Research Policy 33 (2004) 409–424

Does innovativeness matter for international competitiveness

in developing countries?
The case of Turkish manufacturing industries
Emre Özçelik, Erol Taymaz∗
Department of Economics, Middle East Technical University, Ankara 06531, Turkey
Received 4 February 2003; received in revised form 1 September 2003; accepted 4 September 2003

Abstract
We examine the determinants of export performance of firms in the Turkish manufacturing industry. Prominent differences
show up between innovator and non-innovator firms in terms of the impacts of such variables as firm size, advertisement
intensities, ownership structures, and composition of employees. Importance of innovations and R&D activities, conduciveness
of capital intensity, and insignificance of the real wage are meaningful as far as a rational international competition policy is
concerned. Results are suggestive of a technology-oriented and capital-formative development path, if Turkey is to come up
with the international competitive standards.
© 2003 Elsevier B.V. All rights reserved.
Keywords: Innovation; R&D; International competitiveness; Exports

[I]n capitalist reality as distinguished from its text- orientation has basically generated a wide-spread
book picture, it is not that kind of competition which emphasis on the role of technology in developed coun-
counts but the competition from the new commod- tries. As a matter of course, being developed and tech-
ity, the new technology, the new source of supply, nological superiority have for long been the two sides
the new type of organization . . . (Schumpeter, 1942, of the same coin. Thus, relative fewness of the studies
p. 84) which deal with the less-developed economies must
not be surprising. Nevertheless, construction of a ratio-
nale for investigating the interactions between exports
1. Introduction and technology in technologically-backward countries
may also be fruitful in terms of a better determination
Accumulation of studies that have tried to link of relevant strategies. A perspective directed towards
the export performance of economic units (be they the differences among the successful and unsuccessful
countries, industries or firms) with their technological country-specific strategies for improving international
competitiveness may help one to distinguish between
∗ Corresponding author. Tel.: +90-312-210-3034;
the correct and ill-advised policy options. The Turk-
fax: +90-312-210-1244.
ish experience has demonstrated that mere recourse
E-mail addresses: [email protected], to ready-made policies (such as export subsidies or
[email protected] (E. Taymaz). real devaluations) cannot be useful in escaping from

0048-7333/$ – see front matter © 2003 Elsevier B.V. All rights reserved.
doi:10.1016/j.respol.2003.09.011
410 E. Özçelik, E. Taymaz / Research Policy 33 (2004) 409–424

underdevelopment or competitive disadvantages. of which is innovation) has served as a source of inspi-

Searching for alternative strategies seems inevitable. ration for several studies. For instance, Gruber et al.
Innovativeness, as a genuine lever required to raise (1967), Keesing (1967), Baldwin (1971), Branson
living standards, may well be the pertinent key to im- (1971), Lowinger (1975), Stern and Maskus (1981),
proving international competitiveness. Subject-matter and Sveikauskas (1983) revealed the significantly pos-
of this study is, thus, an inquiry into the seemingly itive impact of R&D efforts on US commodity trade
closed doors that could be opened by this key. in general. Similar results were obtained by Hughes
Following Section 2 not only elaborates on the (1986) for the UK and Vestal (1989) for Japan.
course of the alternative treatments of the technology At this point, inclusion of human capital and R&D
factor within the theoretical literature, but also sets as explanatory variables was necessary yet insufficient
forth industrial organizational bases for firm-level insofar as the framework of neofactor theories was
studies. It is in Section 3 where the empirical literature concerned. Treatment of physical and human capital
is surveyed. Section 4 begins with Turkey’s initially and technological factors as static endowments was
successful yet eventually retrogressive export-led inappropriate due to their pertinently dynamic nature.
growth strategy during the 1980s, and draws attention Trying to get rid of the static world, family of neotech-
to (i) substantially neglected formation of gross fixed nological theories of trade was, thus, a further attempt
capital and, (ii) lack of a political conscience as to in this respect.
the significance of a national technology policy. After Neotechnological attitudes basically originate from
describing the data set, the model and the method to Posner’s (1961) technological-gap theory (TGT) and
be utilized for empirical analysis, descriptive statistics Vernon’s (1966) product life-cycle theory (PLCT),
are summed up. In Section 5, we interpret the deter- both of which rely on varying production functions
minants of the export intensity of Turkish firms, and for the same commodities across countries. In the
compare and contrast innovators and non-innovators. TGT, a product innovation provides the innovating
The paper runs its course with a set of concluding domestic firm(s) with a temporary monopoly power
remarks in Section 6, which also embodies a few bits at home and abroad. Profits earned by the innovator
of policy recommendation. above “normal” levels lead to imitation on the part
of foreign firms, which eventually develop compara-
tive advantages in the new commodity. In this way,
2. “Innovative competitiveness” as rationale for this imitation lag is suggestive of technological-gaps
firm-level studies across countries on the basis of the differences in
innovative capabilities. PLCT, on the other side, rep-
“As is frequently observed, it matters a great deal resents a step forward with respect to the TGT. In
today whether a country specializes in the produc- contrast to the strong factor intensity assumption of
tion of potato chips or micro chips. According to con- the factor-endowment theories, PLCT predicts that a
ventional trade theory, however, this choice does not new product will have varying relative input require-
really matter” (Haque, 1995, p. 22). Apart from its ments over its life cycle. “Accordingly, as the product
“conventional” textbook versions, the evolution of in- matures and becomes standardized, comparative ad-
ternational trade theory has witnessed extensions. Ne- vantage may shift from a country relatively abundant
ofactor and neotechnological trade theories are two in skilled labor to a country abundant in unskilled
cases as such. labor” (Chacoliades, 1990, pp. 107–108). On the one
Variants of the neofactor theory, which emerged (static) hand, “countries with a high technological
as a reaction to the well-known Leontief paradox, capacity produce technology-intensive goods”; on the
have basically distinguished between qualified labor other (dynamic) hand, “technology intensity of goods
(human capital) and unskilled labor. Preserving the decreases over time as they become standardized”
assumption of common production functions over the (Wakelin, 1997, p. 17). Hence, the dynamic com-
world, they have included “knowledge” as an addi- ponent of these neotechnological theories basically
tional factor of production. The fact that knowledge relies on the changing input requirements of products.
can be generated through R&D (the expected resultant Changing production technologies across countries
 E. Özçelik, E. Taymaz / Research Policy 33 (2004) 409–424 411

are not a matter of import in this respect. The dy- nature of technological change”. Decisions to spend
namism here refers merely to products, in which on the generation and development of innovative ca-
case a static implication is still dominant in terms pacities are taken at the level of the firm (Wakelin,
of technical stability (Walker, 1979). Technologically 1997, p. 20). It is also a level at which the returns to
capable and relatively innovative countries will con- innovation are collected in terms of new products and
tinuously have a comparative advantage in new prod- markets, cost advantages, and rents. Finally, an im-
ucts, and a comparative disadvantage in standardized perfect world entails different degrees of international
ones. Of course, this leaves no potential dynamics competitiveness for firms from the same industry.
for “catching-up” through learning-by-innovating on At this point, it must be noted that firms are
the part of technologically-backward countries. Thus, not independent or autonomous entities in terms of
“product dynamism”, rather than truly technologi- their innovation-related decision-making processes,
cal dynamism, yields another case for insufficient which inevitably yield spillover effects at the inter-
treatment of the technology factor, which can yet be firm level. Indeed, “interactive learning and collective
tempered by the inclusion of the essentials of the entrepreneurship are fundamental to the process of
Schumpeterian analysis on competitiveness. innovation” (Lundvall, 1995, p. 9) and “innovating
Schumpeter’s seminal attitude, in this respect, may firms are not islands of planned co-ordination in a
be regarded as a case of blending two crucial concepts: sea of market relations” (Oerlemans et al., 1998,
He underscores a dynamic competition for innovation p. 307). Because innovation entails the processing of
in lieu of the static price-competitiveness. Basic unit tacit knowledge, learning in this context may well
of analysis in the Schumpeterian view is the capital- necessitate “cooperation for innovation” in the form
ist business enterprise since the innovation activity of “external linkages”. Freel (2003) is a recent case
as the single most important determinant of com- study comprising a concise review of the literature
petitiveness is basically carried out at the firm-level. on this issue. On the other side of the same coin,
Innovation requires substantial R&D layouts, which, “intrafirm” technological efforts should also be taken
in turn, necessitate the existence of relatively large into account. An original case study by Galende and
firms in a particularly innovative industry. In other Fuente (2003) examines and reveals such internal
words, Schumpeter draws attention to the importance factors of innovation at intrafirm level.
of monopolistic and oligopolistic market structures in Consequently, it is quite reasonable to conceive
creating innovative capabilities that yield competitive firms to be developing their capabilities in confor-
edges. To be sure, this is in sharp contrast with the mity with their firm-specific characteristics especially
perfectly competitive and atomistic firms of the con- in terms of their technological efforts and skills
ventional theory. In conventional theory, there is no (Atkinson and Stiglitz, 1969). In this regard, the idea
reason to delve into the determinants of international that firms do not operate on a common production
competitiveness at the firm-level since its very as- function is one of the major premises of evolutionary
sumptions yield identical degrees of competitiveness theories (Nelson, 1981, 1987; Nelson and Winter,
abroad for the firms in the same industry. 1982). Put differently, technological capabilities of
In contrast, neoScumpeterian conception of com- firms differ from each other, and an evolutionary view-
petitiveness relies on the evolutionary aspects of in- point is promising in explaining the “permanent exis-
novation as a microeconomic process that takes place tence of asymmetries among firms, in terms of their
within the firm.1 This process is identified by “search process technologies and quality of output” (Dosi,
for knowledge and techniques, and the cumulative 1988, p. 1155). Generation of firm-level technologi-
cal capabilities are influenced by such factors as firm
1 Relying largely on the Schumpeterian framework, Nelson and size; organizational and managerial skills; adaptabil-
Winter (1982) developed a seminal theory that led to what may ity to new methods and technologies; and access to
be called the “technological capability approach”, which concep- skills from the market, external technical information
tualizes firms in the face of imperfect knowledge of technological
possibilities (Lall, 2000). With a focus on technology generation
and support, and embodied technology (Lall, 1992,
in Latin America, the collection by Katz (1987) shows up as a p. 169). In this connection, three types of capabili-
leading work within the same domain of inquiry. ties interact with each other to yield competitiveness:
412 E. Özçelik, E. Taymaz / Research Policy 33 (2004) 409–424

Capabilities in production, investment and innovation. industries in which the firms-in-question are located,
In the case of developed countries, the feeding se- etc.
quence generally runs from innovation to investment The causality postulated to run from technologi-
to production. However, developing countries mostly cal factors to export performance has usually been
transfer technology, and thus “they usually reverse verified. In general, studies have been successful in
the sequence and use production capability as the demonstrating that there exist major exporters who
foundation for developing capabilities in investment relate their R&D activity more to exporting over time
and innovation” (Dahlman et al., 1987). Therefore, (Lall and Kumar, 1981), that the propensity-to-export
once the conventional assumption that firms operate of firms engaged in R&D tends to be higher than that
on a common production function is dropped, vary- of the entire branch to which they belong (Hirsch
ing degrees of innovativeness among firms become and Bijaoui, 1985), and that the variation in ex-
self-evident especially in developing economies. port sales are well explained by the variations in
However, the set of arguments above does not imply R&D-to-sales ratio (Hirsch et al., 1988). Exporting
that the priority of developing countries is production firms were also found to have higher labor productivi-
systems rather than knowledge systems. Technology ties (Abd-el-Rahman, 1991), technology showed up as
transfer alone cannot be a long-term development a quite crucial factor in explaining the export behav-
strategy; technology creation must be learned as ior of firms in medium and low-technology industries
well. Transformation from under-development to a (and not in high-tech ones) especially in the case of
developed economy necessitates a due attention to developing countries (Kumar and Siddarthan, 1994),
national systems of innovation in the long-term. Bell and innovating and non-innovating firms turned out
and Albu (1999), for instance, convincingly elaborate to behave differently both in terms of the probability
on “the need to focus on systems of knowledge ac- of exporting and the level of exports implying that
cumulation, rather than just production systems” in the capacity to innovate fundamentally affected the
developing countries. Technological capability, thus, export performance of firms (Wakelin, 1998). More-
must be considered an end per se, rather than a simple over, a recent study demonstrates that innovativeness
by-product of production and investment activities. is conducive to competitiveness in export markets
in general, and that significant differences emerge
between not only the firms of a country, but also
3. Firm-level studies on the determinants of the countries themselves: (i) Innovative UK firms
export performance (as compared with the non-innovative ones) benefit
more from sectoral spill-over effects of innovation
Linking the export performance of firms with activities, while the reverse is true for non-innovative
their technological orientation, relevant studies have German firms, and (ii) scale of innovation activity and
adopted a number of measures and proxies for the de- export propensity are positively related for UK firms,
gree of success in foreign trade and the inclination to whereas a negative relation holds for German firms
innovative behavior. A variety of export performance (Roper and Love, 2002). A study on some Indian en-
measures has been regressed, via several econometric gineering and chemical firms reveals the significance
techniques, on such technology-related variables as of firm-specific determinants of export performance
R&D-to-sales ratio, R&D dummies taking the value with the conclusion that product-centered R&D in
of one if the firm has proved to be an R&D performer, engineering has a negative impact on international
formal R&D expenditures, value of the royalty and li- competitiveness, and that process-centered R&D in
censing fees abroad, the percentage of equity held by chemicals, while not taking India to world stan-
foreign firms (as a measure of access to technology dards of efficiency, does not constitute a handicap in
via direct foreign investment), dummies that distin- terms of product quality and design characteristics
guish between the producers of capital goods and of (Lall, 1986). A study examining Italian manufac-
other goods, labor and capital productivities, skill and turing firms not only reveals the important impact
capital intensity of operations, imports of technol- of R&D activities on export performance, but also
ogy, number of innovations used or produced in the yields that product innovations are more contributive
 E. Özçelik, E. Taymaz / Research Policy 33 (2004) 409–424 413

in the case of small firms, whereas process innova- whether exporting generates efficiency gains. Relying
tions enhance the exports of medium-sized and large on this firm-level panel-data study, which finds that
firms (Sterlacchini, 1999, 2001). Similar results are more innovative firms become exporters and not vice
obtained by Nassimbeni (2001) who finds that export versa, we are to be content with the present frame-
propensity of “small” Italian firms is affected most by work at least for the time being. Our cross-sectional
the ability to generate new products and to develop data at hand comes from the only available innova-
inter-organizational relations, while technological tion survey conducted for the first time at the end of
profiles matter much less. While providing evidence 1998. In this regard, with the accumulation of new
for the expected consequence that export intensity data through prospective surveys in the near future, a
of innovative firms exceed that of the non-innovative time-series dimension may also be available, in which
ones, another study on Italian firms focuses on the case dealing with two-directional causalities becomes
negative impact of exchange rate devaluations on the rigorously feasible.
conduciveness of technological capabilities to export Finally, inclusion of technology-related variables
performance (Basile, 2001). (as potent determinants of export performance) into
The cruciality of technical collaborations and in- any model of international competitiveness is in-
digenous R&D efforts and yet the negative impact of evitable. However, they alone cannot account for the
capital intensity on the export performance of the firms entirety of inter-firm variations. Thus, any such model
(in the Indian automobile industry) have also been ev- is to incorporate some other explanatory variables,
idenced (Bhat and Sethuraman, 1995). Last but not the whereby the wider comprehension can help improve
least, it has also been shown that not only the influ- the empirical results. This, in turn, necessitates an
ence of R&D on both export propensity and growth elaboration through industrial organization with an
is significantly positive, but also there exist reciprocal eye to international economics. At this point, one of
relationships between R&D and exports (Zhao and Li, the most inextricable tasks in front of an empirical
1997). researcher is to take into account such factors as
As a matter of fact, examination of a possible firm size, technical manpower, industrial concentra-
causality in the opposite direction (i.e. from export- tion, product differentiation, unit labor costs, wages,
ing behavior towards technological improvement) is markups, profitability, expenditures on advertising,
a desideratum for studies such as this. The so-called etc. as other possible determinants of export perfor-
“learning-by-exporting” literature has been developed mance, which have usually shown up as significant
in that context. The idea that export-oriented policies regressors in empirical literature (Glejser et al., 1980;
may well expand technological frontiers (especially in Lall and Kumar, 1981; Hirsch and Bijaoui, 1985;
the case of developing countries) provides a rationale Lall, 1986; Hirsch et al., 1988; Abd-el-Rahman, 1991;
for this domain of research. For instance, Dahlman Kumar and Siddarthan, 1994; Bhat and Sethuraman,
and Westphal (1982) provide evidence that Korean 1995; Zhao and Li, 1997; Wakelin, 1998; Wignaraja,
firms were able to generate improvements in product 2002).
quality and design as well as in productivity thanks to
participating in exporting activity. Kırım (1990), in a
case study of 659 largest Turkish manufacturing firms, 4. Descriptive aspects of the Turkish
argues that the attempt of export-led growth during manufacturing industry
the 1980s had significant impacts on the direction of
in-house technological change, albeit not on the rate In the 1960s and 1970s, Turkey adopted an im-
of R&D. All the same, while admitting the possibil- port substitution industrialization strategy, which was
ity of an opposite or a two-directional causality, we able to generate a process of rapid yet unsustain-
would still rather confine the scope of this study to able economic growth. Towards the end of 1970s,
a framework of exporting-by-learning. Examining, a balance-of-payments crisis led the Turkish govern-
on the one hand, whether exporter firms are more ment to implement a stabilization and structural ad-
efficient than their domestic non-exporter counter- justment program, the essence of which turned out to
parts, Clerides et al. (1996), on the other, inquire into be an export-led growth strategy in the 1980s. In this
414 E. Özçelik, E. Taymaz / Research Policy 33 (2004) 409–424

context, ready-made tools were plentiful export sub- It is in the light of above-mentioned arguments
sidies cum incessant real devaluations. In 1983, ex- that we intend to inquire into whether technological
port incentives came up to 36% of the export revenue efforts of Turkish manufacturing firms are conducive
(Uygur, 1991), and from 1979 to 1984 Turkish lira to their export performance. With a weak “national
was devalued against USD by 100% in real terms. The system of innovation” and a negligibly small share
consequence was an export boom in the period under of R&D expenditures in GNP, Turkey is an inter-
consideration. The boom-in-question, however, was esting case of analysis. Her manufacturing industry
achieved at the expense real wages. Indeed, real wage is relatively dynamic and productive implying that
rates (deflated by the consumer price index) could not long-term growth and international competitiveness
catch up with their 1978 levels before the early 1990s are most likely to arise therefrom.2 If “individual”
(Taymaz, 1999). technological efforts of manufacturing firms are
Dramatic real wage deterioration created as such shown to play a role in enhancing export intensi-
was accompanied by a non-increasing gross fixed ties, this may well imply much higher benefits to be
capital formation (GFCF) in manufacturing during reaped under a well-established system of innova-
the 1980s. This was a seemingly controversial phe- tion. Perhaps, it is in this way that the conventional
nomenon since it was the manufacturing industry ready-made attitudes towards international competi-
that led the others in the process of export boom. tion policy can be replaced by technology-centered
At this point, Dani Rodrik solves the paradox in a priorities.
comparative study on the differences between the Our main data set comes from the Innovation Sur-
export-led growth strategies of South Korea and Tai- vey that was conducted the first time by the State In-
wan on the one hand and Turkey and Chile on the stitute of Statistics (SIS) of the Republic of Turkey
other: “[M]odest export booms in Turkey and Chile in 1998. The survey covers the innovation activities
in the 1980s have required cumulative exchange of firms in the period 1995–1997. The questionnaire
rate depreciations contemporaneously of the order of is compatible with the Community Innovation Sur-
100%, a change in relative prices vastly in excess of vey of the European Union, and defines “technologi-
anything observed in East Asia” (Rodrik, 1995, p. cal innovation”, as “technologically new products and
2). The two East Asian countries in question were processes or significant technological improvements
able to blend export-orientation with successful in- in products and processes”.3 An innovation has been
vestment and technology strategies, whereas Turkey implemented if it has been introduced on the market
and Chile solely relied on devaluations and export (product innovation) or used within a production pro-
incentives without any significant efforts to feed up cess (process innovation). Innovation is explicitly de-
the productive infrastructure. fined at the firm-level, i.e. “innovation occurs when a
In this regard, it is a quite convincing contention firm implements a new or improved product or pro-
that “one way of differentiating competitively strong cess which is technologically novel for the firm, not
and weak countries is by the methods they adopt to for the market”. In order to check the quality of re-
gain the competitive edge–productivity increases or re- sponses, firms who claim to be innovative are asked
duced wages” (Haque, 1995, p. 23). While the former to describe their (at most three) product and process
method, by and large, necessitates a search for tech-
nological development in the form of R&D activities
2 See Pamukçu (2003) for an analysis of the determinants of
(as implemented by South Korea and Taiwan, “Asian
innovation-related decisions of plants in the Turkish manufacturing
tigers” as of now); the latter may, for instance, be ac- industry.
complished through a real devaluation of the currency 3 The questionnaire includes 24 questions on product and process

(as a ready-made tool embraced by Turkey and Chile). innovations, sales revenue and employment, internet access, R&D
It turns out that genuine international competitiveness activities and expenditures, aims of innovative activities, sources of
is a matter of innovativeness, which has nothing to information for innovation, obstacles to innovation, research collab-
orations, R&D support, patenting behavior, and organizational in-
do with cost-reductions-via-devaluations (or by way novations. For more information about the survey, see the web site
of artificial incentives, like export subsidies, for that of the SIS, http://www.die.gov.tr/konularr/teknollojikYenilik.htm
matter). (in Turkish). For definitions see OECD (1996).
 E. Özçelik, E. Taymaz / Research Policy 33 (2004) 409–424 415

innovations. Respondents’ descriptions seem to sug- Among the explanatory variables, our focus of at-
gest that their responses are quite reasonable. tention is a number of technology-related variables,
A sample of about 4000 firms stratified by size which we utilize in separate regressions due to the
and industry category was asked to complete the correlations among them. Two basic innovation dum-
questionnaire. The response rate was about 55%. The mies, PRODUCT and PROCESS, are equal to one
SIS performed a non-response analysis and estimated if a firm reported to have introduced any product
sample weights for each respondent. Consistent with or process innovation, respectively, and zero other-
other science and technology indicators, the aggre- wise. Another innovation dummy, INNOVATOR, is
gate innovation rate in the manufacturing industry is equal to one if a firm reported to have introduced
found to be quite low compared to the EU countries any innovation (product or process). INNOVATOR is
(only 23.0%). This finding suggests that firms do not likely to serve as a better explanatory variable since
overstate their innovative performance. The size and PRODUCT and PROCESS are correlated. Share of
sectoral distribution of innovators are consistent with R&D expenditures in total sales, or R&D inten-
a priori expectations: the proportion of innovators sity (RDINT), is another technology-related variable
increases monotonically by size, and firms in chem- (which is correlated with PRODUCT and PROCESS).
icals, non-metallic minerals, metal, and engineering We determine the impact of technology transfers with
industries tend to be more innovative. the inclusion of a dummy (TECHTRAN), which is
The innovation database was matched with the equal to one if the firm acquired technology through
1995–1997 data from the Annual Survey of Manufac- license or know-how agreements, and zero otherwise.
turing Industries as part of a National System of Inno- Finally, regional spill-over effects are also taken into
vation Project (for details, see Taymaz, 2001), where account by REGINN, which is the ratio of innovators
a preliminary version of the econometric analysis re- to the total number of firms in the region (province)
ported in this paper was conducted as a background where the respective firm is located.
study. Size of firms has long been a conventional variable
Export intensity equations are estimated to find out of interest in the empirical literature insofar as its im-
the determinants of export performance. It is obvious pact upon (export) performance is concerned. Number
that the whole sample consists of many firms that do of employees is a conventional measure of firm size.
not export at all. Hence, the dependent variable (share Our regressions also include firm size as an explana-
of exports in total sales, EXPINT) assumes the value of tory variable: SIZE X+ is equal to 1 if the firm em-
zero for non-exporter firms, and positive values for the ploys X or more employees, zero otherwise, where X
exporters. Such being the case, the most appropriate is set at 25, 50, 150, 250 and 500 to end up with five
way of obtaining unbiased and consistent estimators dummies for measuring the influence of firm size. The
is the so-called Tobit estimation procedure, which is rationale behind utilizing size dummies (instead of in-
thus utilized to obtain the inferential results.4 dividual sizes of firms) is to avoid problems that might
arise from potential non-linearities between EXPINT
4 As mentioned by one of our referees, the Tobit estimation and size (number of employees) of firms.
imposes a proportionality restriction on the effect of each regressor It is generally agreed that use of factors also plays
on the probability of exporting and export intensity. The validity an important role. Among our explanatory vari-
of this restriction can be tested against an alternative unrestricted ables, thus, we have capital intensity (CAPINT) (or
form comprising separate Probit and truncated regression models capital-to-labor ratio, which is proxied by the loga-
(Greene, 2003, p. 770). In such a case, the probit model tests the
effects on the “capacity” to export (does the firm export or not),
rithm of the ratio of depreciation allowances to the
and the second model tests the effects on export intensity (how number of employees). Logarithm of the real wages
much does the firm export). Appendix A presents the maximum (WAGE) is also included in order to capture the
likelihood estimation results of the selection model. Since both probable influence of the quality of labor.
methods generate qualitatively same results, we discuss here only Export intensities may also be related to ownership
the Tobit estimation results (the correlation coefficient between
export intensities predicted by two models is 0.968). All other
structures in the firms. Three shares, sum of which is
maximum likelihood estimation results of the sample selection equal to unity, are to be considered in this respect: (i)
models are available from the authors upon request. public ownership, (ii) (domestic) private ownership,
416 E. Özçelik, E. Taymaz / Research Policy 33 (2004) 409–424

and (iii) foreign ownership. In our regression analy- Export intensity is higher for innovators (16.6%)
sis, two of these shares serve as non-omitted explana- than for non-innovators (11.8%). The difference be-
tory variables (PUBLIC and FOREIGN), which are tween innovators and non-innovators is quite obvi-
expected to yield significant impacts with respect to ous so far as their size is concerned (198 versus 89
domestic private ownership. employees on the average, respectively). The well-
Along with advertisement intensity, ADVERINT known Schumpeterian hypothesis, in this regard,
(advertisement expenditures/sales), two variables are seems to be descriptively supported. Interpretation
incorporated into regressions in order to determine of the capital intensity variable is interesting: Inno-
whether the structure of production affects export per- vators use capital-intensive production techniques,
formance: Share of inputs subcontracted to suppliers while non-innovators rely, by and large, on labor
(SUBINPUT) and share of output subcontracted by rather than capital. Technology transfer is practiced
customers (SUBOUT). Finally, composition of the la- more commonly by the innovators than by non-
bor force is also taken into account by including the innovators (7.6 versus 2.6%, respectively). This may
shares of “administrative”, “technical” and “female” indicate that technology transfer and innovativeness
personnel in all employees (ADMINSH, TECHSH and are complements. Logarithm of real wage is 2.5
FEMALESH, respectively). in innovators and 1.76 in non-innovators indicating
Since we are interested in inter-firm variations, that innovators pay much higher wages. Ownership
descriptive statistics at the firm-level are rather in- of the firms, advertisement intensity, subcontracted
formative before analyzing the inferential results. input and output shares, as well as the shares of
Means of the variables are separately provided in administrative and technical personnel differ only
Table 1 for (i) the whole sample (all firms), (ii) the slightly between innovators and non-innovators. Of
firms that reported to have introduced product and/or course, these variables may still play significant
process innovations (innovators), and (iii) the firms roles in the determination of inter-firm variations
without innovations (non-innovators). All statistics in export intensity. Finally, it is noteworthy that
are weighted by sample factor weights. the share of female personnel is seven percentage

Table 1
Variable definitions and descriptive statistics
Label Definition All firms Innovators Non-innovators

EXPINT Export/sales ratio 0.129 0.166 0.118

PRODUCT Product innovator 0.149 0.649 0.000
PROCESS Process innovator 0.182 0.790 0.000
INNOVATOR Innovator 0.230 1.000 0.000
RDINT RD expenditures/sales ratio (102 ) 0.176 0.742 0.007
RDGINN Regional innovation intensity 0.332 0.372 0.320
SIZE Number of employees 114 198 89
CAPINT (ln) depreciation allowances per employee −0.254 0.462 −0.502
TECHTRAN Technology transfer dummy 0.038 0.076 0.026
WAGE (ln) Real wage rate 1.952 2.501 1.764
PUBLIC Share of public ownership 0.054 0.036 0.060
PRIVATE Share of private ownership 0.931 0.936 0.929
FOREIGN Share of foreign ownership 0.015 0.028 0.011
ADVERINT Advertisement expenditures/sales ratio 0.005 0.009 0.004
SUBINPUT Subcontracted output/sales ratio 0.042 0.036 0.045
SUBOUT Subcontracted input/inputs ratio 0.063 0.042 0.071
ADMINSH Share of administrative personnel 0.202 0.212 0.199
TECHSH Share of technical personnel 0.066 0.069 0.064
FEMALESH Share of female personnel 0.216 0.164 0.234
Source: EXPINT, PRODUCT, PROCESS, INNOVATOR, RDINT, REGINN and SIZE from SIS, Innovation; Survey, 1995–1997. Other
variables, SIS, Annual Survey of Manufacturers, 1995–1997.
Note: Weighted means.
 E. Özçelik, E. Taymaz / Research Policy 33 (2004) 409–424 417

Table 2
Sample characteristics by industry
31 32 33 34 35 36 37 38 39

Food Textile Wood Paper Chemicals Non-metallic Metal Engineering Other

EXPINT 0.070 0.246 0.028 0.033 0.056 0.091 0.096 0.072 0.052
PRODUCT 0.088 0.092 0.195 0.191 0.357 0.171 0.146 0.215 0.145
PROCESS 0.149 0.124 0.090 0.205 0.254 0.293 0.311 0.244 0.203
INNOVATOR 0.167 0.154 0.195 0.236 0.390 0.332 0.369 0.319 0.203
RDINT 0.000 0.001 0.000 0.000 0.002 0.008 0.001 0.003 0.000
RDGINN 0.304 0.340 0.341 0.331 0.344 0.358 0.309 0.340 0.329
SIZE 99 132 75 114 126 109 212 84 119
CAPINT −0.270 −0.376 −0.919 0.499 0.304 0.012 0.016 −0.435 0.010
TECHTRAN 0.016 0.017 0.117 0.013 0.093 0.031 0.030 0.084 0.110
WAGE 2.365 1.415 1.700 2.761 3.053 1.899 2.467 2.115 2.635
PUBLIC 0.149 0.015 0.070 0.097 0.059 0.050 0.037 0.048 0.196
PRIVATE 0.818 0.979 0.929 0.898 0.907 0.933 0.957 0.935 0.804
FOREIGN 0.033 0.006 0.001 0.005 0.034 0.017 0.006 0.018 0.000
ADVERINT 0.003 0.004 0.015 0.007 0.009 0.007 0.003 0.005 0.002
SUBINPUT 0.003 0.079 0.008 0.032 0.016 0.009 0.014 0.032 0.006
SUBOUT 0.003 0.117 0.006 0.047 0.040 0.000 0.043 0.049 0.000
ADMINSH 0.298 0.149 0.158 0.251 0.289 0.186 0.178 0.204 0.267
TECHSH 0.055 0.066 0.041 0.131 0.064 0.050 0.068 0.075 0.028
FEMALESH 0.120 0.381 0.081 0.080 0.122 0.066 0.027 0.098 0.083
Source: See Table 1.

points higher for non-innovators as compared to intensity variable indicates that food, textiles, wood
innovators. and engineering sectors are labor-intensive. The av-
Insofar as our sample is concerned, some descrip- erage export intensity of these labor-intensive sectors
tive characteristics of the nine sub-sectors are reported equals 10.5%, whereas that comes up to only 6.6%
in Table 2. Unsurprisingly, one of the most discernible for the remaining capital-intensive ones. This sec-
differences is the relatively much higher average ex- toral aspect, of course, is somewhat supportive of the
port intensity of the textile sector, traditional export factor-endowment theory since Turkey is most likely
leader. The average export intensity of the eight sec- to be a labor-abundant country. Technology transfer
tors, excluding textile, equals 6.23%, whereas textile and export intensity seem to be independent from each
exports are about 25% of total textile sales. Interest- other. When it comes to investigate real wages in the
ingly however, the average of the percentage of firms sectors, it is markedly the manufacture of chemicals
that introduced product and process innovations is (the leader in innovativeness) that pays the highest
considerably low for textiles (15.4%); indeed, it is the wages. The lowest wages, on the other side, are paid
lowest value among the nine sub-sectors. Average per- by textiles, the traditional export leader. These two
centage of innovators is the highest in the case of the descriptive aspects as to real wages somewhat confirm
manufacture of chemicals (39%), which in turn has a conventional a priori expectations.
modest export intensity (5.6%). In the light of these We discuss our econometric results in the following
facts, it can be argued that, at the sectoral level, in- section. Before that, however, we had better attract
novativeness per se does not seem to necessarily con- the attention of the reader to an important aspect
tribute to exports. Moreover, R&D intensities of all of the models we estimated by the Tobit procedure.
sectors are negligibly small. But it must be noted that All regressions, on which we elaborate next, include
the share of R&D in overall innovation expenditures sectoral dummy variables at the ISIC two-digit level.
is about 10–15% in Turkey (Taymaz, 2001). Sectoral dummies not only capture some unobserv-
The average size of firms in the sectors is seem- able aspects of the sectors, but also reflect, at least
ingly uncorrelated with export intensity. The capital partially, the impact of factor proportions. Our results
418 E. Özçelik, E. Taymaz / Research Policy 33 (2004) 409–424

suggest that such dummies are quite explanatory as remains intact irrespectively of the inclusion of basic
far as the export performance is concerned at the innovation variables (PRODUCT, PROCESS, INNO-
sectoral level. Coefficients of the dummies tend to be VATOR) and R&D intensity (RDINT). In other words,
relatively lower for capital-intensive sectors. In other explanatory variables are insensitive to changing tech-
words, export-intensity of labor-intensive sectors is nology variables.
relatively higher. Keeping this important sectoral as- It is up to 150 employees that a larger firm size
pect in mind, we should also draw attention to the implies a significantly higher export performance. Be-
intra-sectoral heterogeneity of the firms. Needless to yond that size, export intensity seems to be indepen-
say, a firm-level analysis is desirable in this regard. dent of the number of employees. Those firms which
This is what we attempt in what follows in order to employ more than 500 employees also tend to have
shed light on the micro-dynamics between competi- higher export intensities; however, such large firms are
tiveness and innovativeness. quite few in number, and they may be operating in
relatively low-technology industries, while basically
producing for export markets. Besides, capital inten-
5. Determinants of international competitiveness: sity (CAPINT) and wage (WAGE) variables may be
estimation results interpreted together: The former is significantly con-
ducive to export performance, whereas the latter has
It is in Table 3 where Tobit estimation results for no impact. This could be related to labor quality. In
firm-level determinants of export intensity are pre- this regard, if Turkey is a labor-abundant country, then
sented. The first three models involve “all firms”; the positive influence of capital intensity on the ex-
that is, both innovators and non-innovators. Model port performance of Turkish firms turns out to be rem-
4 comprises innovators only, whereas Model 5 is iniscent of the well-known Leontief paradox, albeit
for non-innovators only. The data for the dependent on different grounds (that is, as contrasted with the
variable (export intensity) belongs to the year 1997, capital-abundancy of the United States against the cap-
whereas the explanatory variables are measured in ital intensity of her import-competing sectors).
terms of averages in the period 1995–1997. Hence, Lower public ownership along with higher for-
it is reasonable to expect to capture the lagged ef- eign ownership (PUBLIC and FOREIGN) implies
fect of explanatory variables on export performance. higher export intensity. In this regard, state-owned
In the models considered, basic innovation variables enterprises (establishment objective of which was
(PRODUCT and PROCESS) and R&D intensity vari- import substituting industrialization) can be said to
able (RDINT) have been cautiously incorporated. be naturally less export-oriented, whereas existence
Since “innovation” is the expected resultant of R&D of foreign share-holders seems to be influential on
activities, they tend to exhibit high correlations with exporting efforts. Negatively significant impact of ad-
each other. Therefore, innovation and R&D variables vertisement intensity (ADVERINT) is an interesting
have been separately utilized within the regressions. result. A presumable interpretation is that advertis-
Of course, we did the same in the case of the IN- ing basically targets the home market mainly for
NOVATOR variable. Furthermore, Models 4 and 5, consumer goods. Put differently, those firms with
which have an identical set of explanatory variables, higher advertisement intensities are essentially pre-
were utilized basically for comparing and contrasting occupied with meeting the domestic demand. On the
innovators and non-innovators. In those models, we other side, those firms which subcontract their in-
exclude RDINT since respective data are available puts to “subcontractors” tend to have higher export
only for innovators. In what follows, estimation results intensities, whereas subcontractor firms have lower
are discussed for “all firms” in the first place. Then, export intensity implying that international subcon-
“innovators” and “non-innovators” are compared and tracting is not well-developed. Composition of labor
contrasted. force is also important: Share of female personnel
So far as “all firms” in the sample are concerned in all employees (FEMALESH) is conducive to ex-
(Models 1–3), one of the most outstanding results is port performance, whereas shares of administrative
that statistical significance of the explanatory variables and technical personnel (ADMINSH and TECHSH)
Table 3
Determinants of export intensity, 1995–1997 (Tobit estimation)
Model 1 (all firms) Model 2 (all firms) Model 3 (all firms) Model 4 (innovators) Model 5 (non-innovators)
Coefficient t-statistic Coefficient t-statistic Coefficient t-statistic Coefficient t-statistic Coefficient t-statistic

PRODUCT 0.028 0.900

PROCESS 0.052 1.810∗

E. Özçelik, E. Taymaz / Research Policy 33 (2004) 409–424

INNOVATOR 0.065 2.641∗∗
RDINT 2.575 3.519∗∗
REGINN 0.160 1.880∗ 0.161 1.899∗ 0.188 2.239∗∗ −0.024 −0.308 0.219 1.444
CAPINT 0.031 3.780∗∗ 0.031 3.752∗∗ 0.033 4.086∗∗ 0.035 3.403∗∗ 0.031 2.459∗∗
TECHTRAN 0.010 0.190 0.007 0.134 0.006 0.121 0.050 1.045 0.055 0.574
WAGE 0.005 0.590 0.005 0.630 0.006 0.717 0.009 1.164 0.005 0.318
ADVERINT −2.003 −2.630∗∗ −1.947 −2.565∗∗ −1.877 −2.486∗∗ −1.625 −2.528∗∗ −1.996 −1.356
PUBLIC −0.367 −6.180∗∗ −0.367 −6.177∗∗ −0.378 −6.379∗∗ −0.171 −2.374∗∗ −0.477 −5.127∗∗
FOREIGN 0.448 4.570∗∗ 0.452 4.613∗∗ 0.449 4.586∗∗ −0.042 −0.426 0.851 4.995∗∗
SUBINPUT 1.117 9.120∗∗ 1.107 9.046∗∗ 1.111 9.084∗∗ 0.971 6.011∗∗ 1.110 5.995∗∗
SUBOUT −0.354 −4.380∗∗ −0.358 −4.439∗∗ −0.361 −4.426∗∗ −0.426 −4.240∗∗ −0.361 −2.978∗∗
ADMINSH −0.226 −2.500∗∗ −0.230 −2.538∗∗ −0.228 −2.522∗∗ −0.273 −2.659∗∗ −0.160 −1.123
TECHSH −0.244 −2.190∗∗ −0.246 −2.202∗∗ −0.243 −2.188∗∗ 0.186 1.268 −0.361 −1.868∗
FEMALESH 0.332 6.010∗∗ 0.333 6.036∗∗ 0.343 6.203∗∗ 0.496 6.453∗∗ 0.325 3.932∗∗
SIZE 25+ 0.165 3.330∗∗ 0.165 3.325∗∗ 0.166 3.356∗∗ 0.071 0.817 0.174 2.509∗∗
SIZE 50+ 0.102 4.040∗∗ 0.101 4.020∗∗ 0.111 4.420∗∗ 0.039 1.284 0.122 3.073∗∗
SIZE 150+ 0.022 0.620 0.023 0.624 0.026 0.721 0.042 1.105 0.031 0.536
SIZE 250+ −0.008 −0.160 −0.008 −0.167 −0.005 −0.099 0.023 0.486 −0.055 −0.675
SIZE 500+ 0.090 1.860∗ 0.092 1.886∗ 0.095 1.956∗ 0.011 0.254 0.164 1.896∗
Pseudo R2 23.21 23.22 23.42 32.95 24.88
log likelihood −1057.5 −1057.4 −1054.7 −351.54 −554.47
No. of observations 1529 1529 1529 683 846
No. of exporters 968 968 968 515 453
All models include sectoral dummy variables at the ISIC two-digit level.
∗ Means statistically significant at the 10% level, two-tailed test.
∗∗ Means statistically significant at the 5% level, two-tailed test.

419
420 E. Özçelik, E. Taymaz / Research Policy 33 (2004) 409–424

have a negative influence. These findings suggest that SUBINPUT and FEMALESH are positive, whereas
Turkish manufacturing firms are more competitive in those of PUBLIC and SUBOUT are negative.
activities that require home-based skills, but less com- Surprisingly, there are also two impotent variables:
petitive in activities that require technical skills. Neither WAGE nor TECHTRAN has to do with the
Since it is the principal aim of this study to deal with export intensity of innovators or that of non-innovators
the possible impacts of technological capabilities on (or that of the sample as a whole). Moreover, even
the export performance of firms, technology-related though REGINN has a somewhat significant impact in
variables must be discussed in detail. With respect to the case of all firms (Models 1–3), it interestingly turns
Model 1, process innovations are conducive to exports, out to be an insignificant regressor when regressions
whereas product innovations do not have a significant are run separately for innovators and non-innovators.
influence. However, it is to be noted that PRODUCT To be sure, REGINN is higher for innovators; and if
turns out to be significant determinant, when PRO- innovators actually concentrate within certain regions,
CESS is dropped from Model 1. Positive significance the significant impact in the case of all firms may be
of INNOVATOR and RDINT can be observed in Mod- disappearing when the whole sample is grouped into
els 2 and 3, respectively.5 Besides, one should also innovators and non-innovators.
consider the facts that (i) regional innovation intensity Despite such important similarities, differences are
(REGINN) somewhat contributes to the export perfor- no less between innovators and non-innovators. The
mance, and (ii) technology transfers through license most prominent difference arises from size. Size mat-
or know-how agreements (TECHTRAN) seem to have ters only for non-innovators. Once non-innovators turn
no significant impact. out to be innovators, number of employees does not
One of the objectives of this study is to detect influence exporting behavior. Up to 150 employees in
the similarities and differences between innovators non-innovators, a larger firm size yields a significantly
(Model 4) and non-innovators (Model 5). Irrespec- higher export performance. Beyond that size, export
tively of being an innovator or non-innovator, there are intensity seems to be independent of size. In this con-
a number of explanatory variables that significantly nection, the average number of employees in innovator
affect export intensity in the same direction. Interest- and non-innovator firms may be important: 198 and
ingly, those variables have significant impacts in the 89, respectively. An average innovator is already twice
same direction for the whole sample, too. In this sense, larger than an average non-innovator, which may im-
these variables may be regarded as the most potent ply that most innovators have already surpassed a size
determinants of firm-level export performance in the threshold beyond which export intensities have noth-
Turkish manufacturing industry: Impacts of CAPINT, ing to do with firm size. Moreover, like in the case of
all firms, non-innovators that employ more than 500
5 Marginal effects of basic innovation variables as well as R&D employees also tend to have higher export intensities;
intensity have also been calculated for their mean values. In this however, there are only a few so large non-innovators,
respect, for an exporter firm, introducing a product innovation which may be operating in relatively low-technology
raises export intensity by 1.2% point; and a process innovation
does the same by 2.3% point (Model 1). For instance, an aver-
export industries.
age firm with an export intensity of 13% would be able to raise Advertisement intensity (ADVERINT) is an in-
it up to 16.5% (a 3.4% point increase), if it introduced a prod- significant regressor for non-innovators, while it
uct innovation along with a process innovation. Similarly, a 1.0% significantly yet negatively influences the export in-
increase in R&D intensity generates a 1.1% increase in export tensity of innovators. The interpretation we set forth
intensity (Model 3). On the other side, when a non-exporter firm
introduces a product innovation, its probability of becoming an
for all firms may also be valid for innovators: Since
exporter increases by 3.1% point; whereas the respective contri- they may basically produce to meet domestic de-
bution of a process innovation is 5.8% point (Model 1). These mand, it is no surprise that advertising by innovators
are quite substantial because the proportion of exporters is about targets the home market. Exporter innovators estab-
37% (weighted average). Consequently, a 1.0% increase in R&D lish different ways of connections with their foreign
intensity yields a 2.9% increase in the probability of becoming
an exporter (Model 2). To be sure, implementation of a national
customers.
technology policy is to seriously take into account such marginal In contradiction, while the share of foreign
effects as informational guidelines. ownership (FOREIGN) does not affect the export
 E. Özçelik, E. Taymaz / Research Policy 33 (2004) 409–424 421

intensity of innovators, it is conducive to that of the As a prominent difference between innovators and
non-innovators. Prior motive of foreign share-holders non-innovators, size does not matter for the former
may be export-orientation rather than innovativeness. insofar as their export performance is concerned. Tak-
Moreover, innovators may have already attained a par- ing into account the pertinently large size of inno-
ticular level of international competitiveness to which vators, it may be argued that number of employees
foreign ownership does not have much to contribute. contributes to export performance only up to a certain
Finally, with negative impacts, share of administra- size threshold (which is 150 for non-innovators). Once
tive personnel (ADMINSH) is significant only for in- non-innovators turn out to be innovators, exports be-
novators, and share of technical personnel (TECHSH) come independent of the firm size.
is somewhat significant only for non-innovators. Export performance of non-innovators is positively
Non-innovators can have competitive disadvan- influenced by the share of foreign ownership, while
tages in those products which use relatively skilled that of innovators remain intact with respect to the
labor. same variable. Foreign impulse to improving exports
is an important factor for non-innovators, whereas
innovators may have already developed their own pe-
6. Concluding remarks culiar motives irrespectively of foreign or domestic
ownership. The structure of international marketing
One of our preliminary contentions was that links, thus, can also be different between innovators
firm-level analysis is required to understand important and non-innovators.
dynamics between innovativeness and competitive- With its negative impact, public ownership shows
ness. By their very nature, technological processes up as one of the most potent determinants of export
are sequences of cumulative adaptation experienced intensity in the Turkish manufacturing firms. Put dif-
within the firm. Accordingly, it is a good idea to con- ferently, exports basically arise from the private sector,
ceive the generation of competitive advantages within and this is no surprise since state economic enterprises
a micro-dynamic context. The entire process, thus, were established to produce for the domestic market
should be perceived as a two-sided propagation rather with the objective of import substitution.
than a unilateral feeding from innovativeness towards Two other important aspects of Turkish manufac-
competitiveness. To be frank, we do not claim to have turing firms are observed with respect to the “share
comprehensively examined the dynamics as such. of inputs subcontracted to suppliers” and “share of
The cross-sectional nature of the data set prevented output subcontracted by customers”; the former with
us from doing so. As innovation surveys accumulate a positive and the latter with a negative impact upon
over time, it will be possible to add a time-series export intensities. Exporting behavior seems to be
dimension to the analysis, in which case we will ea- characterized by two steps: First, purchase of unfin-
gerly undertake the work of modeling simultaneous ished products within local networks; second, their
causations. This is, indeed, the genuine challenge in sale abroad after processing. In other words, subcon-
front of us. tractor manufacturers basically sell at home, and what
It has been verified to a great extent that inno- they sell is exported after being processed by non-
vations and R&D activities are crucial for the in- subcontractors.
ternational competitiveness of Turkish manufacturing Last but not the least, persistent insignificance of
firms. However, technology transfers (through license real wage is also worthwhile considering. Turkey
or know-how agreements) do not show up as signif- has conventionally implemented devaluations (basi-
icant determinants of export performance. Therefore, cally to accommodate high inflation) with a hope to
promotion of in-house innovativeness seems a good improve her international competitiveness via real
idea insofar as the priorities of a rational technology cost reductions (e.g. the alleged advantage of “cheap
policy is concerned. All the same, technology trans- labor”). Nevertheless, real wage was able to signif-
fers must not be easily overlooked since own innova- icantly affect export intensity in none of the five
tion activities and technology transfers are likely to be regressions we considered. In contradistinction, cap-
“complementary” processes. ital intensity turned out to be invariably significant
422 E. Özçelik, E. Taymaz / Research Policy 33 (2004) 409–424

and conducive in the very same regressions. Along Acknowledgements

with the important impact of technology-related fac-
tors, these two results regarding wages and capital Earlier versions of this paper were presented at
intensity yield a quite crucial warning to be obeyed two international conferences (ERC/METU Interna-
by the policy-makers at all costs: Turkey as well as tional Conference in Economics V, Ankara, September
similar developing countries must escape from the 10–13, 2001 and Economic Research Forum 8th An-
illusion of temporary export booms achieved by such nual Conference, Cairo, January 15–17, 2002). Many
ready-made tools as devaluations and export subsi- thanks go to the participants in those sessions for their
dies, and construct a coherent technology policy cum helpful comments. Cem Somel of METU deserves
a national development strategy that will generate special thanks for his valuable suggestions at a later
permanent increases in gross fixed capital formation, stage. Finally, we would like to thank two anonymous
and thus in productivity and international competitive- referees for their formative influence on producing the
ness. final version of this paper.

Appendix A

Maximum likelihood estimation of exporting and export intensity models

Exporting (1 if exports, 0 otherwise) Export intensity (export/sales ratio)

Coefficient t-statistic Coefficient t-statistic

PRODUCT 0.143 1.540 0.049 1.340

PROCESS 0.208 2.460∗∗ 0.079 2.390∗∗
REGINN 0.388 1.590 0.168 1.760∗
CAPINT 0.110 4.590∗∗ 0.038 4.050∗∗
TECHTRAN 0.063 0.410 0.005 0.090
WAGE 0.035 1.400 0.013 1.330
ADVERINT −4.888 −2.190∗∗ −1.998 −2.290∗∗
PUBLIC −0.716 −4.200∗∗ −0.273 −4.090∗∗
FOREIGN 1.037 3.710∗∗ 0.439 4.030∗∗
SUBINPUT 3.384 9.350∗∗ 1.288 9.170∗∗
SUBOUT −1.306 −5.690∗∗ −0.452 −5.010∗∗
ADMINSH −0.506 −1.910∗ −0.194 −1.880∗
TECHSH −1.299 −3.600∗∗ −0.463 −3.280∗∗
FEMALESH 0.995 6.070∗∗ 0.362 5.650∗∗
SIZE 25+ 0.467 3.330∗∗ 0.177 3.250∗∗
SIZE 50+ 0.493 6.550∗∗ 0.181 6.140∗∗
SIZE 150+ 0.141 1.330 0.052 1.270
SIZE 250+ −0.068 −0.490 −0.025 −0.460
SIZE 500+ 0.293 2.060∗∗ 0.080 1.440
log likelihood −747.0
LR test for independent equations 303.9∗∗
No. of observations 1529
No. of exporters 968
∗ Means statistically significant at the (10%) level, two-tailed test.
∗∗ Means statistically significant at the 5% level, two-tailed test.
 E. Özçelik, E. Taymaz / Research Policy 33 (2004) 409–424 423

References Katz, J.M. (Ed.), 1987. Technology Generation in Latin American

Manufacturing Industries. Macmillan, London.
Abd-el-Rahman, K., 1991. Firms’ competitive and national Keesing, D., 1967. The impact of research and development on
comparative advantages as joint determinants of trade United States trade. Journal of Political Economy 75, 38–45.
composition. Weltwirtschaftliches Archiv 127, 83–97. Kırım, A., 1990. Technology and exports: the case of the Turkish
Atkinson, A.B., Stiglitz, J.E., 1969. A new view of technological manufacturing industries. World Development 18, 1351–1362.
change. The Economic Journal 79 (4), 573–578. Kumar, N., Siddarthan, N.S., 1994. Technology, firm size and
Baldwin, R.E., 1971. Determinants of the commodity structure of export behavior in developing countries: the case of Indian
U.S. trade. American Economic Review 61, 126–146. enterprises. The Journal of Development Studies 31, 289–309.
Basile, R., 2001. Export behavior of Italian manufacturing firms Lall, S., 2000. Export Performance, Technological Upgrading
over the nineties: the role of innovation. Research Policy 30, and Foreign Direct Investment Strategies in the Asian Newly
1185–1201. Industrializing Economies: With Special Reference to Singapur.
Bell, M., Albu, M., 1999. Knowledge systems and technological Investment and Corporate Studies, Santiago, Chile.
dynamism in industrial clusters in developing countries. World Lall, S., 1992. Technological capabilities and industrialization.
Development 27 (9), 1715–1734. World Development 20 (2), 165–186.
Bhat, S., Sethuraman, T.V., 1995. Technology transfer and export Lall, S., Kumar, R., 1981. Firm-level export performance in
performance: a case study of Indian automobile industry. The an inward-looking economy: the Indian engineering industry.
Indian Economic Journal 43, 1–19. World Development 9, 453–463.
Branson, W.H., 1971. U.S. comparative advantage: some further Lall, S., 1986. Technological development and export performance
results. Brookings Papers on Economic Activity 3, 754–759. in LDCs: leading engineering and chemical firms in India.
Chacoliades, M., 1990. International Economics. McGraw-Hill. Weltwirtschaftliches Archiv 122, 80–92.
Clerides, S., Lath, S., Tybout, J., 1996. Is ‘Learning-by-Exporting’ Lundvall, B., 1995. National Systems of Innovation: Towards a
Important? Micro-Dynamic Evidence from Colombia, Mexico Theory of Innovation and Interactive Learning. Pinter, London.
and Morocco. NBER Working Paper No. 5715, Cambridge, Lowinger, T.C., 1975. The technology factor and the export
MA. performance of U.S. manufacturing industries. Economic
Dahlman, C.J., Larson, B.R., Westphal, L.E., 1987. Managing Inquiry 13, 221–236.
technological development: lessons from the newly Nassimbeni, G., 2001. Technology, innovation capacity, and the
industrializing countries. World Development 15 (6), 759–775. export attitude of small manufacturing firms: a Logit/Tobit
Dahlman, C.J., Westphal, L.E., 1982. Technological effort in model. Research Policy 30, 245–262.
industrial development. In: Stewart, F., James J. (Eds.), The Nelson, R.R., 1981. Research on productivity growth and
Economics of New Technology in Developing Countries. productivity differences: dead ends or new departures. Journal
Frances Pinter, London. of Economic Literature 19 (3), 1029–1064.
Dosi, G., 1988. Sources, procedures and microeconomic effects of
Nelson, R.R., 1987. Innovation and economic developments:
innovation. Journal of Economic Literature 26 (3), 1120–1171.
theoretical retrospect and prospect. In Katz, J.M. (Ed.), 1987.
Freel, M.S., 2003. Sectoral patterns of small firm innovation,
Technology Generation in Latin American Manufacturing
networking and proximity. Research Policy 32, 751–770.
Industries. Macmillan, London.
Galende, J., de la Fuente, J.M., 2003. Sectoral patterns of small
Nelson, R.R., Winter, S.J., 1982. An Evolutionary Theory of
firm innovation, networking and proximity. Research Policy 32,
Economic Change. Harvard University Press, Cambridge, MA.
715–736.
Glejser, H., Jacquemin, A., Petit, J., 1980. Exports in an imperfect OECD, 1996. Oslo Manual: Proposed Guidelines for Collecting
competition framework: an analysis of 1446 exporters. and Interpreting Technological Innovation Data. OECD, Paris.
Quarterly Journal of Economics 94, 507–524. Oerlemans, L., Meeus, M., Boekema, F., 1998. Do networks
Greene, W.H., 2003. Econometric Analysis. fifth ed., Prentice Hall, matter for innovation? The usefulness of the economic network
New Jersey. approach in analysing innovation. Tijdschrift voor Economische
Gruber, W.H., Mehta, D., Vernon, R., 1967. The R&D factor in en Sociale Geografie 89, 298–309.
international trade and international investment of United States Pamukçu, T., 2003. Trade liberalization and innovation decisions
industries. Journal of Political Economy 75, 20–37. of firms: lessons from post-1980 Turkey. World Development
Haque, I.u. 1995. Technology and competitiveness. In Haque I.u. 31 (8), 1443–1458.
(Ed.), Trade, Technology and Competitiveness. The World Bank Posner, M.V., 1961. International trade and technical change.
Economic Development Institute, Washington, DC. Oxford Economic Papers 13, 323–341.
Hirsch, S., Bijaoui, I., 1985. R&D intensity and export Rodrik, D., 1995. Trade Strategy, Investment and Exports: Another
performance: a micro view. Weltwirtschaftliches Archiv 121, Look at East Asia. NBER Working Paper No. 5339, Cambridge,
238–251. MA.
Hirsch, S., Kalish, S., Katznelson, S., 1988. Effects of knowledge Roper, S., Love, J.H., 2002. Innovation and export performance:
and service intensities on domestic and export performance. evidence from the UK and German manufacturing plants.
Weltwirtschaftliches Archiv 124, 230–241. Research Policy 31, 1087–1102.
Hughes, K., 1986. Exports and Technology. Cambridge University Schumpeter, J.A., 1942. Capitalism, Socialism and Democracy.
Press, Cambridge. Harper Torchbooks, New York.
424 E. Özçelik, E. Taymaz / Research Policy 33 (2004) 409–424

Sterlacchini, A., 1999. Do innovative activities matter to small Trade Policies in Developing Countries During the 1990s. May,
firms in non-R&D-intensive industries? An application to export Antalya.
performance. Research Policy 28, 819–832. Vernon, R., 1966. International investment and international trade
Sterlacchini, A., 2001. The determinants of export performance: a in the product cycle. Quarterly Journal of Economics 80, 190–
firm-level study of Italian manufacturing. Weltwirtschaftliches 207.
Archiv 137, 450–472. Vestal, J.E., 1989. Evidence on the determinants and factor content
Stern, R.M., Maskus, E.K., 1981. Determinants of the structure of characteristics of Japanese technology trade, 1977–1982.
U.S. foreign trade, 1958–76. Journal of International Economics Review of Economics and Statistics 71, 565–571.
11, 207–224. Walker, W., 1979. Industrial Innovation and International Trading
Sveikauskas, L., 1983. Science and technology in united states Performance. JAI Press, Greenwich, CT.
foreign trade. The Economic Journal 93, 542–554. Wakelin, K., 1997. Trade and Innovation: Theory and Evidence.
Taymaz, E., 1999. Trade liberalization and employment generation: Edward Elgar, Aldershot.
the experience of Turkey in the 1980s. In: Revenga A. Wakelin, K., 1998. Innovation and export behaviour at the firm
(Ed.), Turkey: Economic Reforms, Living Standards, and level. Research Policy 26, 829–841.
Social Welfare Study. vol. II, Technical Papers, World Bank, Wignaraja, G., 2002. Firm size, technological capabilities and
Washington, DC. market-oriented policies in Mauritius. Oxford Development
Taymaz, E., 2001. National System of Innovation: The Processes of Studies 30, 87–104.
Innovation and Technological Change in Turkish Manufacturing Zhao, H., Li, H., 1997. R&D and export: an empirical analysis
Industries (in Turkish). TUBITAK/TTGV/DIE, Ankara. of Chinese manufacturing firms. Journal of High Technology
Uygur, E., 1991. Policy, trade and growth in Turkey: 1970–1990. Management Research 8, 89–106.
Paper presented at the UNCTAD Expert Group Meeting on