,1358 -0,2870 1,0000 0,1816 0,1860 0,0856 1,0000 -0,1087 0,0508 -0,0524 0,0894 1,0000

ELECTRONICS (n=137) Herf Entry Copatf Acit Linkcit Herf Entry Copatf Acit Linkcit 1,0000 0,1415 1,0000 0,0522 -0,1521 1,0000 0,1746 -0,0045 0,0832 1,0000 -0,3266 -0,1044 -0,0151 0,2376 1,0000

MACHINERY (n=120) Herf Entry Copatf Acit Linkcit Herf Entry Copatf Acit Linkcit 1,0000 0,0747 1,0000 0,1001 -0,3416 1,0000 0,0517 0,0163 0,1120 1,0000 0,0517 0,0602 -0,0072 0,2041 1,0000

24

S-ar putea să vă placă și