Export variety, technological content and the economic performance of countries. The case of Portugal

Export variety, technological content and the economic performance of countries. The case of Portugal Francisco Rebelo Master Dissertation in International Economics and Management Supervisor: Ester Gomes da Silva 2012 i Acknowledgements A master dissertation is the outcome of a long period of research, which can often be a very lonely process. However, it would not have been accomplished without the collaboration of many people, who helped me in a variety of different ways. First and foremost, I would like to acknowledge the contribution of my supervisor, Professor Ester Gomes da Silva, who always supported me from the beginning of the thesis project until the conclusion of the dissertation. I would like to thank for her constant availability throughout this work, for the helpful suggestions and for sharing her knowledge and experience in a variety of subjects. I also owe a great debt of gratitude to Professor Aurora Teixeira for the countless suggestions and support, ever since the choice of the topic. Thanks are also owing to Professor Ana Teresa Lehmann for her helpful comments on my thesis proposal. Also to my Professors from the Master Programme, as their knowledge in the international economics and business areas was very important and helpful in numerous ways. Last but not least, a special acknowledgement goes to my family and friends for their continuous support and encouragement. To all these people I express my deepest gratitude! ii Abstract Although the analysis of the relationship between international trade and economic growth has an important tradition in the economic literature, the specific focus on a related matter, the link between export variety and economic growth, remains a relatively unexplored field of research. A few studies have recently approached this issue from a neo-Schumpeterian framework, which emphasizes the irreversibility and path dependency features of international trade flows and their connection with the economic performance of countries. In line with these recent developments, we investigate the joint impact of growth in variety and change in the technological content of exports in economic growth, focusing on a rather unexplored context, the Portuguese case. We consider a long time span, of almost half a century (1967-2010), marked by considerable change and increasing integration of Portugal in the world economy. The econometric analysis is performed with resort to cointegration techniques, using export data from the CHELEM database, combined with macroeconomic indicators from other sources. The evolution in the technological content of exports is addressed through the use of technological classification schemes. The evidence obtained shows that increasing related variety has led to a significant growth bonus in Portugal during the last four decades, but only in the case of technology advanced sectors. The impact of export variety on economic growth seems therefore to be conditioned by the technological intensity of the products involved. Keywords: Trade; variety; economic growth; technical change; path dependency. JEL-Codes: F10; O11; O30; O52. iii Resumo Apesar da análise da relação entre comércio internacional e crescimento económico ser um tema recorrente na literatura económica, um tópico específico deste campo de investigação permanece relativamente inexplorado: a relação entre o aumento na variedade das exportações e o crescimento económico. Alguns estudos abordaram recentemente esta questão numa perspetiva teórica neo-Schumpeteriana, que enfatiza as características de irreversibilidade e de dependência temporal dos fluxos de comércio internacional e a sua ligação com o desempenho económico dos países. Em linha com estes recentes desenvolvimentos, este estudo procura investigar o impacto conjunto de dois fatores, o crescimento da variedade e a alteração no conteúdo tecnológico das exportações, no crescimento económico, abordando um contexto relativamente inexplorado, o caso português. A análise considera um longo período de tempo, de aproximadamente meio século (1967-2010), que foi marcado por mudanças significativas e por uma integração crescente da economia portuguesa na economia mundial. A análise econométrica é efetuada com recurso a técnicas de cointegração, utilizando informação sobre o valor das exportações da base de dados CHELEM, recorrendo a outras fontes de informação para os restantes indicadores macroeconómicos. A análise da evolução do conteúdo tecnológico das exportações é concretizada por intermédio da utilização de taxonomias sectoriais de base tecnológica. Os resultados revelam que o aumento da variedade das exportações teve um impacto positivo no crescimento económico português durante as últimas quatro décadas, mas apenas no caso de sectores relacionados e tecnologicamente avançados. O impacto da variedade das exportações no crescimento económico parece assim estar dependente da intensidade tecnológica dos produtos em consideração. Palavras-Chave: Comércio; variedade; crescimento económico; mudança tecnológica; path dependency. Códigos JEL: F10; O11; O30; O52. iv Index of Contents Acknowledgements .......................................................................................................... i Abstract ............................................................................................................................ ii Resumo ............................................................................................................................ iii Index of Contents ........................................................................................................... iv Index of Tables ............................................................................................................... vi Index of Figures ............................................................................................................ vii Abbreviations ................................................................................................................. ix Introduction ..................................................................................................................... 1 1. Export variety and economic growth. A critical appraisal of the literature ......................................................................................................... 4 1.1. The concept of variety ........................................................................................... 4 1.2. The relationship between variety and economic growth ....................................... 5 1.3. Empirical studies focusing on the relationship between export variety and economic growth .......................................................................................... 10 1.4. An assessment of the evidence found in the Portuguese case and thesis’ main research issues ............................................................................................ 14 2. Trends in export variety and long-term economic growth: the case of Portugal, 1967-2010 ........................................................................ 20 2.1. The economic background: growth patterns and main features of the process of economic integration of Portugal over the last four decades ............ 20 2.2. The structure of Portuguese exports, 1967-2010 ................................................. 27 2.2.1 Changes in the direction of trade ............................................................... 27 2.2.2. Changes in the composition of exports ..................................................... 30 2.3. Trends in export variety ....................................................................................... 41 v 2.4. Comparison of the Portuguese case with other cohesion countries’ experiences .......................................................................................................... 46 3. Export variety and economic growth: an econometric assessment ........................................................................................................... 57 3.1. The model ............................................................................................................ 57 3.2. Estimation method ............................................................................................... 59 3.3. Econometric results .............................................................................................. 63 Conclusion ..................................................................................................................... 68 References ...................................................................................................................... 72 Appendix ........................................................................................................................ 81 vi Index of Tables Table 1: Studies that analyse the relationship between variety and economic growth .. 11 Table 2: Concentration of Portuguese international trade ............................................. 27 Table 3: Export shares by type of good (%) .................................................................. 30 Table 4: Exports shares by industry (%) ........................................................................ 32 Table 5: OECD taxonomical categories ........................................................................ 33 Table 6: Evolution of Portuguese exports according to Tidd and Bessant’s (2009) categories (1967-2010; %) ............................................................................. 35 Table 7: Top 10 ISIC-4 industries in total exports ........................................................ 40 Table 8: Manufacturing exports by technological content (average 2008-2010) .......... 48 Table 9: Unit root tests – variables in levels .................................................................. 62 Table 10: Unit root tests – variables in first differences ................................................ 62 Table 11: Lag length selection ....................................................................................... 64 Table 12: Johansen’s cointegration test results .............................................................. 65 Table 13: Normalized cointegrating coefficients ........................................................... 65 Table 14: Classification of industries ............................................................................ 81 vii Index of Figures Figure 1: Portuguese exports and GDP growth in constant 2005 prices (1967-2010) .. 20 Figure 2: Exports and GDP growth trend (Portugal, 1967-2010) ................................. 21 Figure 3: Trade openness (Portugal, 1967-2010) .......................................................... 21 Figure 4: Portuguese exports and imports as a percentage of GDP (1967-2010) ......... 22 Figure 5: Portuguese exports and imports as a percentage of GDP in constant 2005 prices (1967-2010) ....................................................................................... 22 Figure 6: Real effective exchange rate based on unit labour costs relative to EU-15 countries (Portugal, 2000 = 100, 1967-2010) .............................................. 23 Figure 7: Trade Balance as a percentage of GDP (Portugal, 1967-2010) ..................... 24 Figure 8: Coverage of imports by exports (Portugal, 1967-2010) ................................. 24 Figure 9: Portuguese exports by receiving continent (%) ............................................. 28 Figure 10: Portuguese export shares outside EU-27 ...................................................... 28 Figure 11: Main destinations of Portuguese exports (%) .............................................. 29 Figure 12: Manufacturing exports by technological content ......................................... 34 Figure 13: Revealed comparative advantage by technological content (Portugal, 19672010) ............................................................................................................ 36 Figure 14: Revealed comparative advantage within Tidd and Bessant’s (2009) industry groups (Portugal, 1967-2010) ...................................................................... 36 Figure 15: Balassa specialization coefficient by technological content ........................ 37 Figure 16: Balassa specialization coefficient within Tidd and Bessant’s (2009) industry groups ........................................................................................................... 38 Figure 17: Total export variety (Portugal, 1967-2010) ................................................. 43 Figure 18: Related variety (Portugal, 1967-2010) ......................................................... 43 Figure 19: Unrelated variety (Portugal, 1967-2010) ..................................................... 44 viii Figure 20: Related variety by technological content (Portugal, 1967-2010) ................. 45 Figure 21: Related variety using Tidd and Bessant’s (2009) industry groups ............... 46 Figure 22: Manufacturing exports by technological content ......................................... 47 Figure 23: Evolution of exports of Tidd and Bessant’s (2009) industry groups ........... 50 Figure 24: Total entropy, related and unrelated variety ................................................ 52 Figure 25: Related variety by technological content in Spain, Ireland, Greece and Italy .............................................................................................................. 53 Figure 26: Related variety using Tidd and Bessant’s (2009) industry groups in Spain, Ireland, Greece and Italy .............................................................................. 54 Figure 27: Portuguese labour productivity per hour worked in 1990 US dollars .......... 60 Figure 28: Natural logarithm of the average number of years of formal education of the working age population (Portugal, 1967-2010) ........................................... 60 Figure 29: Natural logarithm of the share of investment in GDP (Portugal, 19672010) ............................................................................................................ 61 5 than products and services belonging to different sectors (Saviotti and Frenken, 2008). Unrelated variety, on the other hand, refers to the variety between the main sectors of the economy, representing the entry of new products and services that are unrelated to the pre-existing ones. Capabilities required to produce related variety are similar to the already existing on the economy, and thus easier to acquire than the capabilities necessary to the production of unrelated varieties. Moreover, as the capabilities and institutions of a specific sector can be easily transferred to related sectors, an increase in related variety is easier to accomplish than an increase in unrelated variety (Saviotti and Frenken, 2008).4 1.2. The relationship between variety and economic growth The notion of variety addressed in the previous section is mostly related to neoSchumpeterian and evolutionary streams of research. Under this approach, “economic development is perceived as a dynamic, cumulative, open-ended process far from equilibrium paths that is subject to historical contingencies which cause the process to be path-dependent and irreversible” (Kruger, 2008, p. 344). According to the views expressed within this theoretical frame, three major types of relationships can be envisaged between variety and economic growth (Frenken et al., 2007).5 The first type, centred on the inter-relatedness features of variety, knowledge spillovers and economic growth, states that spillovers can occur not only between firms within a sector but also between sectors. This means that the composition of the economy may affect growth, with countries specializing in a particular composition of complementary sectors experiencing higher growth.6 A second type of relationship sees variety within the context of a portfolio strategy that can be used to protect a country from external 4 The conceptual distinction between related and unrelated variety is reflected in their measurement, with the former being generally measured at lower levels of aggregation. Frenken et al. (2007), for example, measure unrelated variety at the two-digit level and related variety at the five-digit level within two-digit sectors, whereas Saviotti and Frenken (2008) measure unrelated variety using one-digit export data and related variety with three-digit export data. 5 As indicated earlier, variety is related to the notion of structural change, whose impact on economic growth has been documented by a vast amount of research (e.g., Kuznets, 1971; Fagerberg, 2000; Wang and Szirmai, 2008; Silva and Teixeira, 2011). 6 The inflow of external knowledge may also promote economic growth, if combined with regional absorptive capacity. The cognitive proximity between the external knowledge and the existing knowledge base should be neither too small (as it would not add something very relevant to the local knowledge base) nor too large. In fact, if the external knowledge is unrelated to the existing competences, the region may not be able to absorb it and thus will probably not reap benefits from its use (Boschma and Iammarino, 2009). 6 shocks. Following this line of reasoning, specialization is generally related to a higher vulnerability to demand shocks in growth and employment variables. Because unrelated variety refers to sectors that do not possess substantial input-output linkages, in the presence of a sector-specific shock, the economy is less likely to be disturbed as a whole (Boschma and Iammarino, 2009). The third type of relationship, stemming from Pasinetti’s (1981, 1993) seminal work on the relationship between growth and structural change, addresses the long-term effect of variety over the economic system. Labour that has become redundant in pre-existing sectors of an economy, due to productivity increases and demand saturation, as predicted by Engel’s law, can only be absorbed by the emergence of new sectors, which promotes growth in the long-run. Based on this latter type of relationship, Saviotti and Frenken (2008) put forward two main hypotheses regarding the links between variety and the economic performance of countries. The first one states that “growth in variety is a necessary requirement for long-term economic development”, whereas the second claims that “variety growth, leading to new sectors, and productivity growth in pre-existing sectors, are complementary and not independent aspects of economic development” (Saviotti and Frenken, 2008, p. 206). The rationale behind these hypotheses lies on the imbalance between productivity growth and demand growth, as derived in Pasinetti’s (1981, 1993) theoretical scheme. In fact, assuming that the set of activities of an economy remains constant over time, the combination of growing productivity with the tendency towards demand saturation would inevitably lead to structural unemployment, as it would be possible to produce all goods and services with a decreased proportion of inputs (including labour). The emergence of new sectors thus works as a means to compensate for the release of resources determined by productivity growth. Moreover, search activities are required to generate new goods and services,7 which means that an increase in the efficiency of pre-existing sectors is required, in order to allocate resources to these activities (Saviotti and Frenken, 2008). In the context of an open economy, however, the problem of demand saturation may not constitute such a significant bottleneck, at least in the short run (Saviotti and Frenken, 2008). Countries that gain market shares with international trade can continue to 7 According to Saviotti and Mani (1998, p. 255), search activities are defined as “activities that scan the external environment in order to find either alternatives to existing routines or completely new routines”. 7 specialize in a number of sectors, provided that exports in these sectors keep growing. Either way, as new sectors keep emerging worldwide, the share of trade of a country’s sectors of specialization will ultimately decrease, even if it achieves a monopoly in one or more sectors. Specialization in pre-existing sectors will likely run into diminishing returns, and therefore, even in the context of an open economy, export variety growth is still expected to promote long-run economic growth (Saviotti and Frenken, 2008). The mechanism by which the emergence of new sectors generates long-run economic growth is further elaborated in a model developed by Saviotti and Pyka (2004a, 2004b). Defining each sector as a population of firms that produce a differentiated product, the emergence of an innovation capable of establishing a new market generates an adjustment gap, which represents the size of the potential market of a given product.8 The first firm entering a new market enjoys a temporary monopoly which is eroded by the entry of imitators, inducing the incumbent firms to exit and create new innovations. Competition may arise not only within the sector (intra-population or intra-sector), but also in other sectors (inter-population or inter-sector), as goods and services produced by firms are represented not only by technical characteristics, resulting from firms’ inventive activities, but also by service characteristics, which are those ultimately desired by consumers. In these terms, economic growth results from the emergence of new activities, the disappearance of the old ones and the changing weight of all economic activities, which is in broad agreement with the Schumpeterian notion of creative destruction. Taking into account the aforementioned relationship between variety and economic growth, countries can follow multiple paths according to the economic policies they pursue (Saviotti, 2003; Saviotti and Frenken, 2008). Saviotti (2003) distinguishes between three main catch-up strategies. The first one refers to the case in which both the relative output variety and the unit prices and/or the quantities produced by a country increase. The country increases exports from more high-technology sectors at a faster pace than the world economy, which constitutes a strategy of creative de-specialization. A second strategy, defined as virtuous specialization, refers to the case in which the 8 Saviotti (2003) also argues that an innovation may not be immediately better than pre-existing products and services, generating instead a niche that gradually can improve and only later become a developed market by itself, as a result of the learning effects gathered in the niche. 8 relative output variety of a country decreases, but the country produces or sells larger quantities of more expensive products.9 Finally, the third strategy occurs when a country stimulates relative output variety in order to compensate for the falling share of total revenues, either by decreasing prices or quantities. This is a strategy of vicious despecialization, in which the country increases its range of low-technology products. Saviotti (2003) stresses that in a context marked by the worldwide continuous emergence of new sectors, countries that pursue a catch-up strategy based on specialization must be aware that national variety will have to increase, even at a smaller rate than world variety. In other words, due to the continuous emergence of new sectors, it is not possible to sustain that a particular specialization strategy can be kept indefinitely. A different approach to the analysis of export variety and its relationship with economic growth is performed under a stream of research more in line with mainstream economics. Within this frame of analysis, several studies have been produced, following Krugman (1979, 1980, and 1981) and Helpman’s (1981) monopolistic competition models, in which important gains arise from the increasing variety of products resulting from international trade. Based on the aggregate production function modelling framework, these studies endogenize variables such as product or export variety, which become important drivers of the economic performance of countries (e.g. Feenstra et al., 1999; Funke and Ruhwedel, 2001a; Funke and Ruhwedel, 2001b; Funke and Ruhwedel, 2005; Feenstra and Kee, 2008). Within this approach, many studies draw upon the methodology developed by Feenstra (1994), who derives an exact price index from a CES unit-cost function that allows for changes in the sets of product varieties and quality change in some of the varieties. The measure of product variety used in these studies is designed to assure consistency across countries and over time. Feenstra and Kee (2008), for example, weight the set of goods exported by a given country to the USA, using the worldwide exports from all countries to the USA as a benchmark. 9 The virtuous specialization path can assume two further sub-divisions: if prices increase, but the country sells fewer quantities, there is a move towards high-technology or up-market goods; whereas the opposite situation may be associated with the exploration of scale economies or increased efficiency (Saviotti, 2003). 9 Funke and Ruhwedel (2001a, 2005), in their turn, adapt a semi-endogenous growth model, stemming from the work of Jones (1995), which considers technological change endogenous and the long-run growth of the economy dependent of exogenous variables (Jones, 1995; Funke and Ruhwedel, 2005). In the model, an increase in product variety increases per capita income levels by the fuller realization of dynamic economies of scale. Moreover, the larger the gap between the so-called “technology leader” and the catching-up economy, the higher the scope for differential variety growth, since it is easier to introduce products that are already available in other countries, in terms of R&D investment, than to develop totally new ones. Funke and Ruhwedel (2005) also demonstrate that government policies leading to the opening up of the economy have a long-run level effect, but not a long-run growth effect, as in the original Solow model. Another influential work within this stream of research was developed by Feenstra and Kee (2008). The authors introduce a monopolistic competition model with heterogeneous firms that aims at explaining the relationship between export variety and productivity levels. Firms must have higher productivity than a “cut-off” productivity level to actually produce, being the optimal “cut-off” defined at the socially optimal level. However, in each sector only the most productive firms become exporters, whereas the others find it profitable to produce exclusively for their domestic markets. Therefore, if the share of exporting firms increases, and consequently, there is a rise in the share of exported varieties, then the average productivity will increase and countries will grow faster. To summarize, quite independently of the specific theoretical stream of research undertaken (neo-Schumpeterian, evolutionary or mainstream economics), there is a systematic finding according to which there is a positive relationship between international trade variety and the economic performance of countries, whether measured by productivity or employment variables. This relationship is, however, approached in rather different ways by each of the streams of research, which rely on distinct economic explanations, as they depart from different theoretical paradigms. 10 1.3. Empirical studies focusing on the relationship between export variety and economic growth Recently, a number of empirical studies have focused on the relationship between product variety and the economic performance of countries.10 In line with the streams of research explored in the previous section, these studies use different indicators to measure variety: some resort to the indicator derived from a CES production function developed by Feenstra (1994) (e.g. Feenstra et al., 1999; Funke and Ruhwedel, 2005; Feenstra and Kee, 2008), while others use the entropy coefficient (e.g. Frenken et al., 2007; Saviotti and Frenken, 2008; Boschma and Iammarino, 2009). The unit of analysis is also variable across studies. Some studies consider groups formed by several (heterogeneous) countries (e.g. Funke and Ruhwedel, 2005; Feenstra and Kee, 2008; Saviotti and Frenken, 2008), whereas others take the region as the geographical unit of analysis (e.g. Frenken et al., 2007; Boschma and Iammarino, 2009; Boschma et al., 2012). Looking first to the group of studies that measure export variety with recourse to the CES production function, a common finding is that export variety influences positively the economic performance of countries. Such evidence is found by Feenstra et al. (1999) with respect to South Korea and Taiwan. The authors find that export variety increases productivity growth in secondary industries, although the evidence found with respect to the primary sector is somewhat mixed. Funke and Ruhwedel (2001a), in their turn, find a significant and positive correlation between product variety of 18 OECD countries relative to the United States, and their relative per capita income levels. Countries that direct trade policies outwards have access to a greater variety of products and technologies, and thus are able to grow at faster rates. In another work, which uses data from 10 East Asian economies, Funke and Ruhwedel (2001b) also show that countries with greater product variety have better export performances: greater variety is related to an increasing competitive advantage which promotes an increase in export flows. Addison (2003) investigates the correlation between product variety growth and total 10 Table 1 provides a summary description of these studies. 11 Table 1: Studies that analyse the relationship between variety and economic growth Indicator used to measure variety Author(s) Geographical Scope Main Data Sources Disaggregation level Period Method Main Control Variables Derived from a CES function as developed by Feenstra (1994) Feenstra et al. (1999) South Korea and Taiwan US Import Statistics; Statistical Yearbook of the Republic of China; Input – Output Table for South Korea Seven-digit (197288); Ten-digit (1989-91) (Product) 1972-1991 Pooled OLS; SUR Electricity use by each industry (to control for excess capacity); Growth of imports and exports in each industry Funke and Ruhwedel (2001a) 19 members of OECD OECD database International Trade by Commodities Statistics (ITCS); World Development Indicators Six-digit (Product) 1989-1996 Panel data estimation (Fixed effects) Relative investment share Funke and Ruhwedel (2001b) 10 East Asian countries OECD International Trade by Commodities Statistics (ITCS) Six-digit (Product) 1989-1997 GMM Real effective exchange rate Addison (2003) 29 countries (13 rich and 16 poor) World Bank; OECD International Sectoral Data Base Five-digit (Product) 1979-1986 Pooled OLS Period average growth in the real exchange rate Funke and Ruhwedel (2005) 14 Eastern European countries OECD International Trade by Commodities Statistics (ITCS); World Development Indicators Five-digit (Product) 1993-2000 Panel data cointegration Relative investment share, Transition indicator EBRD Feenstra and Kee (2008) 48 countries World Development Indicators; UNIDO; US import statistics Seven-digit (198088); Ten-digit (1989-2000) (Product) 1980-2000 Non-linear Three Stage Least Squares Relative labour–land ratio, Capital– land ratio, Relative land area and non-traded good prices Entropy coefficient Frenken et al. (2007) The Netherlands NUTS 3 Statistics Netherlands (CBS); University of Groningen; LISA Five-digit (Sector employment data) 1996-2002 Pooled OLS Investment, R&D, Capital–labour ratio growth, Human capital, Wage level, Business area growth, Dwellings growth Saviotti and Frenken (2008) 20 OECD countries OECD; University of Groningen – The Groningen Growth and Development Centre Three-digit (Sector) 1964-2003 Pooled OLS Not included Boschma and Iammarino (2009) Italy NUTS 3 Italian National Institute of Statistics (ISTAT) Three-digit (Sector) 1995-2003 Pooled OLS Population density Boschma et al. (2012) Spain NUTS 3 Spanish Statistical Institute’s (INE); Instituto Valenciano de Investigaciones Económicas (Ivie); Dirección General de Aduanas - Agencia Tributaria Six-digit (Product) 1995-2007 Pooled OLS Population density; Labour productivity; Human capital; Level of employment Hartog et al. (2012) Finland NUTS 4 Statistics Finland Five-digit (Sector) 1993-2006 GMM Regional population density; Human capital; R&D expenditures 12 factor productivity growth, considering a set of 29 countries, which includes both developed and developing countries. Along with the finding of a significant relationship between the two variables, the author acknowledges the important role played by imitation of variety as a source of productivity growth in developing countries, especially when accompanied by relatively high educational attainment. In the case of developed countries, however, the role of product variety is less important as a source of productivity growth. A positive impact of product variety on economic growth is also found by Funke and Ruhwedel (2005), in their study of 14 Eastern European transition countries. The authors show that these countries have benefited in a significant manner from the increase in variety of products following the transition towards a market-based organization. Feenstra and Kee (2008) find that export variety is quite effective in explaining the time-series variations in productivity within countries between 1980 and 2000, especially in the case of OECD countries, but that it fails to explain the large absolute differences in productivity between countries in a given moment in time. In the second group of studies, more in line with neo-Schumpeterian reasoning, the entropy coefficient is usually used in the computation of variety. Despite using a different methodology, a positive (and significant) relationship is also typically found between variety and growth. Frenken et al. (2007), for example, find that related variety is a source of Jacobs externalities in their study based on the Dutch economy, due to knowledge spillovers that enhance growth and employment. Moreover, unrelated variety is negatively related to unemployment growth, which confirms the “portfolio effect”, according to which the diversification into different sectors works as a means of protection against external shocks in demand. In a different work, based on data from 20 OECD economies, Saviotti and Frenken (2008) find that countries with the highest level of economic growth present also the highest levels of export variety. The results are sensitive, however, to the type of variety considered: whereas related variety emerges as a determinant of growth in the short run, whether measured by GDP per capita or labour productivity growth, unrelated variety only achieves significance in a broader time horizon. A different impact of related and 13 unrelated variety is also found in a recent study of Boschma et al. (2012), based on data from Spanish regions between 1995 and 2007. Although total variety has a positive effect on regional growth, when it is decomposed in its related and unrelated components, only the former remains statistically significant. Moreover, the use of alternative measures of relatedness between industries, such as Porter’s (2003) cluster classification and the products’ proximity index developed by Hidalgo et al. (2007), leads to a stronger effect of related variety than unrelated variety on regional growth. A similar finding had also been obtained by Boschma and Iammarino (2009) in an earlier study focusing on the Italian experience. In this case, related variety has always a positive impact on value-added growth, while unrelated variety has a positive and significant effect only in two specifications.11 Moreover, if a province has similar import and export specialization patterns, its value-added growth will be lower, as this may be a sign that one of several situations may occur: first, that no value is added in the region, reflecting only pure transit flows; second, that the external knowledge brought through imports did not contribute to the local knowledge base, or finally that export sectors might have suffered from greater international competition (Boschma and Iammarino, 2009). Very recently, Hartog et al. (2012) addressed the impact of related variety on regional employment growth taking into account the technological intensity of industries (separating the high-tech from the medium/low-tech categories) for the case of Finland, finding that only related variety among high-tech industries has a positive and significant effect on regional employment growth. As a general conclusion, the survey undertaken shows that, quite irrespectively of the geographical scope, period or methodology adopted, empirical studies find systematically a positive and significant relationship between increasing variety and the economic performance of countries. The main difference relies on the impact of related and unrelated variety: whereas the former has generally a positive effect on economic growth, the effect of the latter differs across studies, being in many cases statistically non-significant. 11 Taking labour-productivity growth or regional employment growth as the dependent variable, related variety has a positive and significant effect in only one specification, while unrelated variety is always insignificant. 14 1.4. An assessment of the evidence found in the Portuguese case and thesis’ main research issues A few studies addressed the changes in the composition of international trade flows in the Portuguese case covering the 20th century and the more recent years (e.g. Afonso and Aguiar, 2005; Amaral, 2006; Amador et al., 2007; Cabral, 2008; Leite, 2010; Freitas and Mamede, 2011). These studies describe the main patterns of Portuguese international trade and its comparison with other European countries’ experiences, drawing several conclusions about the changing composition of Portuguese exports over time. The analysis of variety is performed, however, in essentially descriptive terms, with no attempt to address the causality chains between export variety, technology and economic growth. According to the evidence reported in these earlier works, a general trend of increasing diversification is characteristic of the evolution of Portuguese exports over the last half a century. Until the 1960s, Portuguese exports were to a great extent composed by consumption goods, in which an important role was played by wine, particularly Port wine. The spread of the industrialization process in the 1950s and 1960s led to an impressive change in the composition of Portuguese exports, and by the end of the 1960s non-food consumption goods achieved predominance (Afonso and Aguiar, 2005). Export variety was still rather limited, however, with clothing and footwear representing approximately a quarter of total exports, a share similar to that observed with respect to Port wine in the end of the 19th century. The available evidence also indicates that a progressive movement of reinforcement of durable goods with higher technology and value added has taken place during the period under study. Amador et al. (2007) show that the share of low-tech sectors in total manufacturing exports has been continuously decreasing over time (from more than 75 per cent in 1967-1969, to around 40 per cent in 2000-2004), with the decline being particularly sharp in the case of “Food products, beverages and tobacco” and “Textiles, textile products, leather and footwear”. There has been also a significant increase in the exports of medium-high-technology sectors, particularly in the case of “Motor vehicles, trailers and semi-trailers”, mostly related to foreign direct investment. The shares of 21 Figure 2: Exports and GDP growth trend (Portugal, 1967-2010) Note: Trend obtained from annual data using a Hodrick-Prescott (1997) filter (λ = 100) Source: European Commission (AMECO database) and own calculations The strong relevance of exports growth as a source of Portuguese economic growth in the last decades is acknowledged by several studies, which refer the growing openness to international trade as one inescapable feature in the development path pursued by the Portuguese economy after the Second World War (e.g., Afonso and Aguiar, 2005; Amaral et al., 2007). In line with this general view, Figure 3 depicts the increasing openness of the Portuguese economy since 1967, which is particularly evident when measured at constant prices. Figure 3: Trade openness (Portugal, 1967-2010) Source: European Commission (AMECO database) and own calculations 0% 10% 20% 30% 40% 50% 60% 70% 80% 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 Current Prices 2005 Constant Prices 0% 2% 4% 6% 8% 10% 12% 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 Export growth GDP growth 22 0% 5% 10% 15% 20% 25% 30% 35% 40% 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 Exports Imports 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 Exports Imports Starting from approximately 17% of GDP in 1967, total trade flows at 2005 prices increased to more than 70% of Portuguese GDP in 2010. In current prices, the external trade to GDP ratio started slightly above 40% and reached about 65% in the beginning of the eighties, fluctuating around that value until the late 1990s. In the more recent years there has been a new tendency of increase, with a peak being found in 2008, at about 3/4 of total GDP. Figure 4: Portuguese exports and imports as a percentage of GDP (1967-2010) Source: CHELEM database and own calculations Figure 5: Portuguese exports and imports as a percentage of GDP in constant 2005 prices (1967-2010) Source: AMECO database and own calculations 23 The increase in the trade openness ratio was somewhat to be expected, given the combined influence of the reduction in both the real costs of trade and distance, through the improvement in road and telecommunications infrastructures, the increase in intraindustry trade, resulting from greater demand of variety from consumers, and the growing importance of vertical specialization activities, which materialized in greater imports of intermediate goods (Amador et al., 2007). But the increase in international trade flows and the ways in which it varied over time are, of course, eminently related to the specificities of the Portuguese recent economic history. Moreover, although exports and imports have increased over time, imports increased at a faster pace (Figures 4 and 5), a pattern which essentially reflects the serious competitiveness problems of the Portuguese economy. In fact, from the late eighties onwards, the Portuguese real effective exchange rate has appreciated nearly 40 p.p. (Figure 6), which is indicative of a strong deterioration of the competitiveness of Portuguese exports. Figure 6: Real effective exchange rate based on unit labour costs relative to EU-15 countries (Portugal, 2000 = 100, 1967-2010) Source: AMECO database Over the 40 years under study, the trade balance was always negative,20 due in part to the country’s strong energy dependence, especially in the aftermath of the oil crises 20 In fact, the trade balance was negative over the whole 20th century, with the exception of the 1941-43 period, when Portugal benefited from an increase in the price of wolfram and other raw materials exports (Afonso and Aguiar, 2005). 0 20 40 60 80 100 120 140 160 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 24 (Figure 7). This imbalance is also seen in the exports to imports ratio, which despite some improvement in the first half of the 1980s, following the IMF adjustment, assumes always values below 1, and has been rather stable from the mid-eighties onwards (cf. Figure 8). In 2010, it assumes a value around 0.64, which means that Portuguese exports can buy less than two thirds of total imports. Figure 7: Trade Balance as a percentage of GDP (Portugal, 1967-2010) Source: CHELEM database and own calculations Figure 8: Coverage of imports by exports (Portugal, 1967-2010) Source: CHELEM database and own calculations -18% -16% -14% -12% -10% -8% -6% -4% -2% 0% 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 25 The changes operated with respect to international trade flows and overall patterns of Portuguese economic growth may be better understood by resorting to the temporal delimitation developed by Lopes (1996), who distinguishes between three phases in the 1960-1992 period. A first phase, between 1960 and 1973, is marked by a strong increase in both imports and exports. This is a period characterized by significant trade liberalization following the joining to the European Free Trade Association (EFTA) in 1960, and the Free Trade Agreement established with the European Economic Community (EEC) in 1972. The second phase, between 1973 and 1985, is characterized by a more irregular growth of both imports and exports, as depicted in Figure 5. Imports decreased sharply after 1974, when the country made its transition to democracy, following the Carnation Revolution, and the modest recovery in 1976 and 1977 was interrupted in 1978, as a consequence of the adjustment policies defined under the first IMF agreement. In the subsequent years there was another spurt of imports, which was again slowed down with the second IMF agreement. Exports, on the other hand, suffered a sharp decrease after the revolutionary period, due to the combined influence of the international crisis and of the huge internal turmoil that characterized the post-revolutionary period. This pattern was reversed after 1976, when the Portuguese domestic currency (escudo) was depreciated and Portuguese firms started regaining confidence from clients abroad, but exports declined again in 1980, with the appreciation of the currency and the international recession stemming from the second oil price shock. This situation was reversed with the real depreciation of the escudo resulting from the second IMF agreement. The third phase, comprising the period between 1985 and 1992, is marked by a strong expansion of international trade, following Portugal’s accession to the European Economic Community in 1986. The removal of trade barriers, together with the decrease in oil prices and the increase in terms of trade, has considerably stimulated domestic demand, which boosted imports. Portuguese export capacity was also strengthened due to the growing of export-oriented foreign direct investment and to the modernization of Portuguese firms, often supported by PEDIP programmes (Lopes, 1996). 26 Giving continuity to Lopes’ periodization, a fourth phase can be established, based on the more recent evolution of the Portuguese economy. The last decade of the 20th century is marked by the adoption of exchange rate stabilization and restrictive monetary policies in order to reduce inflation, which gave rise to an increase in nominal wages in Portugal higher than in the EU (Lopes, 2004). The increased labour costs were not fully compensated by productivity increases or by exchange rate depreciation, as happened between 1977 and 1990, which resulted in a considerable loss of competitiveness. This point is stressed by Mateus (2006), who shows that Portuguese relative labour costs increased approximately 12% relative to EU-15 countries between 1986 and 2002. The last decade of the 20th century is also marked by a faster increase of domestic demand than industrial production. Imports of industrial and agricultural goods have increased without a corresponding movement in the export side being made, and economic growth has been mainly based on services and construction activities (Lopes, 2004). Moreover, the increase in exports was essentially produced until the eighties, remaining relatively stagnant afterwards (Figure 4). From the late 1980s onwards, the exports to GDP ratio has practically stalled, increasing slightly between 2004 and 2007, but decreasing again in 2008 and, most noticeably, in 2009. In fact, the export shares’ of the last decade are very similar to those of the years following Portugal’s accession to the EU (1986-1990), which seems to reflect some competitiveness problems of the Portuguese economy, as illustrated by the evolution of the real effective exchange rate (Figure 6). The more recent years of the period under study are also marked by a negative impact of the international crisis started in 2007-2008, which has affected both Portuguese exports and imports (Leite, 2010). Moreover, the trade balance has worsened as a result of the evolution of the energy account, whose deficit increased from 2,6% in 2003 to 4,9% in 2008 (Freitas et al., 2009). As a matter of fact, Portugal is one of the most strongly dependent countries within the EU on energy imports (mainly oil and natural gas), which makes the economy extremely vulnerable to fluctuations in their prices (Leite, 2010). 27 2.2. The structure of Portuguese exports, 1967-2010 2.2.1 Changes in the direction of trade The analysis of the Portuguese export markets in the 40-year period under analysis reveals a contrasting picture between the late 1960s-early 1980s span and the subsequent period. In fact, whereas the first sub-period is marked by an increasing diversification of exports markets, as shown by the increase of the HerfindahlHirschman index until 1981, the second evidences a clear concentration trend (cf. Table 2). The tendency towards geographical concentration since the early 1980s is also corroborated by the computation of the combined share of Portugal’s four and eight larger trade partners, which at the turn of the 20th century represented about 60% and 80% of total exports, respectively. Table 2: Concentration of Portuguese international trade 1967 1972 1977 1982 1987 1992 1997 2002 2007 2010 Equivalent number (Herfindahl-Hirschman) 7.9 9.7 13.2 12.9 10.7 9.5 9.5 9.1 8.8 9.1 Top-4 partners (CHELEM countries / zones; %) 60.5 55.1 45.0 47.8 54.8 59.8 60.3 60.3 55.9 56.4 Top-8 partners (CHELEM countries / zones; %) 74.9 72.5 65.8 67.8 75.5 77.4 78.1 80.4 76.8 73.3 Source: CHELEM database and own calculations Note: The equivalent number was computed considering the countries, groups of countries and geographic zones at the highest disaggregation level provided by the CHELEM database. The overall concentration trend was accompanied by the progressive reinforcement of European countries as Portugal’s trade partners (Figure 9). Representing about 50% of Portuguese total exports in 1967-69, exports to EU-27 countries reached more than 80% at the turn of the century (Figure 10). 28 Figure 9: Portuguese exports by receiving continent (%) Source: CHELEM database and own calculations Figure 10: Portuguese export shares outside EU-27 Source: CHELEM database and own calculations The movement towards an increasing relevance of European markets has stalled, however, in the early 2000s, when exports’ growth relative to non-European countries became higher than for European counterparts. In 2008, non-European markets represented almost 29% of total exports, about 15 p.p. above the corresponding figure in 1999. “African LDCs” have inclusively surpassed the UK’s share, becoming the fourth largest destination of Portuguese exports.21 The 2008 share of EU-27 countries in 21 The “African LDCs” zone considered in CHELEM includes the following countries: Angola, Benin, Burkina Faso, Burundi, Cameroon, Cape Verde, Central African Republic, Chad, Comoros, Cote d'Ivoire, Democratic Republic of Congo (formerly Zaire), Djibouti, Equatorial Guinea, Eritrea, Ethiopia, Gambia, Guinea, Guinea-Bissau, Kenya, Liberia, Madagascar, Malawi, Mali, Mauritania, Mozambique, Niger, Rwanda, Sao Tome and Principe, Senegal, Sierra Leone, Somalia, Sudan, Tanzania, Togo, Uganda and Zambia. Among these countries, which accounted for approximately 7% of Portuguese exports in 2010, Angola holds the highest share, being the destination of 5.2% of Portuguese exports in 2010 (ITC database). This notwithstanding, these values decreased from those registered in 2009: almost 9% and 7.2%, respectively. 0% 20% 40% 60% 80% 100% 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 Europe Africa America Asia/Oceania 0% 10% 20% 30% 40% 50% 60% 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 29 Portuguese exports was the smallest since 1981. In 2009 and 2010, exports growth directed to European countries was once again higher, with the consolidation of France as Portugal’s second largest destination market, but it is unlikely that this tendency will endure, given the low growth prospects of the main European trade partners. Within the European market, Spain is nowadays the main destination of Portuguese exports, a position accomplished in the late 1990s and which has been reinforced overtime (Figure 11).22 The decline of the Portuguese import shares in Germany (which was Portugal’s top export market between 1989 and 1999) seems to be related to the reforms taking place in Eastern Europe countries since 1991 and their application to the EU (Amaral, 2006). With respect to the origins of Portuguese imports, the most important countries are Spain, Germany and France. Other countries, such as China, South Korea and some Central and Eastern European countries (e.g., the Czech Republic and Poland) have increased their shares in total Portuguese non-energy imports. Conversely, more developed countries, such as the UK, the USA, Japan and Switzerland, have all become less important markets. The “African LDCs”, which held shares of approximately 15% on the late 60s, decreased considerably their relevance as import markets, particularly after the mid-seventies, when the decolonization process took place, representing less than 1% of Portuguese non-energy imports in the last two decades. Figure 11: Main destinations of Portuguese exports (%) Source: CHELEM database and own calculations 22 Exports to Spain suffered a major boost with the accession of both countries to the EEC in 1986, with its share in total exports increasing from approximately 4% in 1985 to almost 13% in 1990. This growth trend has continued over time: in 2010 Spain was the destination of more than 25% of Portuguese exports. 0% 5% 10% 15% 20% 25% 30% 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 Spain France Germany (Incl. FDR and GDR) UK 30 2.2.2. Changes in the composition of exports The structure of Portuguese exports underwent important changes since 1967. A preliminary assessment of these changes can be made taking into account the export shares by product type (primary, basic manufacturing, intermediate, equipment, consumption), as depicted in Table 3. In the late 1960s, the bulk of exports were accounted by a few products, intermediate goods, such as yarns, and articles in wood, and consumption goods, like beverages, preserved meat, fish and fruits. All these products became less relevant in exports during the period under analysis. Conversely, consumption goods experienced an important increase during the eighties and nineties, although a strong decline was observed in the more recent years. A strong increase in mixed products (including plastic, refined petroleum and leather products) and equipment goods shares (e.g., electrical apparatus) has also come into place, particularly during the nineties. Basic manufacturing, including ceramics, glass and iron steel, experienced a gradual increase in export shares until the end of the 20th century, contrasting with the declining shares of primary goods, particularly in the case of non-edible agricultural products. Table 3: Export shares by type of good (%) Source: CHELEM database and own calculations 19671970 19711975 19761980 19811985 19861990 19911995 19962000 20012005 20062010 Primary goods 9.5 8.1 7.6 4.7 4.6 3.9 2.8 3.3 4.9 Basic manufacturing 4.6 4.6 6.1 7.2 6.5 6.4 6.0 7.2 8.7 Intermediate goods 34.7 33.4 31.0 28.3 25.2 22.9 23.2 26.5 25.3 Equipment goods 4.4 7.9 10.2 9.0 9.1 11.3 13.3 13.4 12.5 Mixed products 7.3 7.6 9.1 14.3 16.5 18.2 15.8 15.4 17.6 Consumption goods 32.3 33.0 34.0 34.8 37.2 36.5 38.4 31.6 24.3 N.e.s. products 7.2 5.4 2.1 1.6 0.9 0.6 0.5 2.6 6.7 37 Additional evidence regarding the relative importance of intra and inter-industry trade flows is presented in Figure 15, which plots Balassa specialization coefficients by technological content. It can be seen that in the early 2000s low-tech exports and imports were rather similar, which indicates strong intra-industry specialization. This is the only category with values above zero, meaning that exports exceeded imports, but the positive gap has been declining since the mid-eighties. Likewise, supplierdominated industries were the only category in Tidd and Bessant taxonomy in which Portugal was a net exporter in the beginning of the period (Figure 16). The Balassa specialization coefficient of this industry group has decreased, however, since the accession to the EU and, at the end of the period under analysis, exports and imports were rather equivalent, that is, there was intra-industry specialization. The Balassa coefficient for medium-low-tech sectors has also been approaching zero, though starting from negative values, and again suggesting the presence of intra-industry specialization. On the contrary, the clear negative values for medium-high and high-technology sectors in the OECD (2002) taxonomy, and in Tidd and Bessant’s (2009) scale-intensive, science-based and specialized supplier categories, point to the weak competitive position of Portugal in these industries. The evolution of the coefficient towards the end of the period suggests a proximate level of inter and intra-industry trade; in the case of high-tech industries, however, inter-industry trade still prevails. Figure 15: Balassa specialization coefficient by technological content Source: CHELEM database and own calculations -1,00 -0,75 -0,50 -0,25 0,00 0,25 0,50 0,75 1,00 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 High-Tech Medium High-Tech Medium Low-Tech Low-Tech 38 Figure 16: Balassa specialization coefficient within Tidd and Bessant’s (2009) industry groups Source: CHELEM database and own calculations The analysis of changes taking place in the composition of exports can also be made by looking at the evolution of top export industries over time. Table 7 reports the export shares of the top 10 ISIC-4 industries in four different periods: the beginning and end of the sample (1967-1969 and 2008-2010), the period of accession to the EU (1985-1987) and the period in which the introduction of the Euro was made (1999-2001).29 In the late sixties, as indicated earlier, Portuguese exports were predominantly low-tech (eight out of the ten top industries had this classification). These industries accounted for more than 50% of total exports, with important contributions being made by “Prepared textile fibres; fabrics” and “Wines”. Despite the substantial revealed comparative advantage of this latter industry, in which Portugal still maintains a strong inter-industry specialization, its export share suffered a strong decline over time. The second period (1985-1987) is still marked by the prevalence of low-technology industries – the top 8 low-tech industries still represent more than a half of total exports – but in this case “Wearing apparel except fur” and “Footwear”, are the most important export industries. At the turn of the century, a number of industries with higher technological content appear in the rank, such as “Motor vehicles”,30 “TV & radio receivers, recorders” or “Other electric equipment”. 29 The Euro was introduced in 1999, but coins and banknotes only entered circulation three years later. 30 Despite its presence in the previous period, it now becomes the top industry, contributing for almost 12% of total exports. -1,00 -0,75 -0,50 -0,25 0,00 0,25 0,50 0,75 1,00 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 Supplier-dominated Scale-intensive Science-based Specialised supplier 39 Since the late 1980s there has been an overall decrease in low-tech export shares. However, more than two thirds of the top 10 products are still low or medium-low-tech, which is indicative of the fairly low technological content of Portuguese exports. The “Motor vehicles” industry is still at the top, but its competitive position has been relatively weakened, as shown by the decrease in the Balassa specialization coefficient. “Refined petroleum products” appears as the second more important industry, representing more than 5% of total exports. A relevant contribution is also made by “Pulp, paper and paperboard products”, which presents a strong revealed comparative advantage, and in which there is a proximate level of inter and intra-industry trade. Wearing and footwear also remain as important export industries, but their revealed comparative advantage has decreased over time. The analysis of top export industries also corroborates the trend of increasing diversity in Portuguese exports. The top 10 export industries’ shares have been continuously decreasing over time: starting from almost 60% of total exports, this value declined to 56%, 51% and 37% in the subsequent periods under analysis. This point is further explored in the following section, in which a formal computation of variety and its decomposition between related and unrelated variety is undertaken. 40 Table 7: Top 10 ISIC-4 industries in total exports Source: CHELEM database and own calculations 1967-1969 1985-1987 1999-2001 2008-2010 Industry Share in Exports (%) RCA Balassa Industry Share in Exports (%) RCA Balassa Industry Share in Exports (%) RCA Balassa Industry Share in Exports (%) RCA Balassa Prepared textile fibres; fabrics LT SD 13.3 8.0 0.7 Wearing apparel, except fur LT SD 15.1 30.8 0.9 Motor vehicles MHT SI 11.9 1.1 -0.2 Motor vehicles MHT SI 5.8 0.8 -0.3 Wines LT SI 8.2 664.0 1.0 Footwear LT SD 7.6 23.3 0.9 Wearing apparel, except fur LT SD 7.5 3.8 0.4 Refined petroleum products MLT SI 5.1 1.7 0.0 Wearing apparel, except fur LT SD 7.2 21.7 0.9 Pulp, paper and paperboard LT SD 6.2 5.4 0.6 Footwear LT SD 6.8 7.6 0.6 Parts for motor vehicles MHT SI 4.9 1.7 0.0 Preserved fish & fish products LT SI 6.3 3.4 0.4 Knitted fabrics & articles LT SD 5.3 13.0 0.8 Knitted fabrics & articles LT SD 4.7 4.0 0.4 Footwear LT SD 3.6 4.0 0.4 Jewellery and related articles MLT SD 6.2 1.4 0.0 Prepared textile fibres; fabrics LT SD 4.6 1.1 -0.1 Pulp, paper and paperboard LT SD 4.1 3.1 0.3 Wearing apparel, except fur LT SD 3.6 1.7 0.0 Other products of wood, cork, straw LT SD 5.5 155.6 1.0 Made-up textile articles except apparels LT SD 4.6 62.9 1.0 TV & radio receivers, recorders HT SB 3.5 2.2 0.1 Pulp, paper and paperboard LT SD 3.3 2.8 0.3 Preserved fruit and vegetables LT SI 3.8 52.6 0.9 Other products of wood, cork, straw LT SD 3.5 30.9 0.9 Other products of wood, cork, straw LT SD 3.3 15.5 0.8 Knitted fabrics & articles LT SD 2.8 2.9 0.3 Pulp, paper and paperboard LT SD 3.7 3.3 0.4 Wines LT SI 3.4 270.4 1.0 Made-up textile articles except apparels LT SD 3.3 16.3 0.8 Manufactured basic iron and steel MLT SI 2.7 0.8 -0.4 Made-up textile articles except apparels LT SD 2.8 15.3 0.8 Motor vehicles MHT SI 3.3 0.6 -0.4 Other electrical equipment MHT SB 2.9 3.7 0.4 Plastics products MLT SS 2.6 1.5 0.0 Basic chemicals, excluding fertilizers MHT SB 2.3 0.9 -0.2 Basic chemicals, excluding fertilizers MHT SB 2.8 0.7 -0.3 Parts for motor vehicles MHT SI 2.7 0.8 -0.3 TV & radio receivers, recorders HT SB 2.5 1.9 0.1 41 2.3. Trends in export variety Following a common procedure in the literature (cf. Frenken et al., 2007; Saviotti and Frenken, 2008; Boschma and Iammarino, 2009; Boschma et al., 2012), we use entropy measures to proxy the degree of export variety.31 The entropy coefficient (H) refers to the expected information content or uncertainty of a probability distribution (Frenken, 2007), being calculated by the following expression: (1) where is the number of sectors of composing the economic system and stands for the share of sector in total exports. An important advantage stemming from the use of the entropy coefficient is that it can be decomposed at each sectoral level, which avoids collinearity problems (Jacquemin and Berry, 1979). The minimum value of the entropy index (0) represents total specialization, in which exports are totally concentrated in one sector ( , ; , )., whereas higher values of this index indicate greater relative diversification. The highest value corresponds to the situation of equal shares ( ). Unrelated variety, which accounts for variety between sectors, is measured at higher levels of aggregation. Export shares at these aggregation levels ( ) are obtained by summing the shares at lower levels of aggregation ( ), where Sg stands for a sector at a higher level of aggregation: (2) 31 The use of entropy measures derives from the selected theoretical framework, based on neoSchumpeterian streams of research. As indicated earlier, studies more in line with mainstream economics measure export variety using an indicator developed by Feenstra (1994). 42 Thus, a measure of unrelated variety (UV), or between-group entropy, is given as follows: (3) Related variety (RV), which accounts for variety within sectors, is computed as the weighted average of the entropy values within groups ( ): (4) where (5) Unrelated variety is computed at the two-digit sectoral level, whereas related variety is computed as the weighted sum of the entropy at the four-digit level within each twodigit category. Since no mutual information exists between related and unrelated variety, that is, the two dimensions do not tend to co-occur, total entropy equals the sum of related and unrelated variety:32 (6) Figure 17 depicts the results obtained from the computation of the entropy coefficient. In line with the analysis performed in the previous section, it can be seen that Portuguese export variety has increased since 1967, although this increase has been essentially produced during the last two decades. Total entropy increased in the seventies but experienced a decline in the following decade, reaching a trough in 1987. From this period onwards, there has been a systematic increase in export variety, which reaches its maximum in 2010. 32 See Theil (1972) and Frenken (2007) for more details on the properties of entropy indices. 43 Figure 17: Total export variety (Portugal, 1967-2010) Source: CHELEM database and own calculations Distinguishing between related and unrelated variety, it can be seen furthermore that they behave quite differently over time (cf. Figures 18 and 19). More precisely, whereas unrelated variety shows an upward trend during the whole period under study, related variety decreases from the beginning of the sample until the late 80s, increasing afterwards. Export variety within sectors increases only in the last two decades, reaching in 2010 values slightly above those registered in the beginning of the period. Figure 18: Related variety (Portugal, 1967-2010) Source: CHELEM database and own calculations 4,60 4,80 5,00 5,20 5,40 5,60 5,80 6,00 6,20 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 1,00 1,10 1,20 1,30 1,40 1,50 1,60 1,70 1,80 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 44 Figure 19: Unrelated variety (Portugal, 1967-2010) Source: CHELEM database and own calculations Recalling the theoretical arguments outlined in Part 1, knowledge spillovers are more likely to occur between cognitively proximate firms (Nooteboom, 1999), stimulating productivity growth, whereas the impact of unrelated variety is essentially produced over employment. Since we are interested in analysing the impact of variety on economic growth, the decomposition of variety in the aforementioned technological and innovative categories is restricted to the related variety part. Related variety among low-tech industries has decreased markedly since the beginning of the period (Figure 20), although a slight recovery took place over the last years under analysis. A decreasing trend is also found since the late 1970s in Tidd and Bessant (2009) least innovative category, supplier-dominated industries (Figure 21). Related variety in scale-intensive sectors declined markedly between 1967 and 1990 but it increased afterwards, reaching figures similar to those observed at the beginning of the period. Medium-low and medium-high-tech industries, on the other hand, show an upward trend in related variety, which is particularly strong in the case of the latter category. Tidd and Bessant (2009) categories with higher innovative potential also show an increase in variety since the beginning of the period, which was stronger in the case of specialized supplier industries, despite the decrease that took place in the more recent years. 3,00 3,20 3,40 3,60 3,80 4,00 4,20 4,40 4,60 4,80 5,00 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 45 Related variety among high-tech industries suffered minor changes during the period under study, showing a slight tendency of increase from the mid-1990s until 2006, but declining afterwards. The more recent years present figures similar to the ones registered during the seventies and early eighties. An analysis of the decomposition of entropy in the end of the period reveals that lowtech and medium-low-tech industries still account for the more than a half of related variety, despite the fact that the highest value in individual terms is now registered in medium-high-tech sectors. A similar conclusion is drawn for Tidd and Bessant (2009) taxonomy, with supplier-dominated and the scale-intensive industries, the lowest categories in technological and innovative potential, representing approximately 60 per cent of export related variety. Finally, it is also important to point out that, from 1967 to 2010, both export variety and overall exports have increased more in the Portuguese case than in the EU-15, which suggests that, according to Saviotti (2003), a creative de-specialization path has been in place. In other words, Portuguese exports, which were predominantly based on low-tech industries, show a faster pace in more technologically advanced sectors than the EU15.33 Figure 20: Related variety by technological content (Portugal, 1967-2010) Source: CHELEM database and own calculations 33 This evolution could also be expected due to the smaller initial values observed for Portugal. 0,00 0,20 0,40 0,60 0,80 1,00 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 High-Tech Medium High-Tech Medium Low-Tech Low-Tech 46 Figure 21: Related variety using Tidd and Bessant’s (2009) industry groups Source: CHELEM database and own calculations 2.4. Comparison of the Portuguese case with other cohesion countries’ experiences Being known the main features of Portuguese export trends, it seems worthwhile looking at the same evidence from a comparative perspective, taking into account the experiences of other southern European and cohesion countries, namely Spain, Ireland, Greece and Italy. Figure 22 shows that the technological content of exports of all these countries has improved significantly since 1967. In fact, high-tech exports have increased substantially their shares, whereas the opposite has occurred with low-tech ones. This notwithstanding, high-tech exports’ shares of Portugal, Spain, Greece and Italy still remain well below EU-15 levels. The major exception is Ireland, which changed profoundly its export structure between 1967 and 2010: in the late sixties more than two thirds of its exports were concentrated in low-technology sectors, whereas in 2010 the high-technology sectors represented more than 55% of total manufacturing exports (Table 8). Medium-high-technology exports have also experienced important changes in Ireland, a pattern also found in Spain and Portugal. Italy also displays an important share of exports in these sectors, with values very close to the EU-15 average, regardless of the negative evolution occurred in the mid-seventies and in the beginning of the eighties. 0,00 0,20 0,40 0,60 0,80 1,00 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 Supplier-dominated Scale-intensive Science-based Specialised supplier 53 0,00 0,20 0,40 0,60 0,80 1,00 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 Spain High-Tech Medium High-Tech Medium Low-Tech Low-Tech 0,00 0,20 0,40 0,60 0,80 1,00 1,20 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 Ireland High-Tech Medium High-Tech Medium Low-Tech Low-Tech 0,00 0,20 0,40 0,60 0,80 1,00 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 Greece High-Tech Medium High-Tech Medium Low-Tech Low-Tech 0,00 0,20 0,40 0,60 0,80 1,00 1,20 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 Italy High-Tech Medium High-Tech Medium Low-Tech Low-Tech ´ Figure 25: Related variety by technological content in Spain, Ireland, Greece and Italy Source: CHELEM database and own calculations 54 0,00 0,20 0,40 0,60 0,80 1,00 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 Spain Supplier-dominated Scale-intensive Science-based Specialised supplier 0,00 0,20 0,40 0,60 0,80 1,00 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 Ireland Supplier-dominated Scale-intensive Science-based Specialised supplier 0,00 0,20 0,40 0,60 0,80 1,00 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 Greece Supplier-dominated Scale-intensive Science-based Specialised supplier 0,00 0,20 0,40 0,60 0,80 1,00 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 Italy Supplier-dominated Scale-intensive Science-based Specialised supplier Figure 26: Related variety using Tidd and Bessant’s (2009) industry groups in Spain, Ireland, Greece and Italy Source: CHELEM database and own calculations 55 experienced an increase in the export shares of the top categories, particularly high-tech and science-based industries, suggesting that a virtuous specialization path has been in place. Summing up, several significant changes took place in the Portuguese export structure during the period under analysis. There was a broad upward movement in the technological content of exports, which was also observed in the other countries, especially in the case of Ireland. This notwithstanding, Portuguese high-tech exports remain well below the EU-15 average, while low-tech industries still maintain important export shares, representing nearly one third of total manufacturing exports. Along with this trend, the scale-intensive export shares surpassed supplier dominated shares, a trend that was also observed in Spain, Italy and Greece, whereas in Ireland science-based industries stand out noticeably. The broad upward movement in the technological content of exports should be read with some caution, however. On the one hand, an important part of the growth observed took place in “TV & radio receivers, recorders” industries, which are classified as “science-based”, but which in many cases do not promote significant R&D activities, being mostly based on assembly-line production, intensive in unskilled/low-cost labour.34 On the other hand, although there was a global positive evolution, Portugal did not change considerably its position relative to the other South-European countries, and has enlarged considerably a (negative) distance relative to Ireland. In general, changes observed in Portugal bear some resemblances with those of Spain, Greece and Italy, whereas the evolution of the Irish export structure presents several (and much more positive) distinctive features. The computation of the entropy coefficient shows an increase in Portuguese export variety since 1967, essentially during the last two decades. Although both related and unrelated variety display an upward trend in the end of the period, they behaved quite differently over time, with related variety decreasing until 1989. Decomposing related variety, it can be seen furthermore that the major share of variety within sectors occurs in lower technological and innovative potential categories, although recently mediumhigh-tech industries have become dominant. Export variety evolved in different ways 34 See, in this respect, Hobday (1995). 56 across countries, with Ireland experiencing a decline in total variety, combined with a significant progress in related variety among high-tech industries. On the whole, the positive evolution observed in related variety in all countries is associated with the strong increase observed within the top technological and innovative categories, particularly in the case of Ireland. 57 3. Export variety and economic growth: an econometric assessment 3.1. The model The general econometric specification used in the estimations performed to assess the impact of export variety on Portuguese economic growth is defined in Equation 7. (7) In this expression, is the natural logarithm of labour productivity, defined as GDP per hour worked in period t, UV and RV are the main explanatory variables, representing, respectively, the unrelated and related variety components, CV is a vector of control variables which may influence productivity growth, and is the error term. Related and unrelated varieties are computed in the ways described in Equations 1-5 in Part 2. Since we are interested in crossing the variety and technology dimensions, we also estimate Equation 7 using the decomposition of related variety according to the innovation and technology industry categories defined earlier. More precisely, we decompose related variety into the high-tech (RVHT), medium-high-tech (RVMHT) and medium-low-tech and low-tech (RVMLTLT) categories, and into specialized supplier (RVSS), science-based (RVSB) and supplier-dominated and scale-intensive (RVSDSI) categories. These latter specifications allow us to investigate if the impact of related variety on productivity growth differs across the different technology groups, providing in this way a better grasp on the relationship between variety, technology and growth. Following the theoretical arguments outlined in Part 1, we expect a positive relationship between related variety and labour productivity growth. If a country specializes in a particular composition of complementary sectors, knowledge spillovers will be more likely to occur between them, and the country will probably benefit from higher growth rates. The impact of unrelated variety on productivity growth is less clear-cut, however. Unrelated variety plays an important role in employment, dampening the effects of sector-specific shocks on unemployment growth, but its impact on productivity growth 58 is not readily apparent, since knowledge spillovers are more likely to occur when firms are cognitively proximate (Nooteboom, 1999).35 With respect to the technology classification of exports, we expect knowledge spillovers to be more relevant in high-tech industries, since R&D intensive firms are more capable to absorb and exploit existing information (O’Mahony and Vecchi, 2009; Heindenreich, 2009; Santamaría et al., 2009). Consequently, we expect a positive impact on productivity growth from technology-intensive sectors (OECD taxonomy) and from science-based and specialized supplier industries, the Tidd and Bessant’s (2009) categories with higher innovative potential. In contrast, in the least innovative categories the relationship is expected to be more moderate, particularly in supplierdominated industries. Actually, this category corresponds to a great extent to low-tech industries, such as textiles, wearing apparel or wood and wood products, among others.36 Regarding the control variables, we include a proxy of the human capital stock (HC), defined as the average number of years of formal education of the working age population. The importance of human capital on growth is reflected on the fundamental role played by the so-called “social capabilities”, which determine the country’s capacity to assimilate more advanced technologies from other economies (Abramovitz, 1986).37 Moreover, according to the non-linear model of convergence developed by Verspagen (1991), countries with larger technological backwardness and lower levels of intrinsic learning capability – which, among other variables, depends on the education of the labour force – are more likely to widen their development gaps. In empirical terms, although results are not unanimous in this respect, a vast number of studies have successfully established a (positive) relationship between human capital and economic growth (e.g., Temple, 1999; Agiomirgianakis et al., 2002; Ciccone and Papaioannou, 2009). 35 It is probably for this reason that previous evidence regarding the impact of unrelated variety on productivity growth is rather mixed (cf. Part 1, Section 1.3). 36 As pointed out by Pavitt (1984, p. 356), “supplier dominated firms can be found mainly in traditional sectors of manufacturing” and they “make only a minor contribution to their process or product technology”. Moreover, according to the classification of industries presented in the Appendix, there is a close correspondence between these two categories. 37 The term “social capabilities” was originally introduced by Okawa and Rosovsky (1973, p. 212), “to designate those factors constituting a country’s ability to import or engage in technological and organizational progress”. 59 To account for the influence of physical capital accumulation, the share of investment in GDP (INV) is also included in the regression. Equipment investment may translate into high social returns, as shown by the central role played by mechanization in the economic history of countries and by the external economies generated by equipment investment (De Long and Summers, 1991; Herrerias and Orts, 2012). Data on labour productivity, expressed in 1990 US dollars converted at Geary Khamis PPPs, are taken from The Conference Board Total Economy database, available on-line at http://www.conference-board.org/data/economydatabase/. Data on export flows are taken from the CHELEM database, which provides detailed information regarding export and import flows, both at the macroeconomic and industry levels of analysis, over a rather long time span (from 1967 to 2010). With regard to the control variables, data on education are taken from Bassanini and Scarpetta (2001) for the period between 1971 and 1998, and from Silva and Teixeira (2011) for the period between 1999 and 2003. We extend the computations from these latter authors up to 2009, applying the same methodology and using data from OECD Education at a Glance (several issues). Furthermore, we extrapolate these values, considering the annual average growth rate from 1971 to 2009, to obtain data for the years 1967-1970 and 2010. Data on gross fixed capital formation are taken from the PORDATA database, available on-line at http://www.pordata.pt/. 3.2. Estimation method The variables used in the regression display strong trends, as depicted in Figures 18-21 and 27-29, evolving over time and showing no tendency to revert to their mean levels. In other words, they are non-stationary. 60 Figure 27: Portuguese labour productivity per hour worked in 1990 US dollars (converted at Geary Khamis PPPs and expressed in natural logarithms) Source: The Conference Board Total Economy Database Figure 28: Natural logarithm of the average number of years of formal education of the working age population (Portugal, 1967-2010) Sources: Bassanini and Scarpetta (2001), Silva and Teixeira (2011), OECD Education at a Glance (several issues) and own calculations 0,00 0,50 1,00 1,50 2,00 2,50 3,00 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 1,60 1,70 1,80 1,90 2,00 2,10 2,20 2,30 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 61 Figure 29: Natural logarithm of the share of investment in GDP (Portugal, 1967-2010) Sources: PORDATA database and own calculations Given the non-stationarity of the variables, resorting to classical estimation techniques, such as ordinary least squares (OLS), could lead to spurious regressions (Granger and Newbold, 1974). In fact, if the means and variances change over time, the computed statistics of a regression model will be also dependent on time, and consequently, they will not converge to the population values as the sample increases to infinity. Furthermore, hypothesis testing will be biased towards the rejection of the null hypothesis (Rao, 1994). The use of cointegration techniques is thus required in order to get reliable estimates (e.g. Granger, 1981; Engle and Granger, 1987). Two or more variables are cointegrated if, albeit being individually non-stationary, one or more linear combinations of them are stationary, becoming stable around a fixed mean in the longrun (Dickey et al., 1991). In order to obtain a cointegration relationship between a specific group of variables, the variables must be integrated of the same order. We thus start by performing the Augmented Dickey-Fuller (ADF) (Dickey and Fuller, 1979; 1981) and the PhillipsPerron (PP) (Phillips and Perron, 1988) unit root tests in order to assess the stationarity of the variables under study.38 The results are presented in Tables 9 and 10. 38 The econometric software EViews 7 was used in the estimation. -1,80 -1,60 -1,40 -1,20 -1,00 -0,80 -0,60 -0,40 -0,20 0,00 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 62 Table 9: Unit root tests – variables in levels Series ADF PP Y 0.1915 (0) 0.0930 (21) UV 0.2319 (0) 0.1908 (5) RV 0.7946 (0) 0.8722 (3) RVHT 0.0843 (0) 0.0843 (0) RVMHT 0.2652 (0) 0.2769 (1) RVMLTLT 0.6148 (0) 0.5772 (1) RVSDSI 0.5189 (0) 0.5657 (3) RVSB 0.1040 (0) 0.1209 (2) RVSS 0.5133 (0) 0.5133 (0) HC 1.0000 (0) 1.0000 (4) INV 0.2737 (0) 0.4093 (5) Notes: For the Augmented Dickey-Fuller (ADF) and Phillips-Perron (PP) tests, we present MacKinnon (1996) one-sided p-values. For series Y, UV, RV, RVMHT, RVMLTLT, RVSDSI, RVSB, RVSS and INV we specify a random walk with drift and time trend, while for series RVHT we use a random walk with drift and for series HC we use a random walk. For the ADF test we use the Schwarz Information Criterion, with an upper bound of 9 lags (figures enclosed in parentheses in the ADF column are the lag length). For the PP test, bandwidth selection was made according to the Newey-West (1994) method, using Bartlett kernel (figures enclosed in parentheses in the PP column represent the Newey-West bandwidth). Table 10: Unit root tests – variables in first differences Series ADF PP Y 0.0001 (0) 0.0000 (41) UV 0.0000 (0) 0.0000 (11) RV 0.0000 (0) 0.0000 (4) RVHT 0.0000 (0) 0.0000 (3) RVMHT 0.0000 (0) 0.0000 (3) RVMLTLT 0.0000 (0) 0.0000 (2) RVSDSI 0.0001 (1) 0.0000 (6) RVSB 0.0000 (0) 0.0000 (3) RVSS 0.0000 (0) 0.0000 (4) HC 0.0000 (0) 0.0000 (8) INV 0.0002 (2) 0.0000 (3) Notes: For the Augmented Dickey-Fuller (ADF) and Phillips-Perron (PP) tests, we present MacKinnon (1996) one-sided p-values. For series Y, RV, RVMLTLT, RVSDSI and HC, we specify a random walk with drift and time trend, while for series UV, RVMHT and RVSS we use a random walk with drift and for series RVHT, RVSB and INV we use a random walk. For the ADF test we use the Schwarz Information Criterion, with an upper bound of 9 lags (figures enclosed in parentheses in the ADF column are the lag length). For the PP test, bandwidth selection was made according to the Newey-West (1994) method, using Bartlett kernel (figures enclosed in parentheses in the PP column represent the Newey-West bandwidth). 69 Tidd and Bessant (2009). These taxonomies allow for the characterization of sectors according to their technological content, providing in these terms a deeper understanding of the relationship between export variety, technology and economic growth. During the period under analysis, the Portuguese economy underwent considerable change. International trade flows, in particular, experienced strong transformation, acting simultaneously as recipients and drivers of macroeconomic change. Both imports and exports increased substantially, particularly the former, with a considerable imbalance between them being found in virtually the whole period under study. The trade deficit has not declined with the accession to the EU, and export shares did not increase significantly from that period onwards. With respect to Portuguese exports’ destinations, there was an overall concentration trend since the eighties, which was accompanied by the reinforcement of European countries’ shares. The composition of exports has also changed significantly, with a decreasing trend in primary products and “traditional export sectors”, such as textiles and wearing apparel, being observed since the late 1980s. Conversely, automobiles and vehicle components increased considerably their shares, being today top export industries. Overall, there has been a considerable reduction in the export shares of low technology and innovative potential industries, which also suffered a decline in revealed comparative advantage, whereas mediumhigh-tech and scale-intensive industries’ export shares increased markedly. This pattern was also found in other southern European and cohesion countries, such as Spain and Greece. According to our findings, Portuguese total export variety, measured by the entropy coefficient, increased markedly in the last two decades. Notorious differences arise, however, with respect to the evolution of the related and unrelated variety components. Whereas unrelated variety displays a positive trend over the whole period under analysis, related variety decreased from the beginning of the sample until the late 80s, exhibiting a positive trend from that time onwards. Decomposing related variety according to the technology and innovation taxonomies, it can be seen furthermore that the largest part of variety is accounted by low-tech and medium-low-tech industries, and supplier-dominated and scale-intensive categories, although there is a marked tendency of increase in variety in the top technology and innovation industry categories. 70 The investigation of the impact of export variety on productivity growth was performed with resort to cointegration techniques, due to the non-stationarity of the series. The results show a negative relationship between (broad) export variety and labour productivity growth, which suggests that, contrarily to the theoretical arguments put forward, increasing levels of export variety had a negative impact on Portuguese economic growth. The consideration of a more accurate notion of variety, which crosses the technology and variety dimensions, permits to solve the puzzle, by showing that the effect of increasing variety on productivity growth is conditioned by the technological content and innovative potential of industries. More precisely, an increase in export related variety of technology and innovation intensive industries is positively related to an increase in productivity growth, whereas the opposite stands for low-tech, lowinnovation sectors. These findings suggest that the diversification of the export structure matters for growth, but only when it takes place in the related high-tech and innovative intensive segments of the economy. This in turn requires strong investment in education and industrial-led policies. Taken as a whole, our results point to a number of policy recommendations. According to the third Community Innovation Statistics (CIS) survey, the share of higher education employees in Portuguese manufacturing is one of the lowest among developed countries (Tamura et al., 2005), which is reflected in low levels of absorptive capacity of Portuguese firms (Teixeira and Fortuna, 2010). Since the diversification of the export structure is particularly relevant in the top technological and innovative industries, it seems to be of decisive importance to direct industrial policies towards the development of technological capabilities and human capital improvement. The specific design of these policies seems also to be of the utmost importance. Supplyside policies, involving the promotion of the technological infrastructure, should be accompanied by demand-side measures, including the promotion of the mobility of young researchers, the improvement of the career prospects of public researchers, the provision of better information on employment opportunities in the enterprise sector to students, or the promotion of business R&D (OECD, 2004; Laranja, 2009). Moreover, as the number of small and medium-sized enterprises (SMEs) account for 99.6% of all firms in Portugal in 2005 (IAPMEI, 2008), they should be given special 71 attention. Empirical evidence from Portuguese firms (Nunes et al., 2012) suggests that there are different growth patterns between high-tech and non-high-tech SMEs, with the former experiencing higher average growth rates than the latter. Since high-tech SMEs face greater difficulties to finance growth opportunities through external finance (Nunes et al., 2012), economic policy should support them, dampening their financial restrictions, so that they can boost Portuguese exports in the top technological categories. Finally, in order to increase related variety in the top technology and innovation categories, policy makers should also work on the attraction of FDI projects, which can prove to be very useful in the introduction and absorption of advanced technologies (Liu and Wang, 2003; Sinani and Meyer, 2004). The analysis performed provides rather clear results regarding the inter-relatedness features of export variety, technology and growth. Still, it may be substantiated or extended in a number of ways. First, alternative measures can be used to proxy the degree of relatedness between industries, besides the hierarchy of the International Standard Industrial Classification. Among others, such measures can rely on clusters (Porter, 1998), export profiles (Hidalgo et al., 2007), production knowledge (Bryce and Winter, 2009) or skills, captured by labour flows between industries (Neffke and Henning, 2009). Second, in order to overcome the heterogeneity that exists within each category, product or even between firms, information could be obtained directly from firm data, or using classifications that take into account the Portuguese specificities, which would allow for a more accurate classification of the technological content of exports. Third, the analysis of the impact of export variety and its technological content on economic growth could be replicated in other countries for comparative purposes; in particular, it would be enlightening to see if the results obtained with respect to the Portuguese economy were also reproduced in other cohesion and southern European countries, given their (cultural and economic) proximity to the Portuguese case. 72 References Abramovitz, M. (1986), “Catching up, forging ahead, and falling behind”, The Journal of Economic History, 46 (2): 385-406. Addison, D. (2003), “Productivity growth and product variety: gains from imitation and education”, Policy Research Working Paper, Number 3023, The World Bank. Afonso, O. (1999), Contributo do Comércio Externo para o Crescimento Económico Português, 1960-1993, Lisboa, Portugal: Conselho Económico e Social. Afonso, O.; Aguiar, A. (2005), “A internacionalização da economia”, in Lains, P.; Silva, A. F., (eds.), História Económica de Portugal 1700-2000. O Século XX, Lisboa, Portugal: Instituto de Ciências Sociais da Universidade de Lisboa. Agiomirgianakis, G.; Asteriou, D.; Monastiriotis, V. (2002), “Human capital and economic growth revisited: a dynamic panel data study”, International Advances in Economic Research, 8 (3): 177–187. Amador, J.; Cabral, S.; Maria, J.R. (2007), “International trade patterns over the last four decades: how does Portugal compare with other cohesion countries?”, Working Papers, 14/2007, Banco de Portugal. Amaral, J.F. (2006), “Evolução do comércio externo português de exportação (19952004)”, GEE Papers, No. 1, Gabinete de Estratégia e Estudos, Ministério da Economia e da Inovação. Bassanini, A.; Scarpetta, S. (2001), “Does human capital matter for growth in OECD countries?: Evidence from pooled mean-group estimates”, OECD Economics Department Working Papers, No. 282, OECD Publishing. Bloom, N.; Draca, M.; Reenen, J.V. (2011), “Trade induced technical change? The impact of Chinese imports on innovation, IT and productivity”, NBER Working Paper Series, No. 16717, The National Bureau of Economic Research. Boschma, R.; Iammarino, S. (2009), “Related variety, trade linkages, and regional growth in Italy”, Economic Geography, 85 (3): 289-311. 73 Boschma, R.; Minondo, R.; Navarro, M. (2012), “Related variety and regional growth in Spain”, Papers in Regional Science, 91 (2): 241-257. Broda, C.; Weinstein, D. (2006), “Globalization and the gains from variety”, The Quarterly Journal of Economics, 121 (2): 541-585. Bryce, D.; Winter, S. (2009), “A general interindustry relatedness index”, Management Science, 55 (9): 1570-1585. Cabral, M.H.C. (2008), “Export diversification and technological improvement: recent trends in the Portuguese economy”, GEE Papers, No. 6, Gabinete de Estratégia e Estudos, Ministério da Economia e da Inovação. Cabral, S.; Manteu, C. (2010), “Ganhos da importação de novas variedades: o caso de Portugal”, Boletim Económico, Verão, Banco de Portugal. Chamberlin, E.H. (1933[1950]), The Theory of Monopolistic Competition: A Reorientation of the Theory of Value, 6th ed., Cambridge, USA: Harvard University Press. Ciccone, A.; Papaioannou, E. (2009), “Human capital, the structure of production, and growth”, The Review of Economics and Statistics, 91 (1): 66-82. De Long, J.B.; Summers, L.H. (1991), “Equipment investment and economic growth”, The Quarterly Journal of Economics, 106 (2): 445-502. Dibooglu, S.; Enders, W. (1995), “Multiple cointegrating vectors and structural economic models: an application to the French Franc/U.S. Dollar exchange rate, Southern Economic Journal, 61 (4): 1098–1116. Dickey, D.; Fuller, W. (1979), “Distribution of the estimates for autoregressive time series with a unit root”, Journal of the American Statistical Association, 74 (366): 427–431. Dickey, D.; Fuller, W. (1981), “Likelihood ratio statistics for autoregressive time series with a unit root”, Econometrica, 49 (4): 1057–1072. Dickey, D.; Jansen, D.; Thornton, D. (1991), “A primer on cointegration with an application to money and income”, in Rao, B. (ed.), Cointegration: for the Applied Economist, Basingstoke, UK: Palgrave Macmillan. 74 Engle, R.; Granger, C. (1987), “Co-integration and error correction: representation, estimation, and testing”, Econometrica, 55 (2): 251-276. Fagerberg, J. (2000), “Technological progress, structural change and productivity growth: a comparative study”, Structural Change and Economic Dynamics, 11: 393-411. Feenstra, R. (1994), “New product varieties and the measurement of international prices”, The American Economic Review, 84 (1): 157-177. Feenstra, R.; Kee, H. (2008), “Export variety and country productivity: estimating the monopolistic competition model with endogenous productivity”, Journal of International Economics, 74: 500-518. Feenstra, R.; Madani, D.; Yang, T.; Liang, C. (1999), “Testing endogenous growth in South Korea and Taiwan”, Journal of Development Economics, 60: 317-341. Freitas, M.L.; Mamede, R.P. (2011), “Structural transformation of Portuguese exports and the role of foreign-owned firms: a descriptive analysis for the period 19952005”, Notas Económicas, 33: 20-43. Freitas, M.L.; Salvado, S.; Marques, W. (2009), “O preço da energia e o défice da balança energética em Portugal”, Boletim Mensal de Economia Portuguesa, 5: 3741. Frenken, K. (2007), “Entropy statistics and information theory”, in Hanusch, H.; Pyka, A. (eds.), Elgar Companion to Neo-Schumpeterian Economics, Cheltenham, UK: Edward Elgar. Frenken, K.; Van Oort, F.; Verburg, T. (2007), “Related variety, unrelated variety and regional economic growth”, Regional Studies, 41.5: 685–697. Funke, M.; Ruhwedel, R. (2001a), “Product variety and economic growth: empirical evidence for the OECD countries”, IMF Staff papers, 48 No. 2, International Monetary Fund. Funke, M.; Ruhwedel, R. (2001b), “Export variety and export performance: empirical evidence from East Asia”, Journal of Asian Economics, 12: 493–505. 75 Funke, M.; Ruhwedel R. (2005), “Export variety and economic growth in East European transition economies”, Economics of Transition, 13 (1): 25–50. Granger, C. (1981), “Some properties of time series data and their use in econometric model specification”, Journal of Econometrics, 16: 121-130. Granger, C.; Newbold, P. (1974), “Spurious regressions in econometrics”, Journal of Econometrics, 2: 111-120. Handa, J. (2009), Monetary Economics, New York, USA: Routledge. Hartog, M.; Boschma, R.; Sotarauta, M. (2012), “The impact of related variety on regional employment growth in Finland 1993-2006: high-tech versus medium/low-tech”, Papers in Evolutionary Economic Geography, Number 12.05, Utrecht University. Hatzichronoglou, T. (1997), “Revision of the high-technology sector and product classification”, OECD Science, Technology and Industry Working Papers, 1997/02, Organisation for Economic Co-operation and Development. Hausmann, R.; Hwang, J.; Rodrik, D. (2007), “What you export matters”, Journal of Economic Growth, 12: 1-25. Heidenreich, M. (2009), “Innovation patterns and location of European lowand medium-technology industries”, Research Policy, 38: 483-494. Helpman, E. (1981), “International trade in the presence of product differentiation, economies of scale and monopolistic competition: a Chamberlin-Heckscher-Ohlin approach”, Journal of International Economics, 11: 305-340. Herrerias, M. J.; Orts, V. (2012), “Equipment investment, output and productivity in China”, Empirical Economics, 42 (1): 181-207. Hidalgo, C.A.; Klinger, B.; Barabási, A.-L.; Hausmann, R. (2007), “The product space conditions the development of nations”, Science, 317 (5837): 482-487. Hobday, M. (1995), “East Asian latecomer firms: learning the technology of electronics”, World Development, 23 (7): 1171-1193. Hodrick, R.J.; Prescott, E.C. (1997), “Postwar U.S. business cycles: an empirical investigation”, Journal of Money, Credit and Banking, 29 (1): 1-16. 76 Hotelling, H. (1929), “Stability in competition”, The Economic Journal, 39 (153): 4157. IAPMEI (2008), Sobre as PMEs em Portugal, Lisboa, Portugal: Instituto de Apoio às Pequenas e Médias Empresas e à Inovação. Jacquemin, A.; Berry, C. (1979), “Entropy measure of diversification and corporate growth”, The Journal of Industrial Economics, 27 (4): 359-369. Johansen, S. (1988), “Statistical analysis of cointegration vector”, Journal of Economic Dynamics and Control, 12: 231–254. Johansen, S. (1991), “Estimation and hypothesis testing of cointegrating vectors in Gaussian vector autoregressive models”, Econometrica, 59 (6): 1551-1580. Johansen, S.; Juselius, K. (1990), “Maximum likelihood estimation and inference on cointegration with application to the demand for money”, Oxford Bulletin of Economics and Statistics, 52 (2): 169-210. Jones, C. (1995), “R&D-based models of economic growth”, Journal of Political Economy, 103 (4): 759-784. Khandelwal, A. (2010), “The long and short (of) quality ladders”, Review of Economic Studies, 77 (4): 1450-1476. Kleinknecht, A.; Montfort, K.; Brouwer, E. (2002), “The non-trivial choice between innovation indicators”, Economics of Innovation and New Technology, 11 (2): 109-121. Kruger, J. (2008), “Productivity and structural change: a review of the literature”, Journal of Economic Surveys, 22 (2): 330-363. Krugman, P. (1979), “Increasing returns, monopolistic competition and international trade”, Journal of International Economics, 9: 469-479. Krugman, P. (1980), “Scale economies, product differentiation and the pattern of trade”, The American Economic Review, 70 (5): 950-959. Krugman, P. (1981), “Intraindustry specialization and the gains from trade”, Journal of Political Economy, 89 (5): 959-973. 77 Kuznets, S. (1971), Economic Growth of Nations: Total Output and Production Structure, Cambridge, USA: Harvard University Press. Lall, S.; Weiss, J.; Zhang, J. (2005), “The sophistication of exports: a new measure of product characteristics, QEH Working Paper Series, No. 123, Queen Elizabeth House, Oxford University. Lancaster, K. (1990), “The economics of product variety: a survey”, Marketing Science, 9 (3): 189-206. Laranja, M. (2009), “The development of technology infrastructure in Portugal and the need to pull innovation using proactive intermediation policies”, Technovation, 29: 23-34. Leite, A.N. (2010), “A internacionalização da economia portuguesa”, Relações Internacionais, 28: 119-132. Liu, X.; Wang, C. (2003), “Does foreign direct investment facilitate technological progress? Evidence from Chinese industries”, Research Policy, 32: 945-953. Lopes, J.S. (1996), A Economia Portuguesa desde 1960, Lisboa, Portugal: Gradiva. Lopes, J.S. (2004), A Economia Portuguesa no Século XX, Lisboa: Portugal: Instituto de Ciências Sociais da Universidade de Lisboa. Loschky, A. (2008), “Reviewing the nomenclature for high-technology trade - the sectoral approach”, STD/SES/WPTGS, 2008/9, Organisation for Economic Cooperation and Development. MacKinnon, J. (1996), “Numerical distribution functions for unit root and cointegration tests”, Journal of Applied Econometrics, 11 (6): 601–618. MacKinnon, J.; Haug, A.; Michelis, L. (1999), “Numerical distribution functions of likelihood ratio tests for cointegration”, Journal of Applied Econometrics, 14 (5): 563–577. Martin, J.; Méjean, I. (2011), “Low-wage countries’ competition, reallocation across firms and the quality content of exports”, CEPR Discussion Papers, No. 8231, The Centre for Economic Policy Research. 78 Mateus, A. (2006), Economia Portuguesa: Crescimento no Contexto Internacional (1910-2006), Lisboa, Portugal: Editorial Verbo. Neffke, F.; Henning, M. (2009), “Skill-relatedness and firm diversification”, Papers on Economics and Evolution, Jena, Germany: Max Planck Institute of Economics. Newey, W.; West, K. (1994), “Automatic lag selection in covariance matrix estimation”, Review of Economic Studies, 61: 631–653. Nooteboom, B. (1999), “Innovation and inter-firm linkages: new implications for policy”, Research Policy, 28 (8): 793-805. Nunes, P.M.; Serrasqueiro, Z.; Leitão, J. (2012), “Is there a linear relationship between R&D intensity and growth? Empirical evidence of non-high-tech vs. high-tech SMEs”, Research Policy, 41: 36-53. OECD (2002), The Measurement of Scientific and Technological Activities. Proposed Standard Practice for Surveys on Research and Experimental Development – Frascati Manual, Paris, France: OECD Publications. OECD (2004), OECD Science, Technology and Industry Outlook 2004, Paris, France: OECD Publications. Okawa, K.; Rosovsky, H. (1973), Japanese Economic Growth: Trend Acceleration in the Twentieth Century, Stanford, California, USA: Stanford University Press. O’Mahony, M.; Vecchi, M. (2009), “R&D, knowledge spillovers and company productivity performance”, Research Policy, 38: 35-44. Pack, H. (1994), “Endogenous growth theory: intellectual appeal and empirical shortcomings”, Journal of Economic Perspectives, 8 (1): 55-72. Pasinetti, L., (1981), Structural Change and Economic Growth: a Theoretical Essay on the Dynamics of the Wealth of Nations, Cambridge, United Kingdom: Cambridge University Press. Pasinetti, L. (1993), Structural Economic Dynamics: A Theory of the Economic Consequences of Human Learning, Cambridge, United Kingdom: Cambridge University Press.