An assessment of the technology level and knowledge intensity of regions in Turkey

European Planning Studies ISSN: 0965-4313 (Print) 1469-5944 (Online) Journal homepage: https://www.tandfonline.com/loi/ceps20 An assessment of the technology level and knowledge intensity of regions in Turkey Necmettin Çelik, Sedef Akgüngör & Neşe Kumral To cite this article: Necmettin Çelik, Sedef Akgüngör & Neşe Kumral (2019): An assessment of the technology level and knowledge intensity of regions in Turkey, European Planning Studies, DOI: 10.1080/09654313.2019.1579301 To link to this article: https://doi.org/10.1080/09654313.2019.1579301 Published online: 11 Feb 2019. Submit your article to this journal Article views: 29 View Crossmark data Full Terms & Conditions of access and use can be found at https://www.tandfonline.com/action/journalInformation?journalCode=ceps20 EUROPEAN PLANNING STUDIES https://doi.org/10.1080/09654313.2019.1579301 An assessment of the technology level and knowledge intensity of regions in Turkey Necmettin Çelika, Sedef Akgüngörb and Neşe Kumralc a Department of Economics, Faculty of Economics and Administrative Sciences, İzmir Katip Çelebi University, İzmir, Turkey; bDepartment of Economics, Faculty of Business, Dokuz Eylül University, İzmir, Turkey; c Department of Economics, Faculty of Economics and Administrative Sciences, Ege University, İzmir, Turkey ABSTRACT ARTICLE HISTORY The paper investigates the patterns of technology and knowledge of the regions. The first aim of the paper is to determine cluster templates at the national level. The second aim of the paper is to investigate the technology and knowledge composition of the regional highpoint clusters. The paper identifies patterns of industrial linkages to define cluster templates and regional highpoints. The second part uncovers regional distributions of technology and knowledge. The data comes from Turkey’s 2012 input–output table. The location quotients use industrial employment statistics from the Turkish Statistical Institute. The technological and knowledge intensity classification follows Eurostat. The findings reveal 10 cluster templates in Turkey. Spatial distribution of the highpoint clusters reveals that most regions contain highpoint clusters with low technology and low knowledge-intensive sectors. The results reveal that highpoint clusters in Turkey’s regions contain industries whose technologies do not demand high skills, knowledge and sophistication. Limited existence of high-tech industries and low knowledge intensity in Turkey’s industry composition is a limiting factor for transition to high value-added manufacturing. Special emphasis should be directed towards constructing regional advantage, given the current levels of technology and knowledge intensity. Received 31 July 2018 Revised 10 December 2018 Accepted 28 December 2018 KEYWORDS Industry clusters; inputoutput analysis; principal components analysis; regional specialization; technology and knowledge Intensive Introduction Policies that support clustering initiatives started in the late 1990s mostly owing to Turkey’s preparation of accession to the European Union (EU). The beginnings of discussions and early work in clustering in Turkey are associated with the Competitive Advantage of Turkey (CAT) project that was led by non-governmental organizations and private sector leaders in association with Michael Porter between 1999 and 2001. The activities of CAT were later carried out by International Competitiveness Research Institute (URAK), a local NGO that was founded by private sector leaders and bureaucrats (Bulu & Yalçıntaş, 2014). The CAT project and URAK were pioneer civil arenas for increasing awareness and research on clusters (Dulupçu, Karagöz, Sungur, & Ünlü, 2015). Despite these early efforts, Turkey lacked the presence of a systematic and CONTACT Sedef Akgüngör sedef.akgungor@deu.edu.tr © 2019 Informa UK Limited, trading as Taylor & Francis Group 2 N. ÇELIK ET AL. complete scheme for a national cluster policy until the beginning of the twenty-first century. The first policy that placed a strong emphasis on clusters was the SME Strategy and Action Plan in 2004 prepared by the Ministry of Industry for adoption into EU procedures (Alsaç, 2010). This action plan pointed out the importance of clusters. The revised version of the document accepts clusters as a tool to increase competitiveness (Dulupçu et al., 2015). The role of clusters continues to be mentioned in Turkey’s macro-level medium-term programs as well. The UNDP has been a major international organization that supports clustering initiatives in Turkey. The UNDP projects mostly target clusters for rural areas, such as the marble cluster in Diyarbakır, ready-wear cluster in Adıyaman and organic agriculture in Şanlı Urfa (Bulu & Yalçıntaş, 2014). Another important initiative is the Istanbul fashion and textile cluster initiated by the Istanbul Textile and Apparel Exporter Association. There have been 10 EU-funded cluster projects between 2007 and 2009. Currently, cluster initiatives in Turkey are supported through complementary tools and mechanisms, such as support programs initiated by the Ministry of Science, Industry and Technology, the Ministry of Economics and regional development agencies. Early studies on cluster formation in Turkey reveal that there are emerging regions that are characterized by internationally competitive local production systems, such as Çorum, Bursa, Denizli and Gaziantep (Eraydın, 2002 and Eraydın and Koroglu, 2005). Öz (2002) identifies and elaborates on the performance of the towel/bathrobe cluster in Denizli and furniture cluster in Ankara following Porter’s diamond approach. Öz further emphasizes the importance of cluster initiatives in enhancing competitiveness (Öz, 2004). Since the beginning of the twenty-first century, studies on clusters in Turkey have focused on sector dynamics, sub-sector structure and performance, technological intensity and integration with global value chain. The studies include macro-level reports (Sanayi ve Ticaret Bakanlığı, 2007; TUSIAD, 2005) and studies conducted for regional development agencies such as İZKA (2010), GEKA (2011), BEBKA (2012). Additionally, there are studies that focus on specific regions (see, for example, Albayrak & Erkut, 2010; Çağlar & Kutsal, 2011; Deliktaş & Çelik, 2018;; Karakayacı & Dinçer, 2012 Sungur, 2015) as well studies with the objective of exploring agglomerations within the country as a whole (see, for example, Akgüngör, 2006; Akgüngör, Kumral, & Lenger, 2003; Çiftçi, 2018; Elburz & Gezici, 2012; Falcıoğlu & Akgungor, 2008; Filiztekin, Barlo, & Kıbrıs, 2011; Sungur, 2015; TÜSİAD, 2005). Dincer, Özaslan, and Kavasoğlu (2003), Doğruel (2006), Doğruel and Doğruel (2011), Kaygalak (2011) further demonstrate various industrial agglomerations across the country in the context of regional development policies. The studies above in general focus on pointing out regional agglomerations and specializations and explore the impact of potential growth in regional or local economies. A limited number of studies emphasize mixtures of economic activities by looking at related variety and unrelated variety within the regions (Akgüngör, Kuştepeli, & Gülcan, 2013; Falcıoğlu, 2011) and analyse technology composition and composition of knowledge intensity (Makelainen, 2014; Türkcan, 2015). Considering the importance of clusters in developing and sustaining competitiveness, it is imperative to explore the extent to which the clusters are able to create high value-added products. In Turkey, recent literature on clusters emphasize on the issue of technological EUROPEAN PLANNING STUDIES 3 intensity and sophistication (Falcıoğlu & Akgungor, 2008; Gezici, Yazgı-Walsh, & Kacar, 2017; Kaygalak, 2013; Kaygalak & Reid, 2016; Kazancık, 2007; Kıymalıoğlu & Ayoğlu, 2006; Yolchi & Akseki, 2018). Karadağ, Deliktaş, and Önder (2004) investigate public capital formation on private manufacturing sector performance at national and regional level in Turkey, and demonstrate that public capital affects private output positively in many regions. In general, existing studies confirm that Turkey’s industrial activities contain medium and low technologies. Sectors that require more advanced technologies are located in regions with high socioeconomic development levels (Kazancık, 2007; TÜSİAD, 2005). With regard to technological composition of clusters, Kaygalak (2013) explored the clustering tendency of sectors and identified 11 sectors with tendencies to form clusters. His findings propose that none of the identified 11 sectors are high-tech sectors. Kaygalak and Reid (2016) further demonstrate that increases in sectoral agglomerations tend to be within the medium-low and medium-high technology sectors. The results proposed by Gezici et al. (2017) support a similar pattern. Eyyuboğlu and Aktaş (2017) further argue that techno parks as a tool for attracting high technology sectors are usually located in the western regions of Turkey where socioeconomic development is high. Although the studies above focus on the regional dispersions of technology level, little is known with regard to the technology levels of cluster templates. There is also limited knowledge of the technology and knowledge profiles of the highpoint clusters across the regions. Moreover, the previous studies emphasized technology levels, whereas information on levels of knowledge intensity across the sectors is scarce. Considering the importance of knowledge-based economy on innovation growth, it is imperative to explore the role of knowledge use in economic activities, as knowledge-based capital in many OECD countries is as important as physical capital (OECD, 2013). The first aim of this study is to explore the pattern of forward and backward linkages across Turkey’s industries using the latest available input–output table (2012). The aim is to understand the current pattern of inter-industry transactions to identify industry cluster templates. Using the identified cluster templates, the objective is to explore spatial distribution of economic activity and determine highpoint clusters within the regions. The second aim of this study is to explore the profile of the cluster templates according to technology levels and knowledge intensity. The paper further explores the geographical dispersion of technology and knowledge levels by looking at the sectoral composition of the highpoint clusters across Turkey’s regions. Theoretical background The theoretical foundation of this study is based on cluster theory and constructing regional advantage (CRA) approach, particularly on the relationship between them. The cluster theory explains the dispersion and agglomeration of economic activity at the regional level. For over two decades, clusters and cluster-based economic development have been accepted paradigms in understanding and explaining regional development. The theory and empirical work on agglomeration economies have been well established since the seminal work of Marshall (1920). Since the 1990s, Porter’s diamond model has been a prominent tool for cluster analysis (Porter, 1999). The core-periphery model 4 N. ÇELIK ET AL. of Krugman (1991) and regional innovation systems (Asheim, Boschma, & Cooke, 2011; Cooke, 1992; Cooke, 2012) provide additional perspectives to cluster theory. The CRA approach developed by Asheim et al. (2011), argues that creating regional advantage requires proactive public and private partnerships based on related variety and distributed knowledge bases and networks. The foundation of CRA relies on promoting competitive advantage through differentiation of regions and creating unique products for regional competitiveness. The CRA approach contends that the advantage of each region lies in its unique characteristics such as prevailing knowledge bases and related variety and innovation systems. As such, constructing regional advantage through innovations requires proactive public–private partnerships and public policy. Consequently, the CRA approach offers a theoretical basis to study the role that regional policies can play for regional development. Cluster theory and the CRA approach are interrelated because identification and analysis of the composition of clusters is the first step in understanding the regional knowledge base and industrial structure of the regions. In order to construct regional advantages, it is crucial to define regional industry clusters and their technological intensities. The CRA approach contends that it is important for regional policy makers to implement strategies that do not lead to the reinforcement of existing regional structures (Morgan, 2013). Rather than creating negative lock-it, the focus is on creating new growth paths based on the current industry composition and distributed knowledge bases. The emphasis of the CRA approach is based on regional platform policies which in turn are based on related variety and differentiated knowledge bases where knowledge sharing across networks of firms is significant in creating innovations. Regional innovation policies are thus based on CRA ideas that defy one-size-fits-all policies as well as stand-alone policies (Boschma, 2013). Rather that starting from starch, the CRA approach understands that current firm networks, knowledge bases and related economic activities are important in creating innovative development paths for the regions. Although the high-tech and low-tech dichotomy are not relevant in CRA’s approach toward regional development, it is imperative to know the current structure of the industries within the regions. Documenting the current cluster templates and understanding the technology levels and knowledge intensity compositions in the regional highpoint clusters is a preliminary step for further exploration regarding competitiveness. Moreover, the CRA approach is theoretically consistent with the current debate on smart specialization initiatives (S3) in the EC (Landabaso, 2012; OECD, 2013). Smart specialization is defined as letting entrepreneurs select and prioritize fields where a cluster of activities should be developed, and let entrepreneurs discover the right domains of future specialization (Balland, Boschma, Crespo, & Rigby, 2018; Boschma, 2013). In order to initiate smart specialization policies, it is informative to understand the technical and knowledge intensity composition of the sectors within the region. The CRA approach therefore emphasizes the importance of understanding the current structure of clusters. Starting with baseline characteristics, the aim is to create new paths. The new paths can be developed through understanding and using the existing industry compositions, backward and forward linkages and current knowledge bases. Therefore, it is crucial to have a knowledge of the current composition of the industries across the regions as well as technological levels and knowledge use. EUROPEAN PLANNING STUDIES 5 The paper aims to help construction of regional advantages by exploring the current composition and regional distribution of cluster templates in Turkey. The authors accept that cluster theory and the CRA approach are interrelated. Clusters are not only tools for path extension but they are also important to achieve path renewal and new path creation (Asheim, Isaksen, Martin, & Trippl, 2016). In order to use, renew and create new paths for regional development, the initial step for policy makers is to explore the current structure of industries in the regions (cluster theory) and then look for ways to construct competitive advantages based on related variety and knowledge bases. As outlined above, the theoretical justification of this study is related to constructing regional advantage and smart specialization approaches. Rather than focusing on dimensions of smart specialization policies or constructing regional advantages in Turkey, the primary focus of the study is to understand the regional knowledge bases of technology structure as a first step in implementing strategies for further development by using the new combinations of growth paths within the existing structure. Data and methods The data consists of the latest (2012) input–output (I–O) table of Turkey. The 2012 national I–O table includes the sale and purchase transactions data across 62 industries. In this paper, the authors use the supply and demand table (or more correctly input and output tables) in deriving industry cluster templates. For identification of industry templates, the reason that the I–O table at the national level is used is owing to the non-existence of suitable regional level I–O tables in Turkey. The authors therefore generalize the national level inter-industry dependency to the regional level by assuming that the structure of sectoral interconnections at the national level and regional level are similar. Whereas the I–O tables structure is expected to be different at regional levels, data limitations compelled the authors to assume that the national level structure is valid at the regional level. The authors use the cluster templates identified at the national level in order to explore the regional dispersion of economic activity. In exploring the regional dispersion (location quotients), we use employment data that is reported at the NACE Rev.2 (number of persons employed) compiled from the Turkish Statistical Institute (TURKSTAT) for the 2009–2015 period.1 ‘The first step of the analysis’ is to determine the macro-level industry templates. The method follows the principle component approach initially suggested by the work by Czamanski and Ablas (1979) at the national level. Later, O’hUallachain (1984) used the method to analyse industrial activities at the metropolitan level in the US. The principle component analysis is considered to be an appropriate method to explore connections across industries (Feser & Bergman, 2000; Hofe & Bhatta, 2007). The method is outlined in the Appendix. ‘The second step of the analysis’ is to determine the level of technology and knowledge use in the cluster templates according to the composition of primary and secondary industries within each cluster template. The technology level and knowledge use for the industries are based on Eurostat’s sectoral approach to technical classification. Eurostat’s sectoral approach combines manufacturing industries according to their technological 6 N. ÇELIK ET AL. intensity as measured by R&D expenditure/value added at the NACE 2-digit level. The level of R&D intensity is used as a criterion to group economic sectors into four groups: High-Technology (HT), Medium High-Technology (MHT), Medium Low-Technology (MLT) and Low-Technology (LT) industries. Services are grouped according to the share of tertiary-educated persons at the NACE 2-digit level: Knowledge-Intensive Services (KIS) and Less Knowledge-Intensive Services (LKIS) based on the share of tertiary-educated persons at the NACE 2-digit level.2 ‘The third step of the analysis’ involves determining the spatial distribution of the cluster templates according to the location quotient (LQ) values greater than 1.25 in 2009 and 2015. In determining the highpoint clusters, initially we select the regions where the cluster template as a whole had LQ values of 1,25 and above. Economic theory suggests that a LQ value greater than 1.00 has proportionally more workers than the larger comparison area employed in a specific industry sector.3 As a rule of thumb, LQ values greater than 1.25 is generally accepted as an evidence of regional specialization (Kopczewska, Churski, Ochojski, & Polko, 2017). A high location quotient in a specific industry may translate into a competitive advantage in that industry for the local economy (Baer & Brown, 2006). As discussed above, the cluster templates consist of a combination of industries. Among the industries of the cluster template in a particular region, there may be one or more secondary industries (with low loading scores, suggesting a weak relationship with the underlying cluster) whose LQ value is below 1.25; while none of the primary industries (with high loading scores suggesting a strong relationship with the underlying cluster) has a LQ value above 1.25. If this is the case, it would be difficult to conclude that such a cluster is a highpoint economic activity for that region. The authors therefore eliminated the regions if none of the primary industries of the cluster template have LQ value of 1.25 or above at a specific region. ‘The fourth step of the analysis’ is to investigate the spatial distribution of technology and knowledge use across regional highpoint clusters. In each region, we report the frequency distribution of the primary and secondary industries according to their technology levels and knowledge use. The frequencies across the regions are then used to map out the technology levels and knowledge intensity of the regions according to their number of industries with regards to technology level and knowledge use (HT, MHT, MLT, LT, KIS, LKIS, HT-KIS). The inquiry with regard to the third step above involves spatial distribution of cluster templates across regions. The aim is to determine highpoint clusters. The inquiry with regard to the fourth step is to look deeper by investigating the technology levels and knowledge intensity of the industries (sectoral composition of clusters) within each of the highpoint clusters in the region. Findings The findings reveal 10 groups of industries (cluster templates) each of which involve backward and forward interactions across various sectors. The backward and forward linkages across industries within each group allow us to point out specific groups of activities and therefore makes it possible to give a broad name for the cluster templates described below. EUROPEAN PLANNING STUDIES 7 1. Macro level cluster templates The first finding is at the macro level. Using the 2012 input–output (I–O) table of Turkey and following the methodology outlined in the methods section above, the principal component analysis reveals 10 factors that constitute the foundation of the cluster templates. The analysis explains nearly 85% of the variation in the matrix. Table 1 shows the results of the principle component analysis. Each of the10 factors represents groups of sectors. Table 2 reveals the names of the sectors that constitute the 10 factors identified above. By observing the content of the factors, it is possible to define and give names to the groups of industries within each of the factor group. Following Table 2, the identified cluster templates include: construction, services, logistics, food and agriculture, automotive, media and communication, health, textile, tourism and energy. Primary sectors are the subsectors where factor loading values are 0.60 and above, secondary sectors are the subsectors where factor loadings range from 0.35 to 0.59. The largest number of sectors belong to services cluster templates followed by construction and logistics with a total number of sectors (primary and secondary sectors) of 18, 19 and 12 respectively, suggesting existence of a variety of economic activities. The cluster templates with the smallest number of sectors are textile, tourism and energy, where the total number of sectors are 5, 5 and 4, respectively. Table 2 further reveals the technology and knowledge intensity classification of each of the sectors within each of the identified cluster templates. Table 3 demonstrates the employment statistics of the identified cluster templates for the 2009–2015 period. With regards to employment, the cluster templates with the largest number of workers (primary and secondary sectors combined) are construction, followed by food and agriculture, services and textile clusters. 2. Composition Of macro-level cluster templates according to their technology levels and use of knowledge Using the Eurostat classifications4 on technology level and use of knowledge, we explore the extent to which identified cluster templates involve sectors that are High-Tech (HT), Medium High-Tech (MHT), Medium Low-Tech (MLT) and Low-Tech (LT) and at what level they depend on knowledge-based inputs such as Knowledge-Intensive Sectors (KIS), Less Knowledge-Intensive sectors (LKIS) and High-Tech and Knowledge-Intensive Sectors (HT-KIS). Table 1. Results of the Principal Component Analysis [Eigen value cutoff: 1]. Factor 1 2 3 4 5 6 7 8 9 10 Eigen value % Variance % Cumulative variance 13,869 11,531 6,813 6,110 4,231 3,152 2,437 1,735 1,708 1,579 22,369 18,598 10,989 9,855 6,824 5,083 3,931 2,798 2,755 2,546 22,369 40,968 51,957 61,812 68,636 73,719 77,651 80,449 83,204 85,750 8 N. ÇELIK ET AL. Table 4 shows that 95 sectors are included in the identified clusters. Out of the 95 sectors that constitute the clusters, it is possible to make technological and knowledge classifications 84 of them, while 11 are not included in the Eurostat technological and knowledge classification. There are 2 HT, 7 MHT, 10 MLT, 9 LT and 7 HT-KIS sectors.5 Simple observation of the frequency of industries according to technology level reveals the dominance of the medium and low technology sectors. The findings are consistent with discussions presented in previous research (Avcı, Uysal, & Taşcı, 2016; Eşiyok, 2013; Tübitak, 2016). According to the knowledge intensity composition of the subsectors, 26 are classified as knowledge-intensive sector, 23 are classified as low knowledgeintensive. The only 2 HT sectors in Turkey’s industry cluster composition belong to the automotive and health clusters. With regards to MHT sectors, automotive has 4, construction has 2 and textile has 1 sectors. Textile as a traditional industry in Turkey is still dominated by low-level technologies. Similarly, the construction industry, a large economic base across Turkey for the last several years, uses medium-level technologies. Health and automotive are the only clusters that have high-technology industries. The finding is consistent with Tübitak (2016) where it is stated that automotive and health are among the industries that are most ready for digital transformation. Of the 26 KIS sectors, 9 are in the services cluster, while construction and logistics have 4 KIS sectors each. Services, tourism and media contain sectors that have both HT and KIS. The findings show that services, construction and logistics are associated with highest number of industries that require knowledge and human capital. Table 2. Sectors within the factors identified and factor names (cluster templates). Factor 1: Construction NACE Code Primary sectors 71 23 41-43 16 25 22 77 24 27 37-39 28 84 93 Secondary sectors 47 5-9 68 46 33 Industry Loadings Technology/Knowledge classification Architectural and engineering services; technical testing and analysis services Other non-metallic mineral products Constructions and construction works Wood and of products of wood and cork, except furniture; articles of straw and plaiting materials Fabricated metal products, except machinery and equipment Rubber and plastic products Rental and leasing services Basic metals Electrical equipment Sewerage services; materials recovery services; remediation services Machinery and equipment n.e.c. Public administration and defense services; compulsory social security services Sporting services and amusement and recreation services Retail trade services, except of motor vehicles and motorcycles Mining and quarrying Real estate services excluding imputed rents Wholesale trade services, except of motor vehicles and motorcycles Repair and installation services of machinery and equipment 0.964 KIS 0.951 0.951 0.857 MLT – LT 0.849 MLT 0.829 0.825 0.807 0.766 0.733 MLT LKIS MLT MHT – 0.697 0.653 MHT KIS 0.634 KIS 0.572 KIS 0.536 0.475 0.459 LKIS – LKIS 0.376 MLT EUROPEAN PLANNING STUDIES 9 Factor 2: Services NACE Code Primary sectors 69 66 64 72 94 95 78 53 62 Secondary sectors 61 68 80 65 96 85 84 47 18 45 Industry Loadings Technology/ Knowledge classification Legal and accounting services;,services of head offices; management consulting services Services auxiliary to financial services and insurance services Financial services, except insurance and pension funding Scientific research and development services Services furnished by membership organizations Repair services of computers and personal and household goods Employment services Postal and courier services Computer programming, consultancy and related services; Information services Telecommunications services Real estate services excluding imputed rents Security and investigation services; services to buildings and landscape; office administrative, office support and other business support services Insurance, reinsurance and pension funding services, except compulsory social security Other personal services Education services Public administration and defense services; compulsory social security services Retail trade services, except of motor vehicles and motorcycles Printing and recording services Wholesale and retail trade and repair services of motor vehicles and motorcycles 0.912 KIS 0.887 KIS 0.878 0.755 0.728 0.709 KIS KIS HT-KIS LKIS 0.663 0.653 0.648 LKIS KIS HT-KIS 0.643 0.598 0.577 HT-KIS LKIS KIS 0.563 KIS 0.501 0.469 0.466 LKIS KIS KIS 0.465 LKIS 0.441 0.352 LT LKIS Factor 3: Logistics NACE Code Primary sectors Secondary sectors 49 52 19 45 50 2 5-9 85 53 46 51 65 Industry Loadings Technology/ Knowledge classification Land transport services and transport services via pipelines Warehousing and support services for transportation Coke and refined petroleum products Wholesale and retail trade and repair services of motor vehicles and motorcycles Water transport services Products of forestry, logging and related services Mining and quarrying Education services Postal and courier services Wholesale trade services, except of motor vehicles and motorcycles Air transport services Insurance, reinsurance and pension funding services, except compulsory social security 0.940 0.917 0.912 0.806 LKIS LKIS MLT LKIS 0.758 0.687 0.628 0.534 0.525 0.522 0.399 0.355 KIS – – KIS LKIS LKIS KIS KIS 10 N. ÇELIK ET AL. Factor 4: Food and Agriculture NACE Code Primary sectors 1 10-12 55-56 87 17 3 Secondary sectors 46 47 74-75 79 Industry Loadings Technology/Knowledge classification Products of agriculture, hunting and related services Food, beverages and tobacco products Accommodation and food services Residential care services; social work services without accommodation Paper and paper products Fish and other fishing products; aquaculture products; support services to fishing Wholesale trade services, except of motor vehicles and motorcycles Retail trade services, except of motor vehicles and motorcycles Other professional, scientific and technical services and veterinary services Travel agency, tour operator and other reservation services and related services 0.919 0.914 0.802 0.792 – LT LKIS KIS 0.741 0.718 LT – 0.468 LKIS 0.465 LKIS 0.420 KIS 0.380 LKIS Factor 5: Automotive NACE Code Primary sectors Secondary sectors 30 29 33 28 26 27 37-39 31 24 25 Industry Loadings Technology/Knowledge classification Other transport equipment Motor vehicles, trailers and semi-trailers Repair and installation services of machinery and equipment Machinery and equipment n.e.c. Computer, electronic and optical products Electrical equipment Sewerage services; materials recovery services; remediation services Furniture and other manufactured goods Basic metals Fabricated metal products, except machinery and equipment 0.894 0.840 0.823 MHT MHT MLT 0.645 0.629 0.587 0.523 MHT HT MHT – 0.521 0.451 0.440 LT MLT MLT Factor 6: Media and Communication NACE Code Primary sectors 73 60 59 Secondary sectors 18 90-92 95 Industry Loadings Technology/ Knowledge classification Advertising and market research services Publishing services Motion picture, video and television programme production services, sound recording and music publishing; programming and broadcasting services Printing and recording services Creative, arts, entertainment, library, archive, museum, other cultural services; gambling and betting services Repair services of computers and personal and household goods 0.883 0.859 0.808 KIS HT-KIS HT-KIS 0.577 0.386 LT KIS 0.357 LKIS EUROPEAN PLANNING STUDIES 11 Factor 7: Health NACE Code Primary sectors Secondary sectors 21 86 80-82 78 31 62-63 Industry Loadings Technology/ Knowledge classification Basic pharmaceutical products and pharmaceutical preparations Human health services Security and investigation services; services to buildings and landscape; office administrative, office support and other business support services Employment services Furniture and other manufactured goods Computer programming, consultancy and related services; Information services 0.875 HT 0.865 0.586 KIS KIS 0.579 0.564 0.432 KIS LT HT-KIS Factor 8: Textile NACE Code Primary sectors 13-15 20 96 17 22 Secondary sectors Industry Loadings Technology/Knowledge classification 0.864 LT 0.819 0.630 0.429 0.351 MHT LKIS LT MLT Textiles, wearing apparel, leather and related products Chemicals and chemical products Other personal services Paper and paper products Rubber and plastic products Factor 9: Tourism NACE Code Primary sectors 51 79 Secondary sectors 3 55-56 94 Industry Loadings Technology/Knowledge classification Air transport services Travel agency, tour operator and other reservation services and related services Fish and other fishing products; aquaculture products; support services to fishing Accommodation and food services Services furnished by membership organizations 0.812 0.789 KIS LKIS 0.516 – 0.437 0.362 LKIS LKIS Factor 10: Energy NACE Code Primary sectors 35 36 Secondary sectors 85 72 Industry Loadings Technology/Knowledge classification Electricity, gas, steam and air conditioning Natural water; water treatment and supply services Education services Scientific research and development services 0.903 0.716 – – 0.436 0.414 KIS HT-KIS 12 N. ÇELIK ET AL. Table 3. Employment statistics of the macro level cluster templates (number of workers). Primary Sectors Cluster template Construction Services Logistics Food and Agriculture Automotive Media and Communication Health Textile Tourism Energy Number of Sectors 2009 2010 2011 2012 2013 2014 2015 Average 2009–2015 % Share 2009–2015 13 10 7 6 5 3 2 3 2 2 1,760,461 440,603 1,236,154 1,030,948 352,025 49,995 188,357 895,119 49,451 94,588 2,096,436 444,395 1,260,276 1,158,580 384,396 66,689 212,111 972,350 53,230 94,693 2,455,720 491,064 1,354,137 1,310,633 423,746 76,483 239,713 1,104,863 59,795 98,458 2,766,448 544,351 1,429,407 1,447,583 456,799 81,484 256,225 1,199,641 63,170 103,353 2,917,213 533,624 1,483,908 1,512,959 481,773 93,948 271,959 1,233,447 71,846 112,662 3,190,470 562,190 1,474,156 1,575,249 510,124 100,483 289,599 1,257,360 71,606 122,215 3,353,132 583,671 1,458,002 1,638,996 536,989 90,903 292,134 1,244,591 80,934 130,170 2,648,554 514,271 1,385,149 1,382,135 449,407 79,998 250,014 1,129,624 64,290 108,020 33.06 6.42 17.29 17.25 5.61 1.00 3.12 14.10 0.80 1.35 2009 2010 2011 2012 2013 2014 2015 Average 2009–2015 % Share 2009–2015 18 19 12 10 10 6 7 5 5 4 4,602,268 2,905,457 2,325,705 3,588,965 944,402 185,786 883,569 1,080,254 671,883 264,377 4,873,890 2,878,042 2,335,939 3,593,523 1,048,901 191,309 1,071,531 1,176,690 771,157 290,902 5,598,943 3,245,311 2,533,403 4,060,363 1,171,552 220,745 1,227,344 1,335,024 898,511 303,790 6,144,795 3,460,233 2,713,777 4,356,953 1,276,736 244,818 1,353,693 1,441,002 993,928 339,073 6,278,893 3,515,822 2,791,538 4,392,866 1,322,233 256,031 1,483,480 1,483,470 1,043,745 380,955 6,576,746 3,524,829 2,828,302 4,441,945 1,370,764 257,592 1,605,200 1,524,445 1,088,111 408,705 5,837,469 3,297,942 2,624,830 4,139,273 1,223,831 228,177 1,333,801 1,366,434 946,663 344,956 27.35 15.45 12.30 19.39 5.73 1.07 6.25 6.40 4.44 1.62 Number of Sectors Cluster template Construction Services Logistics Food and Agriculture Automotive Media and Communication Health Textile Tourism Energy Primary and secondary sectors combined 6,786,745 3,555,900 2,845,147 4,540,297 1,432,228 240,960 1,711,793 1,524,151 1,159,305 426,887 EUROPEAN PLANNING STUDIES 13 Table 4. Technology level and knowledge composition of the cluster templates (number of sectors). LKIS KIS HT-KIS LT MLT MHT HT NA TOTAL Services 6 9 3 1 0 0 0 0 19 Construction 3 4 0 1 5 2 0 3 18 Logistics 5 4 0 0 1 0 0 2 12 Automotive 0 0 0 1 3 4 1 1 10 Food 4 2 0 2 0 0 0 2 10 Media 1 2 2 1 0 0 0 0 6 Health 0 3 1 1 0 0 1 0 6 Textile 1 0 0 2 1 1 0 0 5 Tourism 3 1 0 0 0 0 0 1 5 Energy 0 1 1 0 0 0 0 2 4 TOTAL 23 26 7 9 10 7 2 11 95 Note: The numbers denote the frequency distribution of the primary and secondary sectors within the cluster templates according to their technology level and knowledge intensity. The shaded cells show the highest frequency. The technology and knowledge intensity level of the sectors are from Eurostat indicators on ‘High-tech industry and Knowledge – intensive services’ according to NACE Rev.2 classification. NA means that the industry is not listed in the Eurostat technology and knowledge classification. Table 5. Regional highpoint clusters. Cluster template Logistics Food and agriculture Automotive Media and communication Textile Tourism Energy Health Region code Region name LQ2009 LQ2015 % Change Primary sectors with LQ value greater than 1.25 in 2015 (NACE Rev.2 codes) TRA2 TR81 TRB2 TRC3 TR90 TR32 TR22 TR61 TR41 TR42 TR52 TR10 Ağrı Zonguldak Van Mardin Trabzon Aydın Balıkesir Antalya Bursa Kocaeli Konya İstanbul 0.89 1.30 1.28 1.43 1.31 1.20 1.35 1.49 2.00 2.27 1.19 1.54 1.46 1.30 1.13 1.42 1.25 1.25 1.39 1.58 2.11 2.19 1.52 1.64 64.0 0.0 −11.71 −7.0 −4.6 5.0 3.0 6.0 5.5 −3.5 27.7 6.5 49 45, 5, 6, 7, 8, 9 49 49 10, 11, 12 1, 10, 11, 12, 55, 56, 3 10, 11, 12, 55, 56 55, 56 30, 29, 28 30, 29, 33, 28 29, 28 73, 59 TR21 TR63 TRC1 TR41 TR22 TR32 TR61 TR51 TRA1 TRA1 TRC3 TRC2 Tekirdağ Hatay Gaziantep Bursa Balıkesir Aydın Antalya Ankara Erzurum Erzurum Mardin Şanlıurfa 2.58 1.13 1.92 1.54 1.47 1.79 3.08 1.39 1.58 1.23 0.76 1.79 2.60 1.43 2.38 1.49 1.50 1.78 3.28 1.64 1.38 1.47 1.77 2.26 0.7 26.5 24.0 −3.2 2.0 −0.5 6.5 18.0 −12.6 13, 14, 15, 20 13, 14, 15, 96 13, 14, 15 13, 14, 15 NONE 79 51, 79 35, 36 35, 36 NONE NONE NONE 3. Regional Specialization using the macro-level cluster templates Table 5 shows the spatial distribution of the macro-level cluster templates across Turkey’s regions (regional highpoint clusters). In order to determine the specialization of regions, we look at the LQ values that are 1.25 and higher. However, high LQ values may be misleading owing to the fact that high LQ values for some of the regions may come from high employment in secondary industries, most of which may not be directly related to the cluster. This implies that large LQ values might come from industries that may be weakly linked to the cluster template (secondary industries with loading 0.35 and 0.59) while the primary industries (with loadings above 0.60) have lower LQ values. For 14 N. ÇELIK ET AL. Figure 1. Regional distribution of industry clusters (primary sectors with LQ value greater than 1.25 in 2015). example, although the TR22 Region (Balıkesir) has a LQ value of 1.25 in tourism, none of the primary sectors in the cluster template has an LQ value higher than 1.25.6 Similarly, the high LQ values of health cluster in the TRA1 Region (Erzurum), the TRC2 Region (Şanlıurfa) and the TRC3 Region (Mardin) come from secondary sectors and none of the primary sectors have LQ values greater than 1.25.7 The total number of specialized regions in Turkey in at least one cluster are therefore 17 out of a total of 26 regions when we exclude TRA1 Region (Erzurum), TRC2 Region (Şanlıurfa) and TRC3 Region (Mardin). Figure 1 shows spatial distribution of highpoint clusters for Turkey’s regions. - The regions along the eastern and southeastern border of Turkey appear as highpoint clusters for logistics. Ağrı and Van are two important regions on the transit highway to Iran. Ağrı is considered to be the door to Asia. Van is an important crossroads in the east Anatolian region. Similarly, Mardin stands along the intersection of numerous roads that reaches to Iraq. It is also reasonable to see that the LQ value of the logistic cluster template is higher than 1.25 in Zonguldak due to its being the centre of coal production in Turkey, a significant source of energy. - In terms of tea and hazelnut production we see the Trabzon region with food being a highpoint cluster template. With regard to Aydın and Balıkesir agri-food products constitute a significant source of livelihoods. Similarly, Antalya is a centre for production of vegetables and fruits. - The automotive cluster template reveals an LQ value 1.25 and above for Bursa, Kocaeli, Konya. The three regions are considered centres of the automotive industry. Bursa and Kocaeli specialize in automobile assembly. Konya is a region known as a supplier of automotive parts. - With regard to the media and communication cluster template, we see Istanbul as a centre with high specialization, an expected outcome. Istanbul is a centre for many EUROPEAN PLANNING STUDIES 15 services, and media (TV production, movie production, music production etc.) are no exception. Similarly, the majority of mainstream media (TV and radio) as well as headquarters of telecommunication companies are located in İstanbul. - The results demonstrate evidence that the textile cluster templates located in Hatay, Gaziantep, Bursa and Tekirdağ are major sources of industrial production. - Tourism as a highpoint cluster template for Aydın and Antalya is to be expected, as those areas abound with major holiday destinations. - With regard to energy, there are numerous companies that are specialized in distribution of energy (electricity, gas) as well as firms that operate in the field of water treatment and supply services in Ankara, the nation’s capital. Similarly, Erzurum is on a major natural gas pipeline (Bakü-Tiflis-Erzurum), which makes it possible for firms dealing with energy to locate in the region. 4. Technological and knowledge intensity of the regions The regional highpoint clusters identified above do not tell much about the underlying composition of the sectors. It is also worthy to explore how regions are specialized according to the underlying components (sectors) of the highpoint clusters. A deeper look at the composition of the clusters is useful for better understanding of a region’s industry base for further exploration of growth paths of the regions. In order to determine the profile of the clusters according to technological and knowledge intensity, we look at the LQ values of the primary industries of the highpoint clusters with a cutoff point of LQ equal or greater than 1.25. Among the highpoint clusters across the regions, none of the primary industries are high-tech, while the number of MHT industries is 9 (Figure 2). With regard to knowledge intensity, the number of KIS industries is 3 (Figure 3). The industries that constitute the Figure 2. Geographical distribution of industries according technology level. (The numbers denote frequency distribution of the primary sectors of the highpoint clusters in 2015). (LQ cutoff value is 1.25). Note that none of the HT industries have LQ value 1.25 and above. 16 N. ÇELIK ET AL. Figure 3. Geographical distribution industries according to knowledge intensity. (The numbers denote frequency distribution of the primary sectors of the highpoint clusters in 2015). (LQ cutoff value is 1.25). highpoint clusters across the regions are mostly LT and LKIS, reflecting low sophistication with regards to technology and knowledge intensity across regions. As seen in Figure 2, the MHT industries are located in 4 regions: Bursa, Kocaeli, Konya and Tekirdağ. Bursa, Kocaeli and Konya have automotive as highpoint clusters and Tekirdağ has textile as a highpoint cluster. The MHT in Bursa, Kocaeli and Konya are ‘Machinery and equipment n.e.c.’, ‘Motor vehicles, trailers and semi-trailers’ and ‘Other transport equipment sector’. The textile highpoint cluster in Tekirdağ includes ‘chemical and chemical products’ as an MHT sector. The MHT sectors are geographically located along the northwest region diagonal to the centre to Ankara. The KIS industries are located only in 2 regions: Antalya and İstanbul. Tourism is the highpoint cluster in Antalya where ‘air transport services’ is a KIS with a LQ value above 1.25. For Istanbul, media and communication represent the highpoint cluster where ‘advertising and market research services’ and ‘Motion picture, video and television programme production services, sound recording and music publishing; programming and broadcasting services’ are KIS with LQ values above 1.25. As seen above, the MHT and KIS are concentrated in the regions where the highpoint clusters are ‘automotive’, ‘textile’, ‘tourism’ and ‘media and communication’. Other regions usually include MLT, LT and LKIS industries. However, cluster-based regional economic development is not a smooth process. In order to create an innovative growth path for the regions with MHT industries (Bursa, Kocaeli and Konya for automotive; Tekirdağ for textile) there is a need to explore whether there are industries that are related in variety. Only then would it be possible to create innovative combinations of technology and knowledge bases within the regions. Furthermore, for the regions where the majority of LT industries are located, there is a need to create proactive polices that promote new and innovative growth paths, such as for logistics in Ağrı, Zonguldak, Van and Mardin, an food in Trabzon, Aydın, Balıkesir and Antalya. One possible policy recommendation is to initiate strategies for product differentiation particularly EUROPEAN PLANNING STUDIES 17 for the LT regions, and to promote new combinations of technologies and knowledge bases. For example, a policy recommendation could be to create new and innovative products/processes for industries with low technological sophistication so that the industry will be able to integrate into higher global value chains. Similarly, regions with LT industries need to be complemented with related activities/industries so that new combinations of technologies, knowledge, innovativeness can be created. Conclusions The paper presents the composition and regional distribution of cluster templates as a first step in designing policies toward constructing competitive advantage. There are two main conclusions of the paper as summarized below. The first conclusion is related to the overall composition of Turkey’s macro-level cluster templates according to level of technology levels and use of knowledge inputs. The findings reveal that HT and MHT sectors are in automotive, textile, health and construction clusters. The sole HT sectors appear in health and automotive both of which are considered to be most ready for digital transformation across all sectors in Turkey. All other clusters include MLT or LT sectors. With regards to knowledge use, KIS sectors appear in energy, health, services, media, logistics, food and agriculture and tourism cluster, and HT-KIS sectors appear in energy, media, services and health. Overall, the profile of cluster templates contains industries/sectors with low sophistication in terms of technology and knowledge inputs. The second conclusion is related to spatial distribution at the level of technology and knowledge use. 17 out of 26 NUTS II regions in Turkey have at least one highpoint cluster with overall LQ value larger than 1.25; and with LQ value of at least one primary sector larger than 1.25. Within the highpoint clusters of the regions, none include a HT and HT-KIS sector with LQ 1.25 and above. Most sectors across the highpoint clusters of the regions are MLT, LT and LKIS with LQ values 1.25 or above, indicating that regional specialization in Turkey is largely based on industries whose technologies do not demand high skills, knowledge and sophistication. The findings generally point out the lack of technological sophistication as a factor that limits the creation of regional competitive advantage. The results support the importance of recent efforts and government policies that place special emphasis on digital transformation of Turkey’s industries (T.C. Bilim, Sanayi ve Teknoloji Bakanlığı, 2018). The policy papers and existing work (Avcı et al., 2016; Eşiyok, 2013; Öngel, 2018; Tübitak, 2016) point out the importance of structural transformation across Turkey’s industrial base. The results of a recent country-wide survey of the managers of industrial firms reveal that Turkish industry has not even completed the third industrial revolution, in which the use of computer and automation rules industrial production (TÜBITAK, 2016). Implications can be drawn for Turkey’s regional development policies. The major policy recommendation involves the need to develop and apply policies that facilitate conditions for increased opportunities for technology and knowledge use in industrial production. Implications include two components: The first component of policy recommendations is related to technology and knowledge composition of industry cluster templates. The composition of the cluster templates limits N. ÇELIK ET AL. 18 the use of high technologies and high use of knowledge. There is a need for improvement in human capital through education that will enable the spread of new developments and technologies which require sophisticated skills and use of higher-level technologies. More university-industry collaborations and interactions would be a good way to motivate industry toward more state of the art technologies. Innovative activities from collaborations with R&D institutions would enhance positive externalities and promote sectoral and regional competitiveness. There is an urgent need for more enabling conditions that would create stronger links between firms and research institutes, such as Universities and techno-parks. The second implication is related to regional development policies. Selective promotion of regions is critical, and special emphasis should be given to regions where technology use and knowledge intensity use is low. Less developed regions, particularly the regions on the eastern side of the axis starting from northwest (İstanbul) going down to southern Anatolia should receive particular attention. Turkish industry should make efforts to adopt to global interdependencies so that it can integrate into global value chains There is also a need to promote of the regions that already have high-tech and knowledge-intensive sectors. Policies that would advance the cluster developments of these regions, however, should not create further regional imbalances. The emphasis should be to build on the strength of the regions in order to construct competitive advantage through existing knowledge bases and existing regional innovation systems. Notes 1. As explained above, the I-O table is at the national level. In exploring the agglomerations of the cluster templates at the regional level, we use regional employment data of the sectors in the cluster templates. The employment data at the regional level uses the NACE Rev.2 classification. The classification that is used in I–O table is CPA 2008. The authors use Eurostat correspondence tables (http://ec.europa.eu/eurostat/ramon/index.cfm?TargetUrl=DSP_ PUB_WELC) (Access Date 05.11.2018) in order to convert CPA2008 to an NACE Rev.2 classification. The NACE matchings of the sectors from the I–O table are presented in Table 2. Regional employment data is for 2009–2015 (Table 3) 2. http://ec.europa.eu/eurostat/cache/metadata/FR/htec_esms.htm, Access Date: 09.07.2018. 3. http://www.incontext.indiana.edu/2006/march/1.asp, Access Date: 25.10.2018 4. http://ec.europa.eu/eurostat/cache/metadata/Annexes/htec_esms_an3.pdf, Access Date: 09.07.2018 5. The NACE matchings of the sectors from the I-O table are presented in Table 2 6. The two primary sectors of the tourism cluster are, 51: Air transport services and 79: Travel agency, tour operator and other reservation services and related services. The LQ values of the primary sectors of Tourism in the TR22 Region (Balıkesir) are lower than 1,25. We therefore exclude TR22 Region (Balıkesir) since we cannot conclude that tourism is a highpoint economic activity. 7. The only sector of the health cluster for which TRA1 Region (Erzurum), TRC2 Region (Şanlıurfa) and TRC3 Region (Mardin) are specialized are, 80, 81, 82: Security and investigation services; services of buildings and landscape; office administrative Office support and other business support services; and 78: Employment services. 80, 81, 82 and 78 are secondary sectors of the health cluster template. Primary sectors of the health cluster are, 21: Basic pharmaceutical products and pharmaceutical preparations and 86: Basic pharmaceutical products and pharmaceutical preparations. As seen in Table 4, none of the regions has a LQ value of 1.25 or above in the primary sectors of the health cluster. Therefore, we conclude that the health cluster is not a highpoint economic activity in any of Turkey’s NUTS2 regions. EUROPEAN PLANNING STUDIES 19 Disclosure statement No potential conflict of interest was reported by the authors. References Akgüngör, S. (2006). Geographic concentrations in Turkey’s manufacturing industry: Identifying regional highpoint clusters. European Planning Studies, 14(2), 169–197. doi:10.1080/ 09654310500418002 Akgüngör, S., Kuştepeli, Y., & Gülcan, Y. (2013). An overview of industry clusters and the impact of related variety on regional performance in Turkey. European Review of Industrial Economics and Policy, 5, 1–10 Akgüngör, S., Kumral, N., & Lenger, A. (2003). National industry clusters and regional specializations in Turkey. European Planning Studies, 11(6), 647–669. doi:10.1080/ 0965431032000108378 Albayrak, A. N., & Erkut, G. (2010). Türkiye’de Bölgesel Rekabet Gücü Analizi. Yıldız Teknik Üniversitesi Mimarlık Fakültesi E-Dergisi, MEGARON, 5(3), 137–148. Alsaç, F. (2010). Bölgesel Gelişme Aracı Olarak Kümelenme Yaklaşımı ve Türkiye için Kümelenme Destek Modeli Önerisi. Ankara: DPT Uzmanlık Tezleri. Asheim, B. T., Boschma, R., & Cooke, P. (2011). Constructing regional advantage: Platform policies based on related variety and differentiated knowledge bases. Regional Studies, 45(7), 893–904. doi:10.1080/00343404.2010.543126 Asheim, B., Isaksen, A., Martin, R., & Trippl, M. (2016). The role of clusters and public policy in new regional economic development. In D. Fornahl, & R. Hassink (Eds.), The life cycle of clusters: A policy perspective (pp. 13–34). London: Edward Elger. Avcı, M., Uysal, S., & Taşcı, R. (2016). Türk İmalat Sanayinin Teknolojik Yapısı Üzerine Bir Değerlendirme. Sosyal ve Beşeri Bilimler Araştırmaları Dergisi, 17(36), 49–66. Balland, P., Boschma, R., Crespo, J., & Rigby, D. L. (2018). Smart specialization policy in the European Union: Relatedness, knowledge complexity and regional diversification. Regional Studies, 1–17. doi:10.1080/00343404.2018.1437900 Baer, C., & Brown, T. (2006). Location quotients: A tool for comparing regional industry compositions. Indianapolis: IBRC, Indian Workforce Development. BEBKA. (2012). Tr41 Bölgesi İlleri Kümelenme Analizleri. Bursa: Eskişehir, Bilecik Kalkınma Ajansı. Boschma, R. (2013). Constructing Regional Advantage and Smart Specialization: Comparison of Two European Policy Concepts. Papers in Evolutionary Economic Geography (PEEG) 1322, Utrecht University, Department of Human Geography and Spatial Planning, Group Economic Geography, revised November. Bulu, M., & Yalçıntaş, M. (2014). Clusters as a development policy tool: The Turkish experience. In M. Yülek (Ed.), Economic planning and industrial policy in the globalizing economy public administration, governance and globalization (pp. 331–344). London: Springer. Czamanski, S., Ablas, L. A. Q. (1979). Identification of industrial clusters and complexes: A comparison of methods and findings. Urban Studies, 16, 61–80. doi:10.1080/713702464 Cooke, P. (1992). Regional innovation systems: Competitive regulation in the new Europe. Geoforum; Journal of Physical, Human, and Regional Geosciences, 23(3), 365–382. doi:10.1016/ 0016-7185(92)90048-9 Cooke, P. (2012). Knowledge economy spillovers, proximity, and specialization. In B. T. Asheim, & M. D. Parrilli (Eds.), Interactive learning for innovation: A key driver within clusters and innovation systems (pp. 100–114). Basingstoke: Palgrave Macmillian. Çiftçi, M. (2018). Türkiye’de Bölgelerarası Sektörel Verimliliğin Analizi. İşletme Araştırmaları Dergisi, 10(1), 551–580. Çağlar, E., Kutsal, Y. (2011) Bölgesel Sanayi Politika Aracı Olarak özel Ekonomik Bölgeler: Türkiye’de Organize Sanayi Bölgelerinin Etkileri ve Yetkileri. 5. Bölgesel Kalkınma ve Yönetişim Sempozyumu Bildiriler Kitabı (pp. 187–202), Ankara: TEPAV. 20 N. ÇELIK ET AL. Deliktaş, E. & Çelik, N. (2018, May). Yozgat Bölgesi Sektörel Uzmanlaşma Analizi. Paper presented at the 3rd. International BOZOK Symposium. Dincer, B., Özaslan, M. & Kavasoğlu, T. (2003). İllerin ve Bölgelerin Sosyo-Ekonomik Gelişmişlik Sıralaması Araştırması. Devlet Planlama Teşkilatı, DPT 2671. Doğruel, F. (2006). Türkiye’de Bölgesel Politikalar. In A. Eraydın (Ed.), Değişen Mekan: Mekansal Süreçlere İlişkin Tartışma ve Araştırmalara Toplu Bakış: 1923-2003 (pp. 7–24). Ankara: Dost Kitabevi. Doğruel, F., & Doğruel, A. S. (2011). Privatization and Regional Distribution of Manufacturing in Turkey. TEA Discussion Paper 2011/4. Dulupçu, M. A., Karagöz, M., Sungur, O., & Ünlü, H. (2015). Clustering policies in Turkey: The impact of internationalization or the Imitation of Internationals. In R. Baptista, & J. Leitheo (Eds.), Entrepreneurship, human capital and regional development. International studies in Entrepreneurship (pp. 239–262). London: Springer. Elburz, Z., & Gezici, F. (2012). Regional Development Policies and Industrial Employment Change in Turkey: A Shift Share Analysis (1992-2008). 52. ERSA Congress, 21–25 August 2012, Bratislava. Eraydın, A. (2002). Yeni Sanayi Odakları: Yerel Kalkınmanın Yeniden Kavramsallaştırılması (New industrial centers: Re-conceptualization of local economy). Ankara: ODTÜ. Eraydın, A., & Köroğlu, B. A. (2005). Innovation, networking and the new industrial clusters: The characteristics of networks and local innovation capabilities in the Turkish industrial clusters. Entrepreneurship and Regional Development, 17(4), 237–266. doi:10.1080/08985620500202632 Eşiyok, A. (2013). Türkiye İmalat Sanayinin Teknolojik Yapısı: Sürdürülebilir mi? İktisat ve Toplum, 31–32. Eyyuboğlu, B.B. & Aktaş, S.G. (2017). Türkiye’de Teknoparkların Coğrafi Dağılım ve Yoğunluğu (2001-2015). Doğu Coğrafya Dergisi, 35, 75–88. Falcıoğlu, P. & Akgungor, S. (2008). Regional specialization and industrial concentration patterns in the Turkish manufacturing industry: An assessment for the 1980–2000 Period. European Planning Studies 16(2): 303–323. doi:10.1080/09654310701814678 Falcıoğlu, P. (2011). Location and determinants of productivity: The case of the manufacturing industry in Turkey. Emerging Market Finance and Trade, 47(5), 86–96. doi:10.2753/REE1540496X4706S506 Feser, E. J., & Bergman, E. M. (2000). National industry cluster templates: A framework for applied regional cluster analysis. Regional Studies, 34(1), 1–19. doi:10.1080/00343400050005844 Filiztekin, A., Barlo, M. & Kıbrıs, Ö. (2011). Türkiye’de Bölgesel Kalkınma: Farklılıklar, Bağıntılar ve Yeni Bir Mekanizma Tasarımı. TURKONFED, İstanbul. GEKA. (2011). TR32 Düzey II Bölgesi’nde Kümelenme Yaklaşımı. Güney Ege Kalkınma Ajansı. Gezici, F., Yazgı-Walsh, B. & Kacar, S.M. (2017). Regional and structural analysis of the manufacturing industry in Turkey. Annals of Regional Science, 1(59), 209–230. Hofe, R., & Bhatta, S. D. (2007). Method for identifying local and domestic industrial clusters using interregional commodity trade data. The Industrial Geographer, 4(2), 1–27. İZKA. (2010). İzmir Kümelenme Analizi. İzmir Kalkınma Ajansı. Karadağ, M., Deliktaş, E., & Önder, Ö. (2004). The effects of public capital on private sector performance in Turkish regional manufacturing industries. European Planning Studies, 12(8), 1145–1156. Karakayacı, Ö, & Dinçer, İ. (2012). Sanayi Kümelerinde Firma Özellikleri, Bilgi Ağları ve Yenilikçilik. Sigma, 4, 22–36. Kaygalak. (2011). Türkiye’de Sanayi Kümelenmesi: Uşak Örneği (PhD Dissertation). Ege Üniversitesi Sosyal Bilimler Enstitüsü. Kaygalak, İ. (2013). Türkiye’de Sanayi Coğrafyasında Endüstriyel Kümelenme ve Bölgesel Yoğunlaşma Eğilimi. Beşeri Coğrafya Dergisi, 1(1), 67–81. Kaygalak, İ. & Reid, N. (2016). The geographical evolution of manufacturing and industrial policies in Turkey. Applied Geography, 70, 37–48. doi:10.1016/j.apgeog.2016.01.001 Kazancık, L. B. (2007). Bölgesel Gelişme ve Sektör-Bölge Yığınlaşmaları. 2. Bölgesel Kalkınma ve Yönetişim Sempozyumu, İzmir, October, 25–26. EUROPEAN PLANNING STUDIES 21 Kıymalıoğlu, Ü, & Ayoğlu, D. (2006). Türk İmalat Sanayinde Yığılma Ekonomileri. Doğuş Üniversitesi Dergisi, 7(2), 198–209. Kopczewska, K., Churski, P., Ochojski, A., & Polko, A. (2017). Measuring regional specialization: A new approach. New York: Palgrave. Krugman, P. (1991). Increasing returns and economic geography. Journal of Political Economy, 99 (3), 483–499. Landabaso, M. (2012). Research and innovation strategies for smart specialization. United Nations Economic Commission for Europe-Building Strategies for Regions of Innovation, Salle VII, Palais des Nations, Geneva, 12–13 April. Makelainen, K. (2014) Metal and Engineer and Industries and Innovative Clusters in Turkey. RSA European Congress. Izmir. Marshall, A. (1920). Principles of Economics: An Introductory Volume. London: Macmillanand Co. Morgan, K. (2013). The regional state in the era of smart specialization. Ekonomiaz, 83(2), 103–125. OECD. (2013). Supporting investment in knowledge capital, growth and innovation. Geneve: OECD Publishing. O’hUallachain, B. (1984). The identification of industrial complexes. Annals of the Association of American Geographers, 74(3), 420–436. Öngel, F.S. (2018). Türkiye’de Sanayinin Mekansal Dağılımı ve Yeni Sanayi Odakları. TMMOB Makine Mühendisleri Odası. Öz, Ö. (2002). Assessing Porter’s framework for national advantage: The case of Turkey. Journal of Business Research, 55, 509–515. doi:10.1016/S0148-2963(00)00167-3 Öz, Ö. (2004). Clusters and competitive advantage: The Turkish experience. New York: Palgrave MacMillan. Porter, M. (1999). Clusters and the new economics of competition. Harvard Business Review, (November–December), 77–90. Sanayi ve Ticaret Bakanlığı. (2007). Bölgesel Rekabet Edebilirlik Operasyonel Programı. Ankara: Sanayi ve Ticaret Bakanlığı (Republic of Turkey, Ministry of Industry and Trade). Sungur, O. (2015). Tr61 (Antalya, Isparta, Burdur) Bölgesinde Sektörel Yoğunlaşmanın ve Yoğunlaşma Dinamiklerinin Analizi. Yönetim ve Ekonomi Araştırmaları Dergisi, 13(3), 316–341. Türkcan, B. (2015). Knowledge externalities in Turkish high-technology manufacturing industries: Spatial panel data analyses. Journal of Yasar University, 10(40), 6711–6729. T.C. Bilim Sanayi ve Teknoloji Bakanlığı. (2018). Türkiye’nin Sanayi Devrimi. Dijital Yol Haritası. Ankara. TÜBİTAK. (2016). Yeni Sanayi Devrimi: Akıllı Üretim Sistemleri Teknoloji Yol Haritası. Tübitak Bilim, Teknoloji ve Yenilik Politikaları Daire Başkanlığı. TÜSİAD. (2005). Türkiye’de Bölgesel Gelişme Politikaları: Sektör-Bölge Yığınlaşmaları Özet Bulgular. TÜSİAD Basın Bülteni, Ankara. Yolchi, J., & Akseki, U. (2018). Türk İmalat Sanayindeki Uzmanlaşmanın Teknoloji Düzeyine Göre Bölgesel Bir Analizi. Akademik Araştırmalar ve Çalışmalar Dergisi, 10(18), 134–145. Appendix Using the approach developed by Czamanski and Ablas (1979), we first develop matrices X and Y reflecting the technical input–output coefficients. Xij = aij a ji aij a ji ; X ji = ; Yij = ; Yij = pj pi sj sj Where Xij and X ji : good and services purchases by j (i) from i (j) as a ratio of j’s (i’s) total good and services purchases. Therefore, a large value for Xij, reflects that industry j depends on industry i as a source for a large proportion of its total inputs (X Matrix). Yij and Yij : good and services sales from i (j) to j (i) as a proportion of i’s (j’s) total good and services purchases. Therefore, a large value for Yij reflects that industry i depends on industry j as a market for a large proportion of its total sales (Y Matrix). 22 N. ÇELIK ET AL. Each column in X matrix represents the intermediate input purchasing pattern of the column industry and the sum of the columns should add to unity. Similarly, each row in Y matrix shows the intermediate output selling pattern of the row industry and the sum of each row should also add to unity. Four matrices using the matrices X and Y are calculated. Elements of the first matrix (X correlation matrix) are the correlations between the columns of matrix X. This resulting matrix gives the degree to which pair of industries has a similar input-purchasing pattern. Elements of the second matrix (Y correlation matrix) are the correlations between the columns of matrix Y. Matrix Y represents the degree to which pair of industries have similar output selling patterns. The third matrix shows the degree to which the buying pattern of an industry is similar to the selling pattern of the other industries (X–Y correlation matrix). Elements of the X–Y correlation matrix are the correlations between the columns of matrix X and matrix Y. The elements of the fourth matrix (Y–X correlation matrix) are the correlations between the rows of matrix Y and matrix X. The X–Y correlation matrix gives the degree to which the selling pattern of an industry is similar to the buying pattern of other industries (Feser & Bergman, 2000). Finally, the largest values of each cell among the four correlation matrices defined earlier are selected and a symmetric matrix (matrix Lv) is constructed. The columns of the Lv symmetric matrix describe the pattern of relative linkage between the column industry and all other industries. To cluster industries with similar selling and purchasing patterns, the Lv matrix is used for principal component factor analysis with orthogonal rotation. The relative linkage between a given industry and the derived factor can be measured by the generated set of loadings. Following Feser and Bergman (2000) industries with loading 0.60 or higher, on a given cluster can be viewed as strongly linked to that cluster (primary industries), whereas industries with loading 0.35 and 0.59 are accepted as moderately and weakly linked to the other economic activities (secondary industries).