TY - JOUR
AU - Dario, Guarascio,
AB - Abstract This article empirically analyzes the link between public procurement (PP) and innovation activities, by taking into account the moderating effect played by import penetration on PP. Using industry-level information on patent applications for 24 countries over the period 1995–2012, we test the impact of PP on innovation activities and whether and in which direction import penetration on PP impacts on patenting. The econometric analysis relies on Poisson regression techniques aiming to investigate the correlation between patent counts, supply- as well as demand-side determinants, and controlling for country and sector heterogeneity. The obtained results confirm our main hypotheses. The dynamics of patenting is positively affected by the PP, while a high degree of import penetration reduces the innovation enhancing effect exerted by public demand. Our results suggest that public demand may represent an effective tool for industrial policy to stimulate innovative activities, shape the transformation of production systems, and foster industrial renewal. Moreover, the empirical evidence shows that the strategy regarding the degree of openness in PP toward nondomestic firms is a crucial policy choice capable of affecting the innovative potential of public demand. 1. Introduction The term “public procurement” (PP, hereafter) is used to identify the direct purchase of goods and services by the public sector. In 2015, PP accounted for a share of about 12% of the Organisation for Economic Co-operation and Development (OECD) countries’ gross domestic product and almost 30% of their national public spending.1 PP has been used as a policy instrument in pursuit of different goals: increasing aggregate demand, stimulating production and job creation, protecting domestic companies from global competition by encouraging investment and growth, increasing competitiveness of “national champions” by fostering their innovative capacity, and, finally, reducing regional disparities (Edquist and Hommen, 2000). No less significantly, PP is a key industrial and innovation policy tool. In fact, it can accelerate productive system upgrade, encouraging the development of sectors characterized by greater technological intensity, creating new markets, and stimulating demand-driven innovative investments (Rothwell, 1983; Weiss and Thurbon, 2006; Chang, 1994; Geroski, 1990; Wade, 1990; Edquist and Hommen, 2000; Mazzucato, 2013). In recent years, increasing attention has been paid to demand-side policies to foster innovation (Edler and Georghiou, 2007; OECD, 2011; Georghiou et al. 2014), and PP has been identified as a key instrument of innovation policy in both developed and developing countries (Uyarra and Flanagan, 2010; Crespi and Quatraro, 2013; Mazzucato et al. 2015; Edquist, 2015). The key role of PP appears to be particularly important in the current economic situation, as the 2008 crisis has led to a sharp decline in levels of economic activity and employment, especially in the manufacturing sectors. In Europe, the crisis has led to growing polarization in terms of employment, competitiveness, and industrial specialization (Simonazzi et al., 2013, Cirillo and Guarascio, 2015; Lucchese et al., 2016; Andreoni and Chang, 2016). In this scenario, PP might be particularly useful to reverse the economic trend characterizing the majority of European Union (EU) economies (Mazzucato et al. 2015). Public demand for goods and services could play a crucial role in macroeconomic terms, supporting aggregate demand and employment, but also reviving those innovative investments that are essential for sustaining industrial renewal, international competitiveness, and economic growth in the long run (Andreoni and Chang, 2016). Moreover, orienting public demand toward specific areas and firms can help strengthening regions and sectors more seriously harmed by the crisis. Directing public demand toward one or more specific production sectors represents a relevant strategy to promote the emergence or consolidation of production and markets characterized by high growth prospects (Pasinetti, 1981; Pianta, 2015; Mazzucato, 2016). This appears to be of particular relevance in high-tech (HT) industries, where returns on R&D investment are particularly uncertain. Thus, PP can produce an exogenous increase in demand for HT goods and services, stimulating innovation activities to capture these demand flows (Pasinetti, 1981). After the crisis, however, not all countries followed the same strategy. The United States included PP among their main countercyclical tools. Strategically oriented public investments have been put forth to upgrade the infrastructure network as well as to foster HT and green sectors (Wade, 2017). Moreover, the United States has sought to maximize the impact of its government spending with a strategy focused on domestic purchase of products (the so-called “Buy American” strategy compelling the government to purchase only domestically produced goods and services). By contrast, the EU responded to the crisis adopting deflationary policies designed to consolidate member states’ public budgets. For these reasons, many European countries have seen a reduction rather than increase in the scale of PP. Furthermore, the EU institutions in no way encouraged an increase in the domestic content of PP (i.e., no measures like the “Buy American” have been adopted in Europe). The aim of this work is to analyze, at the sectoral level, the relevance of the demand-pull influence of PP on innovation activities. Moreover, by means of panel data econometrics we investigate whether, and, if so, to what extent, import penetration of PP exerts a moderating effect in shaping the relationship between PP and the innovative activities of industries. In so doing, we analyze the possible trade-off between static and dynamic efficiency that should be taken into account in policy choices regarding PP strategies. Such a trade-off reflects the fact that, on the one hand, PP openness is thought to increase international competition leading prices of goods purchased by the public sector closer to marginal costs of production; on the other hand, the demand-pull stimulus toward local innovation activities might be undermined if the share of public demand captured by foreign producers is particularly large. The analysis draws upon a rich data set providing industry-level information on economic performance, PP, international trade, and production, as well as R&D expenditure and patents for all manufacturing industries—two-digit Nace Rev. 1—in 24 countries over the period 1995–2011.2 In particular, industry-level information on PP stems from the World Input–Output Database (WIOD) database, which collects Input/Output tables for a large set of countries.3 The article is organized as follows. In the next section, we review the relevant literature and spell out the research questions that we address in the empirical analysis. Section 3 describes the database adopted and construction of variables and traces out a descriptive picture of the dynamics of PP and import penetration in the EU and in the major OECD economies. Model, econometric strategy, and results are presented and discussed in Section 4, while Section 5 concludes providing some general remarks. 2. Literature background 2.1 The demand-pull effect of PP on innovative activities The role of demand-side factors has traditionally been underestimated in both economic theory and government policy, which largely focused on supply-side factors enhancing innovation, as if markets were always capable of passively absorbing the innovations introduced (Edler and Georghiou, 2007). However, demand conditions crucially affect the desirability and realization of inventions, while expected profitability resulting from the expansion of market demand represents the key stimulus to which inventive activities tend to respond (Kaldor, 1957; Schmookler, 1966; Mowery and Rosenberg, 1979; Kleinknecht and Verspagen, 1990; Dosi et al., 2010; Dawid et al., 2017). Previous evidence showed that both the quantity and the quality of demand may have a positive influence on innovation dynamics. The expansion of demand tends to trigger R&D investments (Kleinknecht, 1996; Brouwer and Kleinknecht, 1999), with innovations showing a pro-cyclical behavior (Geroski and Walters, 1995).4 In Pasinetti (1981), in turn, the development of innovations is seen as embedded in a process of continuous changes of the industrial structure, fundamentally driven by demand. However, the magnitude of the demand-pull effect may vary according to the type of demand flows faced by firms and industries (Crespi and Pianta, 2007; Bogliacino and Pianta, 2012; Guarascio et al. 2016). The quality of demand flows is particularly relevant, as user–producer interactions represent a further source of demand-pull effect on innovation (Von Hippel, 1986). Buyers can anticipate market demand by becoming early adopters and lead users, stimulating innovative activities for continuous improvement of both product and services. In so doing, they get involved in the innovation process and eventually become co-producers of user-driven innovations (Bresnahan and Greenstein, 2001; Malerba, 2007). As demonstrated by early contributions (Mowery and Rosenberg, 1979; Rothwell and Zegveld, 1981), PP can exert a significant influence on innovation activities with respect to both these channels (Cave and Frinking, 2003; Cabral et al. 2006), whether the stimulation of innovations is an explicit goal of procurement activities pursued by governments (Uyarra and Flanagan, 2010). First, considering that (regular) PP accounts for a good part of the overall demand for goods and services, it can play a key role in enlarging the market for new goods and services, thus providing an incentive to invest in innovation. Public demand can create or consolidate a market, thus reducing uncertainty, favoring the development of a critical mass encouraging R&D investment, and enabling dynamic increasing returns, especially in industries characterized by appreciable economies of scale, substantial R&D sunk costs, and high levels of uncertainty (see, for instance, Kaldor, 1981; Malerba, 2007). Second, when innovation becomes an explicit goal of PP, as in the case of innovative procurement, public organizations can influence innovation directly by purchasing new goods and services. In this case, the public sector uses its own demand/need or acts as a catalyst for needs located outside the public agency specifically to induce innovation, often becoming a lead user and co-creator through user–producer interactions (Edquist et al. 2015). The importance of public demand—and, in particular, of PP—as innovation driver is repeatedly stressed by Mazzucato (2016), who emphasizes their essential role in forming and creating new technological opportunities and market landscapes. In this context, the types of demand flows and the technological characteristics of industries receiving them are of primary importance. In particular, the implementation of public investment programs designed to solve specific societal problems—i.e., societal challenges such as climate change, obesity, ageing, etc.—tends to mobilize knowledge-intensive lines of production, stimulating company innovation, and generating economy-wide technological spillovers. Moreover, the orientation of public demand toward HT sectors may strengthen such industries, reinforcing their economic and innovative dynamics with benefits unfolding throughout the economy. From this point of view, and in line with the arguments presented in Mazzucato (2016), PP represents a key (selective) industrial policy tool having among its major objectives the promotion of innovation and technological upgrading. While the literature has so far examined these issues mainly through qualitative (case-based) analyses or (more rarely) micro-level quantitative studies, the present article aims, in the first place, to provide a large-scale empirical assessment of the role of PP in shaping innovation activities at the sectoral level for a large set of countries. In so doing, we are not able to distinguish between regular and innovative PP. However, we put forth a first (but still rough) attempt to empirically distinguish between the two PP types by using an ad hoc PP innovation propensity index beside the standard PP indicator (in the next section, we provide a detailed description of PP indicators adopted for the analysis). 2.2 The role of import penetration The effectiveness of PP in spurring innovation activities also depends on the specific strategies that countries adopt regarding PP, as for instance those related to the degree of market openness. This choice can be of major importance, since if part of domestic public demand is intercepted by other countries’ economies, the impact of PP on the internal production structure could be reduced. Procurement and innovation occur in space and the impacts of procurement on national (local) systems of innovation crucially depend on the spatial patterns of government procurement, on the degree of control over purchasing by the local and regional authorities, and on the extent to which benefits can be retained within a specific economic area through production linkages and knowledge spillovers (Porter, 1990; Uyarra and Flanagan, 2010). By giving preferential treatment to domestic firms in PP, governments can substantially reduce demand uncertainty and increase incentives to pursue innovative investment, as for instance was the case of the US aircraft industry, the Japanese mainframe computer industry, and the Finnish electronics industry (Chang and Andreoni, 2016). Finally, as preferentially acting on the proportions of local demand, PP can favor the role of increasing returns and learning effects in the generation of new knowledge, both within and outside individual firms (Arrow, 1962; Kaldor, 1966; Antonelli, 1999; Andreoni and Scazzieri, 2014; Stiglitz and Greenwald, 2014). For these reasons, when PP is used as an instrument of industrial policy to promote industrial renewal and the development of certain technologies, geographic areas, or types of enterprises—i.e., small- and medium-sized enterprises—governments tend to favor local companies so as to maximize the impact of public demand (see for instance Miyagiwa, 1991; Cohen, 2007; Peneder, 2017). Such an approach has been implemented for many years by US multistage, multicompetitor R&D programs, not only in the defense industry [Defense Advanced Research Projects Agency (DARPA)/Department of Defense (DOD)] but also in other areas such as energy, transport, health, and in the cross-sectoral Small Business Innovation Research Program (SBIR) (Ruttan, 2006; Block and Keller, 2009; Andreoni, 2016). Policies aiming at ensuring the effectiveness of PP—from both the demand- and the supply-side perspective—can be considered policies of “discriminatory procurement.” The discrimination occurs when governments use PP to encourage local producers. Such discrimination may be explicit, as in the case of the “Buy American” clause in the United States—the public purchase of foreign goods is in some cases avoided—or implicit, using price discrimination or taxes. Discrimination typically occurs, however, through tacit agreements between local producers and governments realized by contracting the time of publicizing the tender or imposing technical specifications favoring local producers (Lowinger 1976; Beviglia-Zampetti 1997, Rickard and Kono, 2014). From an economic point of view, protecting local companies by means of a discriminatory use of PP constitutes a barrier to international trade and is equivalent to imposing import restrictions (Brulhart and Trionfetti 2004; Trionfetti 2000). Historically, this tool has been widely used to protect domestic industry and promote the growth of the so-called “national champions” (Cohen, 2007). In this respect, discriminatory PP has been crucial in supporting Asian countries as Japan and Korea in their process of (productive and technological) catching up with more advanced countries (Okimoto, 1989; Singh, 2002; Ruttan, 2006). Moreover, this is a strategy that these and other advanced economies still employ5 and that is also widely used by many developing countries such as China (OECD, 2011). Regarding these aspects, it is worth noting the difference in the attitudes of the United States and the EU toward the use of PP. The United States, in fact, has always maintained a very extensive and stringent set of rules protecting the American market of PP while promoting American businesses (Wade, 2017). Ever since the Buy American Act of 1933, in fact, the US government has identified a core of public purchasing programs in which the federal government and national agencies are obliged to facilitate the purchase of goods and services produced locally. On the contrary, EU policies on PP have been explicitly designed to prevent any discrimination likely to favor domestic producers. The principle behind this approach is that the increased competition among suppliers—due to the opening up to international competition—enables efficiency gains in the production system, greater efficiency in public spending, and thus welfare gains for the EU as a whole. This attitude raises some concern because, on the one hand, the EU does not seem to be having much success in its ambition to open up the global market for PP significantly while, on the other hand, European companies appear to be penalized by different access rules on foreign markets of PP (European Commission, 2011). Building on this discussion, the present article seeks to test the hypothesis that import penetration of PP affects the relationship between public demand and industry innovativeness by reducing the potential of demand-pull effects of PP in triggering innovation activities and spurring industrial renewal. 3. Data and descriptive evidence The empirical analysis is carried out over a panel of manufacturing industries for 24 OECD countries observed over the period 1995–2012. The data set provides a combination of Input–Output data on PP and of industry-level information on patents. Adopting the procedure proposed by Lybbert and Zolas (2014), we first derive information on patents in each industry for all the countries considered across the entire time span selected (details on the procedure adopted to build the data set are supplied below and in the Appendix).6 Second, we rely on the WIOD to build industry-level variables on PP, import penetration on PP, and exports (Timmer et al., 2015). Third, we use data on R&D expenditure drawn from the OECD ANBERD database to control for industries’ investment in innovation activities. Industries’ innovative dynamics is captured by the sector performance on patents. This choice is mainly motivated by the necessity to build a sufficiently long panel data set on innovation performance, which cannot be alternatively derived from more detailed information sources such as innovation surveys. However, while widespread in the literature, the use of patent data presents several drawbacks: (i) the distribution of patents across firms and sectors is highly skewed; (ii) there is a large variance in patent quality, and, most importantly (iii) only a fraction of innovations is patented (Griliches, 1990; Archibugi and Pianta, 1992; Jaffe and Trajtenberg, 2004). With regard to our study, by adopting patents as a measure of innovation, we are able to capture only some more relevant technological innovation activities, thus excluding all the other forms of innovations that are part of the broad definition provided by the Oslo Manual (OECD, 2005). Accordingly, the interpretation of the empirical results should account for such a limited information content of the adopted innovation measure. Since information on industries’ patenting activities is not generally available, an imputation procedure to break country-level information down to the industry-level is needed. In this respect, significant advances have been made thanks to the imputation technique recently elaborated by Lybbert and Zolas (2014).7 We follow their approach, obtaining a 15-year-long panel of industries’ patenting activities.8 Moreover, we compute patent stock using a standard inventory methods accounting for the depreciation of the knowledge stock (see the Appendix for illustration of the methodology adopted). Once patents are assigned to industrial sectors according to the methodology of Lybbert and Zolas (2014), the three data sources—namely, the WIOD, the OECD ANBERD, and the EPO patent database—are merged together. To this end, the ISIC classification in accordance with the Nomenclature statistique des activités économiques (NACE) information is used to match patents with PP, import penetration, export, and R&D data. In Table 1, we provide the list of variables included in the analysis. Table 1. List of variables and sources Variable Unit Source Patents filed at the EPO Absolute value EPO PP Millions of euros (real values) WIOD Innovative PP propensity index Millions of euros (real values) WIOD and EUROSTAT Import penetration on PP Share WIOD Export over value added Share WIOD R&D expenditure Thousands of euros (real values) ANBERD Variable Unit Source Patents filed at the EPO Absolute value EPO PP Millions of euros (real values) WIOD Innovative PP propensity index Millions of euros (real values) WIOD and EUROSTAT Import penetration on PP Share WIOD Export over value added Share WIOD R&D expenditure Thousands of euros (real values) ANBERD View Large Table 1. List of variables and sources Variable Unit Source Patents filed at the EPO Absolute value EPO PP Millions of euros (real values) WIOD Innovative PP propensity index Millions of euros (real values) WIOD and EUROSTAT Import penetration on PP Share WIOD Export over value added Share WIOD R&D expenditure Thousands of euros (real values) ANBERD Variable Unit Source Patents filed at the EPO Absolute value EPO PP Millions of euros (real values) WIOD Innovative PP propensity index Millions of euros (real values) WIOD and EUROSTAT Import penetration on PP Share WIOD Export over value added Share WIOD R&D expenditure Thousands of euros (real values) ANBERD View Large The variable capturing industry-level PP is computed as the sum of goods and services—both intermediate and final—purchased by the public sector from each two-digit industry. In so doing, we follow the strategy adopted by Messerlin and Mirodout (2012). We add to the information on government expenditure on goods and services the purchases made by the following industries: “electricity, gas and water supply” (100%), “post and telecommunications” (50%), “public administrations and defence, compulsory social security” (100%), “education” (100%), and “health and social work” (100%). Therefore, for each sector i, country j, and year t, the PP variable gives the value of goods and services purchased by the government of the country concerned j from that sector i (see equation (2) below). As argued in the literature background section, and at length by Edquist (2015), the PP pro-innovative stance may vary considerably according to the “type” of PP taken into consideration. In particular, it is worth distinguishing between regular and innovation-related PP, with the latter intended as government purchases of goods and services that “explicitly promote innovation.” To capture the heterogeneous propensity of governments toward innovation procurement, we build an indicator—i.e., the innovative PP propensity index—by weighting information on sectoral PP with the country-level share of public expenditure devoted to R&D, as reported by Eurostat. Albeit simple, this indicator allows for combination of, on one hand, information on the relevance and dynamics of PP in each sector and country, and, on the other, information on the priority each government attributes to innovation. The hypothesis behind the use of this proxy is that the relevance of innovation-oriented PP in each year and industry is positively related to the level of engagement of public sector in research and innovation investment. Relying on this indicator, we test whether, and if so to what extent, the PP demand-pull effect is affected by governments’ innovation propensity. Formally, the innovation PP propensity index can be represented as follows (1): IPPijt =Gijt + ∑ijt Gk sec T *R&DSHjt (1) where for each sector-country-year triple, the innovative PP index is equal to the sectoral PP—i.e., the sum of government purchases of final goods and services from a certain sector i plus the sum of intermediate goods purchased from the same sector by the k industries (k = 1,…, 5)—weighted by the share of public expenditure for R&D activities out of the total. The variable identifying the degree of import penetration of PP is calculated as the share of (domestic) PP to sector j which is captured by foreign producers (Messerlin and Mirodout, 2012).9 The import penetration indicator (2) is computed as: Imp_ penijt = (∑kk=1,…,n;k≠jGikt+ ∑kk=1,…,n;k≠jGikt sec T)Tot_PPijt (2) where the numerator is the sum of intermediate and final goods imported by the government of country j from a certain sector i; the denominator is total public demand by the government of country j directed toward sector i—i.e., the first term on the right hand side of equation (1). Hence, (2) measures how much PP—for each country j and sector i—directed toward a certain sector relies on foreign producers rather than on domestic ones. 3.1 The dynamics of PP and import penetration As a preliminary step, we analyze the dynamics of PP and import penetration on PP between 1995 and 2011. Over the considered time span, the share of PP on total aggregate demand increases in Europe, the United States, and Japan (see Figure 1).10 Figure 1. View largeDownload slide Share of PP on total demand in EU27, EU3 (Ger, It, and Fr), the United States, and Japan. Source: Our elaboration on WIOD data. Figure 1. View largeDownload slide Share of PP on total demand in EU27, EU3 (Ger, It, and Fr), the United States, and Japan. Source: Our elaboration on WIOD data. However, some relevant elements of heterogeneity are detectable. During the 1995–2000 period, PP’s share declines in Europe, while it increases in Japan and, to a lower extent, in the United states. Along the 2000s, the EU27’s PP share lies above both the US and the Japanese one, while figures are lower for the EU3 aggregate. The 2008 crisis coincides with a sharp drop of PP shares in all the considered countries with Japan being the only one showing a slight recovery after 2010. Moving to the dynamics of import penetration (Figure 2), it emerges a watershed around the year 2000. In all the economies taken into consideration, this year coincides with the start of a trend of increasing import penetration on PP continuing until the crisis (and in the European case even afterward). The discontinuity observed around the 2000 may be linked to the beginning of a phase during which the intensity of trade and production globalization skyrockets following, for example, events such as China joining the World Trade Organization. Comparatively, the United States show the highest degree of import penetration with Europe remaining a slightly behind. The degree of import penetration on PP, in turn, is significantly lower in Japan in the early 2000s but tends to converge to the US and EU27 levels around 2008. Figure 2. View largeDownload slide Import penetration on PP in EU27, EU3 (Ger, It, and Fr), the United States, and Japan. 1995–2011 Source: Our elaboration on WIOD data. Figure 2. View largeDownload slide Import penetration on PP in EU27, EU3 (Ger, It, and Fr), the United States, and Japan. 1995–2011 Source: Our elaboration on WIOD data. After 2008, the dynamics of import penetration on PP starts diverging. While the EU27 and the EU3 are continuing in a trend of increasing import penetration on PP, in the United States there has been a sharp fall that stops, with a gradual recovery, toward the end of 2009. During the crisis, thus, the degree of openness of PP in Europe and in the United States starts diverging substantially. This evidence can be associated with the effects of “protectionist” measures—i.e., the “Buy American” act—aimed at protecting PP and put forth in the United States early after the crisis explosion. A reinforced “buy American” would seem to explain, at least in part, the sudden divergence in the dynamics of import penetration depicted in Figure 2. A dynamics of divergence characterizes also the trend of import penetration observed in Japan as opposed to the EU27 and EU3 ones. The comparison between Europe and Japan shows that the latter lies well below the two EU aggregates for the whole period considered and, in particular, from the crisis onward. This is possibly due, on the one hand, to the traditional introversion of the Japanese economy, as testified also by the lower degree of import penetration characterizing the 1995–2007 phase.11 Similarly to what has been shown concerning the United States, on the other hand, the marked reduction of import penetration observable after 2008 can be due to the adoption of a “protectionist” strategy put forth by the Japanese government as a response to the drop in aggregate demand following the crisis. Such descriptive evidence confirms the relevance of PP as a major component of effective demand and suggests that different strategies are adopted by different countries in terms of its openness to international markets. The following econometric analysis is designed to identify the role of these variables in shaping industries’ innovation activities. 4. Model, econometric strategy, and results 4.1 The model The relationship between PP and industries’ innovative dynamics and the role exerted by import penetration is investigated relying on the following econometric specification (3): PATijt = β1*PPijt-1+β2*IPPijt-1 + β3*IMPPPijt-1+β4*IMPPPijt-1*PPijt-1+β5*IMPPPijt-1*IPPijt-1+ β6*Xijt-1+ εijt (3) where i stands for sector, j for country, and t for time. The dependent variable PATijt is the patent stock—we considered the stock of patents at time t accounting for both new patents and depreciation of the knowledge stock. PPijt-1 is the PP variable providing information on PP directed to each sector i at time t − 1, while IPPijt-1 is the PP innovation propensity index described above. The degree of import penetration on PP is identified by IMPPPijt-1 ⁠, and the latter is than interacted with both PPijt-1 and IPPijt-1 ⁠. The term Xijt-1 ⁠, in turn, includes a set of controls such as the lagged change in sectoral R&D expenditure and lagged export intensity—i.e., gross exports on sectoral value added—while εijt is the standard error term. The coefficients β1 and β2 identify the relationships between the two PP indicators and innovation activities, while β3 measures the general effect exerted by the degree of openness of public demand. Finally, β4 and β5 capture the (average) effect that import penetration has on the PP–innovation relationship. The use of lags has both theoretical and methodological reasons. Theoretically, we argue that the potential effects of PP on industries’ innovation dynamics—as well as the effect of the other variables included in (3)—come into operation only after a time lag. From a methodological standpoint, using lagged regressors mitigates the risk of simultaneity-related endogeneity. The specification in (3) constitutes an enhancement with respect to the previous models exploring the determinants of innovation in industries. In particular, we follow the contributions examining jointly the impact on industries’ innovative performance in technology-push and demand-pull factors—see, among the others, Crespi and Pianta (2007), Guarascio et al. (2015, 2016), and Guarascio and Pianta (2017), by considering variables that explicitly account for PP and import penetration of PP. The key hypotheses tested by estimating the model in (3) can be summarized as follows. To begin with, we test whether PP is positively associated with industries’ patenting activities, controlling for both sectoral R&D efforts and export propensity. Second, we verify whether the relationship between PP and patents changes when PP is distinguished between a general measure and the one proxying innovation-oriented PP (see above for a description of the index). Third, we investigate whether, and if so to what extent, a relatively more intense import penetration of PP affects the relationship between the latter and industries’ innovativeness. In other words, we aim to test the hypothesis that the degree of openness of PP to foreign markets affects the capacity of public demand to stimulate innovation. Having analyzed the general relationships—i.e., estimating the model over the full sample of countries and industries included in the analysis—we explore two different levels of heterogeneity. First, considering the peculiar attitude of EU countries toward the limitations to the strategic use of PP in sustaining domestic producers previously evidenced, we focus on country-level heterogeneity and test whether the relations emerging from the full sample model vary if only European industries are considered. Second, we concentrate on sectoral heterogeneity and test whether the identified relationships change when distinction is made between industries characterized by different technological intensity. 4.2 Econometric strategy and results The use of patent data as proxies of the innovative activity implies that we have to deal with count variables, that is, variables with nonnegative integer values. Econometric models specifically designed for this kind of variable are the Poisson regression model (PRM) and the negative binomial regression model (NBRM). Given the characteristics of our data and the limited number of zeros in our count variables, we analyze the relation between PP and industries patenting activities relying on the PRM.12 Poisson regression fits models of the number of occurrences (counts) of an event by assuming that the dependent variable has a Poisson distribution. Moreover, it assumes that the logarithm of its expected value can be modeled by a linear combination of unknown parameters. To control for industry-level heterogeneity, we exploit the panel structure of our data by implementing the Poisson fixed effects estimator.13 In this way, we limit the risk that the detected relationships are driven by industry- or country-specific idiosyncratic elements. Finally, to soften potential endogeneity biases and ensure estimations consistency, we include all regressors at their first lag. All the estimations are performed adopting a stepwise procedure based on four different specifications. In the first specification, industries’ patents are regressed against the annual change (expressed as logarithmic difference) in R&D expenditure and export intensity (exports over value added) only. The second step includes the lagged change in PP (logarithmic difference) testing the degree of correlation between public demand and patenting performance. The third specification includes both PP and PP innovation propensity index to check for potential heterogeneity between the two indicators. Finally, we test the full specification having, on the right-hand side, change in R&D expenditure, export intensity, PP, PP innovation propensity index, degree of import penetration, and interaction terms between the two PP indicators and the import penetration variable. Table 2 reports the result of the full sample model. Industries patenting activities show a significant association with both export intensity and R&D expenditure. The strongest result regards the PP–innovation nexus, confirming the positive demand-pull effect exerted by public demand on innovation activities. Looking at sign and significance of the PP coefficients, it emerges that industries characterized by a relatively more intense PP flows have also more intense innovative dynamics in terms of patents. Such positive association between PP and industries patenting activities is confirmed by the coefficient associated with the PP innovation propensity index. Table 2. Fixed effects Poisson estimations—dependent variable: patents stock (full sample model) (1) (2) (3) (4) Export intensity (first lag) 0.00716*** 0.00672*** 0.00649*** 0.00549** (0.00162) (0.00159) (0.00163) (0.00205) Δ R&D expenditure (first lag) 0.0558 0.0526 0.106** 0.102*** (0.0492) (0.0493) (0.0331) (0.0308) Δ PP (first lag) 0.122*** 0.124*** 0.264*** (0.0352) (0.0341) (0.0539) Δ PP innovation propensity index (first lag) 0.0916** 0.113** (0.0293) (0.0388) Import penetration on PP (first lag) 0.000912 (0.000963) Δ PP * import penetration (first lag) −0.00304*** (0.000770) Δ PP innovation propensity index* import penetration (first lag) −0.000534 (0.00105) Observations 6976 6974 6210 6210 (1) (2) (3) (4) Export intensity (first lag) 0.00716*** 0.00672*** 0.00649*** 0.00549** (0.00162) (0.00159) (0.00163) (0.00205) Δ R&D expenditure (first lag) 0.0558 0.0526 0.106** 0.102*** (0.0492) (0.0493) (0.0331) (0.0308) Δ PP (first lag) 0.122*** 0.124*** 0.264*** (0.0352) (0.0341) (0.0539) Δ PP innovation propensity index (first lag) 0.0916** 0.113** (0.0293) (0.0388) Import penetration on PP (first lag) 0.000912 (0.000963) Δ PP * import penetration (first lag) −0.00304*** (0.000770) Δ PP innovation propensity index* import penetration (first lag) −0.000534 (0.00105) Observations 6976 6974 6210 6210 Note: Standard errors in parentheses. * P < 0.05, ** P < 0.01, *** P < 0.001. View Large Table 2. Fixed effects Poisson estimations—dependent variable: patents stock (full sample model) (1) (2) (3) (4) Export intensity (first lag) 0.00716*** 0.00672*** 0.00649*** 0.00549** (0.00162) (0.00159) (0.00163) (0.00205) Δ R&D expenditure (first lag) 0.0558 0.0526 0.106** 0.102*** (0.0492) (0.0493) (0.0331) (0.0308) Δ PP (first lag) 0.122*** 0.124*** 0.264*** (0.0352) (0.0341) (0.0539) Δ PP innovation propensity index (first lag) 0.0916** 0.113** (0.0293) (0.0388) Import penetration on PP (first lag) 0.000912 (0.000963) Δ PP * import penetration (first lag) −0.00304*** (0.000770) Δ PP innovation propensity index* import penetration (first lag) −0.000534 (0.00105) Observations 6976 6974 6210 6210 (1) (2) (3) (4) Export intensity (first lag) 0.00716*** 0.00672*** 0.00649*** 0.00549** (0.00162) (0.00159) (0.00163) (0.00205) Δ R&D expenditure (first lag) 0.0558 0.0526 0.106** 0.102*** (0.0492) (0.0493) (0.0331) (0.0308) Δ PP (first lag) 0.122*** 0.124*** 0.264*** (0.0352) (0.0341) (0.0539) Δ PP innovation propensity index (first lag) 0.0916** 0.113** (0.0293) (0.0388) Import penetration on PP (first lag) 0.000912 (0.000963) Δ PP * import penetration (first lag) −0.00304*** (0.000770) Δ PP innovation propensity index* import penetration (first lag) −0.000534 (0.00105) Observations 6976 6974 6210 6210 Note: Standard errors in parentheses. * P < 0.05, ** P < 0.01, *** P < 0.001. View Large Interestingly, when the degree of import penetration on PP is taken into account, first, it emerges that there is not a direct effect on innovation exerted by PP openness, suggesting that increasing international competition in PP does not spur innovative activities of domestic industries. On the contrary, in line with the hypothesis put forth above, a high degree of import penetration exerts a (negative) moderating effect on the PP–innovation relation, as testified by the coefficient associated with the PP–import penetration interaction term. No significant effects are registered with respect to the interaction between import penetration and the PP innovation propensity index. According to the full sample model results, thus, PP displays a significant and positive impact on industries’ innovative dynamics. Interestingly, both general PP and innovation-oriented PP variables are jointly significant, suggesting that the proposed indexes are capable of capturing different demand-pull effects on innovation played by public sector. In turn, those industries facing a domestic public demand that comparatively relies more on foreign producers are characterized by a relatively weaker innovative dynamics in terms of patents. In commenting these results it should be noted that the choices on PP strategies and their actual effectiveness in terms of innovation impact might be influenced by the different characteristics of production structures of the economies and by their positioning in the global value chains of productions (Gereffi, 1999; Gereffi et al., 2001). In particular, the potential of PP to stimulate innovation activities of domestic companies can be strongly reduced, even if “buy national” strategies are implemented, when large part of value added incorporated in final goods is generated abroad. While we are not able to investigate these aspects in our model, our results in any case suggest that, on average, the moderating effect of import penetration on the innovation enhancing contribution of PP is relevant. Before moving to the analysis of country and sectoral heterogeneities, we perform a robustness check on this first set of result, by replicating the estimations in Table 2 controlling for two potential sources of bias.14 Given the risk of multicollinearity between R&D and exports, we run the estimations including and excluding these two variables verifying if and to what extent this has any influence on the regressors of interest. Besides, we implement a standard variance inflation factor (VIF) test to check the degree of multicollinearity of all explanatory variables. As Table A1 (Appendix) shows, the inclusion/exclusion of R&D and exports does not affect the results of the baseline model. Looking at both models in Table A1, the magnitude and significance of the coefficients associated with our key variables (PP, PP–innovation-propensity index, and interaction terms) result to be stable. Remarkably, when exports are excluded the import penetration indicator turns out to be significant, contrarily to what emerged from the baseline model. This result might be driven by the fact that when exports are excluded the import penetration indicator is likely to partly capture the effect associated with the general variability of international economic openness across different industries. Moreover, the outcome of the VIF test confirms that multicollinearity is not affecting econometric results, as VIF values are below the 10 thresholds for all regressors.15 Second, we test the robustness of the baseline model’s specification with respect to the R&D variable. More specifically, R&D might be influenced by the dynamics of demand—on this point, see, for example, Brouwer and Kleinknecht (1999) and Piva and Vivarelli (2007)—casting doubts on the appropriateness of the specification in (3). Therefore, we estimate a Two-Stage Least Squares (2SLS) regression, where at the first stage the change in R&D expenditure is regressed against its lag—used as instrument in accordance with the large literature emphasizing the persistency of both firm- and industry-level R&D activities16—and all explanatory variables in (3) including the change in PP.17 As reported in Tables A3 and A4, all the baseline model results are robust to the test on R&D showing stability in terms of significance and even an increase in magnitude for the PP coefficient. Overall, the robustness checks reported here provides support to both the specification adopted as well as the results on the examined key variables. Table 3. Fixed effects Poisson estimations—dependent variable: patents stock (European industries only) (1) (2) (3) (4) Export intensity (first lag) 0.00762*** 0.00721*** 0.00683*** 0.00532*** (0.00162) (0.00160) (0.00162) (0.00155) Δ R&D expenditure (first lag) 0.0432 0.0366 0.0360 0.0331 (0.0286) (0.0278) (0.0275) (0.0273) Δ PP (first lag) 0.110*** 0.104*** 0.231*** (0.0269) (0.0253) (0.0378) Δ PP innovation propensity index (first lag) 0.0678** 0.165*** (0.0257) (0.0431) Import penetration on PP (first lag) 0.00159 (0.00120) Δ PP * import penetration (first lag) −0.00232** (0.000754) Δ PP innovation propensity index* import penetration (first lag) −0.00226* (0.000899) Observations 4881 4880 4880 4880 (1) (2) (3) (4) Export intensity (first lag) 0.00762*** 0.00721*** 0.00683*** 0.00532*** (0.00162) (0.00160) (0.00162) (0.00155) Δ R&D expenditure (first lag) 0.0432 0.0366 0.0360 0.0331 (0.0286) (0.0278) (0.0275) (0.0273) Δ PP (first lag) 0.110*** 0.104*** 0.231*** (0.0269) (0.0253) (0.0378) Δ PP innovation propensity index (first lag) 0.0678** 0.165*** (0.0257) (0.0431) Import penetration on PP (first lag) 0.00159 (0.00120) Δ PP * import penetration (first lag) −0.00232** (0.000754) Δ PP innovation propensity index* import penetration (first lag) −0.00226* (0.000899) Observations 4881 4880 4880 4880 Note: Standard errors in parentheses. * P < 0.05, ** P < 0.01, *** P < 0.001. View Large Table 3. Fixed effects Poisson estimations—dependent variable: patents stock (European industries only) (1) (2) (3) (4) Export intensity (first lag) 0.00762*** 0.00721*** 0.00683*** 0.00532*** (0.00162) (0.00160) (0.00162) (0.00155) Δ R&D expenditure (first lag) 0.0432 0.0366 0.0360 0.0331 (0.0286) (0.0278) (0.0275) (0.0273) Δ PP (first lag) 0.110*** 0.104*** 0.231*** (0.0269) (0.0253) (0.0378) Δ PP innovation propensity index (first lag) 0.0678** 0.165*** (0.0257) (0.0431) Import penetration on PP (first lag) 0.00159 (0.00120) Δ PP * import penetration (first lag) −0.00232** (0.000754) Δ PP innovation propensity index* import penetration (first lag) −0.00226* (0.000899) Observations 4881 4880 4880 4880 (1) (2) (3) (4) Export intensity (first lag) 0.00762*** 0.00721*** 0.00683*** 0.00532*** (0.00162) (0.00160) (0.00162) (0.00155) Δ R&D expenditure (first lag) 0.0432 0.0366 0.0360 0.0331 (0.0286) (0.0278) (0.0275) (0.0273) Δ PP (first lag) 0.110*** 0.104*** 0.231*** (0.0269) (0.0253) (0.0378) Δ PP innovation propensity index (first lag) 0.0678** 0.165*** (0.0257) (0.0431) Import penetration on PP (first lag) 0.00159 (0.00120) Δ PP * import penetration (first lag) −0.00232** (0.000754) Δ PP innovation propensity index* import penetration (first lag) −0.00226* (0.000899) Observations 4881 4880 4880 4880 Note: Standard errors in parentheses. * P < 0.05, ** P < 0.01, *** P < 0.001. View Large Table 4. Fixed effects Poisson estimations—dependent variable: patents stock (HT-SB and Specialized Supplier industries) (1) (2) (3) (4) Export intensity (first lag) 0.00957** 0.00871** 0.00845** 0.00953* (0.00320) (0.00312) (0.00311) (0.00466) Δ R&D expenditure (first lag) −0.00204 −0.0235 0.284* 0.292** (0.200) (0.193) (0.118) (0.112) Δ PP (first lag) 0.237*** 0.216** 0.562*** (0.0714) (0.0722) (0.122) Δ PP innovation propensity index (first lag) 0.142* 0.280* (0.0639) (0.120) Import penetration on PP (first lag) −0.000362 (0.00238) Δ PP * import penetration (first lag) −0.00733*** (0.00215) Δ PP innovation propensity index* import penetration (first lag) −0.00344 (0.00225) Observations 1591 1591 1415 1415 (1) (2) (3) (4) Export intensity (first lag) 0.00957** 0.00871** 0.00845** 0.00953* (0.00320) (0.00312) (0.00311) (0.00466) Δ R&D expenditure (first lag) −0.00204 −0.0235 0.284* 0.292** (0.200) (0.193) (0.118) (0.112) Δ PP (first lag) 0.237*** 0.216** 0.562*** (0.0714) (0.0722) (0.122) Δ PP innovation propensity index (first lag) 0.142* 0.280* (0.0639) (0.120) Import penetration on PP (first lag) −0.000362 (0.00238) Δ PP * import penetration (first lag) −0.00733*** (0.00215) Δ PP innovation propensity index* import penetration (first lag) −0.00344 (0.00225) Observations 1591 1591 1415 1415 Note: Standard errors in parentheses. * P < 0.05, ** P < 0.01, *** P < 0.001. View Large Table 4. Fixed effects Poisson estimations—dependent variable: patents stock (HT-SB and Specialized Supplier industries) (1) (2) (3) (4) Export intensity (first lag) 0.00957** 0.00871** 0.00845** 0.00953* (0.00320) (0.00312) (0.00311) (0.00466) Δ R&D expenditure (first lag) −0.00204 −0.0235 0.284* 0.292** (0.200) (0.193) (0.118) (0.112) Δ PP (first lag) 0.237*** 0.216** 0.562*** (0.0714) (0.0722) (0.122) Δ PP innovation propensity index (first lag) 0.142* 0.280* (0.0639) (0.120) Import penetration on PP (first lag) −0.000362 (0.00238) Δ PP * import penetration (first lag) −0.00733*** (0.00215) Δ PP innovation propensity index* import penetration (first lag) −0.00344 (0.00225) Observations 1591 1591 1415 1415 (1) (2) (3) (4) Export intensity (first lag) 0.00957** 0.00871** 0.00845** 0.00953* (0.00320) (0.00312) (0.00311) (0.00466) Δ R&D expenditure (first lag) −0.00204 −0.0235 0.284* 0.292** (0.200) (0.193) (0.118) (0.112) Δ PP (first lag) 0.237*** 0.216** 0.562*** (0.0714) (0.0722) (0.122) Δ PP innovation propensity index (first lag) 0.142* 0.280* (0.0639) (0.120) Import penetration on PP (first lag) −0.000362 (0.00238) Δ PP * import penetration (first lag) −0.00733*** (0.00215) Δ PP innovation propensity index* import penetration (first lag) −0.00344 (0.00225) Observations 1591 1591 1415 1415 Note: Standard errors in parentheses. * P < 0.05, ** P < 0.01, *** P < 0.001. View Large 4.3 The test on the European economies As discussed before, different approaches to PP as well as different degree of openness of PP to foreign producers may significantly impact on the public demand–innovation relation. Trying to capture potential heterogeneities between Europe and other countries included in our sample—i.e., Japan, Korea, Russia, Turkey, and the United States—we estimate the model in (3) on the subsample of European industries (Table 3).18 The European test shows both similarities and differences with respect to the full sample model. First of all, the positive relation between PP and industries’ innovation performance is confirmed. Even in Europe, PP confirms its role in enhancing innovation activities at the industry level. This result holds looking at both the general PP variable as well as the PP innovation propensity index. The major difference regards the role of import penetration on PP. Differently from the previous estimation, the moderating effect of import penetration is now detected with respect to both the interaction terms. This result could reflect the different treatment in terms of market protection reserved to precommercial PP across countries. Indeed, while so far the European Union has taken very limited advantage of procurement for research and development and innovation (ECWG, 2006), countries like Japan, Korea, the United States, and others largely used this instrument to favor domestic innovation activities (Kattel and Lember, 2010). Overall, the test on European industries provides support to the hypothesis that PP is a key driver of industries’ innovation performance. This is true irrespectively from the variable used to measure PP. When it comes to import penetration, however, the latter seems to penalize the pro-innovative effect of public demand with respect to both PP and PP innovation propensity index. In this respect, the potential effectiveness of a strategic use of both PP and innovation-oriented PP to relaunch European industries appears of particular relevance. 4.4 The role of industries technological heterogeneity As already argued, the relationship between PP and industries’ innovation dynamics may vary substantially according to industries technological characteristics. Hence, we test whether the association detected pooling all industries is reshaped when sectors grouped according to their different technological characteristics are separately analyzed. The main hypothesis is that the positive association between PP and industries innovative activities is stronger for those sectors characterized by a relatively more intense use of knowledge and technology, and that the moderating effect of import penetration can be more severe in these industries. In so doing, we rely on different taxonomies to explore the role of industries’ technological heterogeneity. First, we cluster industries using the revised Pavitt taxonomy—proposed by Bogliacino and Pianta (2016)—as in Guarascio et al. (2016), by including in the HT cluster the industries belonging to the Science Based (SB) and Supplier Specialized (SS) Pavitt’s groups—i.e., industries identified as those that more intensively rely on innovation. We then include in the Low-Tech (LT) cluster the industries belonging to the Supplier Dominated (SD) and Scale Intensive (SI) groups. According to the Pavitt taxonomy’s rationale, SB and SS industries are characterized by: (i) a relatively more intense use of knowledge and technology in their production processes as compared to other sectors; (ii) a greater relevance of HT intermediate inputs; (iii) n higher openness to the foreign markets. Conversely, SI and SD industries are characterized by a comparatively lower propensity toward innovation and by an intensive use of LT inputs originating principally from the domestic market. The full list of industries included in the Bogliacino and Pianta’s (2016) revised Pavitt taxonomy is reported in the Appendix. Second, as an additional test, we group industries adopting the taxonomy proposed by Peneder (2010).19 The latter characterizes industries by the distribution of diverse innovation modes at the firm level. Industries are grouped in five categories: (i) high innovation intensity—industries populated by highly creative firms focusing on product innovation, performing intensively intramural R&D and where appropriability regimes depends on the use of patents; (ii) intermediate-to-high innovation intensity—industries with an intermediate share of creative firms mostly involved in process innovations, investing intensively in R&D but also relying on external acquisitions to access innovation-related knowledge; (iii) intermediate innovation intensity—industries are characterized by a strong heterogeneity concerning firms relative innovativeness, weak intramural R&D expenditure, and not so frequent use of formalized means of knowledge protection as patents; (iv) intermediate-to-low innovation intensity—industries with a high share of firms with adaptive behavior, dependent on the external to acquire new technologies and displaying rare use of patents and internal knowledge; (v) low innovation intensity—industries characterized by a predominant share of firms performing neither innovation activities nor applying any measures for appropriation. To perform our second test, thus, we replicate the analysis grouping industries as medium high-tech—i.e., industries belonging to the Peneder’s (2010) high, medium-to-high, and medium clusters; and LT—i.e., industries belonging to the medium-low and low clusters. Tables 4 and 5 report results relative to HT and LT industries, respectively. In the former, when looking at Columns 3 and 4, it can be noted that the variables capturing industries export intensity and R&D efforts are both positively and significantly associated with the dynamics of patents. The same relationships appear to be lesser robust in LT industries, with the variable capturing industries export intensity losing its significance when the model is fully specified (Column 4). This finding is in line with the comparatively lower degree of internationalization characterizing this group of sectors as opposed to the SB and SS ones. As expected, PP has a strong impact on SB and SS’ industries’ patenting performance, while the PP innovation propensity index shows a weaker (albeit positive) statistical association with the dependent variable. Interestingly, the innovation enhancing role of PP is confirmed also in LT industries, though the magnitude of this effect appears to be lower than in HT sectors. This result is particularly important, since it points to a potential “upgrading effect” of PP. That is, the presence of relevant PP flows seems to stimulate the development of innovations even in industries traditionally less prone to this kind of activities. In this sense, one may argue that a large-scale and well-targeted PP can not only work on the “intensive margin”—i.e., pushing innovative activities of those industries that are already characterized by the introduction of new products, new processes, and new patents but also on the extensive one—i.e., favoring the upgrading of LT industries and the diffusion of innovative activities among firms populating the SD and SI groups. Table 5. Fixed effects Poisson estimations—dependent variable: patents stock (LT-SI and SD industries) (1) (2) (3) (4) Export intensity (first lag) 0.00526*** 0.00496** 0.00456** 0.00321 (0.00153) (0.00151) (0.00162) (0.00193) Δ R&D expenditure (first lag) 0.0683* 0.0674* 0.0689* 0.0670* (0.0286) (0.0295) (0.0313) (0.0292) Δ PP (first lag) 0.0830* 0.0894** 0.195*** (0.0328) (0.0327) (0.0448) Δ PP innovation propensity index (first lag) 0.0767* 0.0890* (0.0320) (0.0407) Import penetration on PP (first lag) 0.00136 (0.000911) Δ PP * import penetration (first lag) −0.00230** (0.000714) Δ PP innovation propensity index* import penetration (first lag) −0.000342 (0.00123) Observations 5385 5383 4795 4795 (1) (2) (3) (4) Export intensity (first lag) 0.00526*** 0.00496** 0.00456** 0.00321 (0.00153) (0.00151) (0.00162) (0.00193) Δ R&D expenditure (first lag) 0.0683* 0.0674* 0.0689* 0.0670* (0.0286) (0.0295) (0.0313) (0.0292) Δ PP (first lag) 0.0830* 0.0894** 0.195*** (0.0328) (0.0327) (0.0448) Δ PP innovation propensity index (first lag) 0.0767* 0.0890* (0.0320) (0.0407) Import penetration on PP (first lag) 0.00136 (0.000911) Δ PP * import penetration (first lag) −0.00230** (0.000714) Δ PP innovation propensity index* import penetration (first lag) −0.000342 (0.00123) Observations 5385 5383 4795 4795 Note: Standard errors in parentheses. * P < 0.05, ** P < 0.01, *** P < 0.001. View Large Table 5. Fixed effects Poisson estimations—dependent variable: patents stock (LT-SI and SD industries) (1) (2) (3) (4) Export intensity (first lag) 0.00526*** 0.00496** 0.00456** 0.00321 (0.00153) (0.00151) (0.00162) (0.00193) Δ R&D expenditure (first lag) 0.0683* 0.0674* 0.0689* 0.0670* (0.0286) (0.0295) (0.0313) (0.0292) Δ PP (first lag) 0.0830* 0.0894** 0.195*** (0.0328) (0.0327) (0.0448) Δ PP innovation propensity index (first lag) 0.0767* 0.0890* (0.0320) (0.0407) Import penetration on PP (first lag) 0.00136 (0.000911) Δ PP * import penetration (first lag) −0.00230** (0.000714) Δ PP innovation propensity index* import penetration (first lag) −0.000342 (0.00123) Observations 5385 5383 4795 4795 (1) (2) (3) (4) Export intensity (first lag) 0.00526*** 0.00496** 0.00456** 0.00321 (0.00153) (0.00151) (0.00162) (0.00193) Δ R&D expenditure (first lag) 0.0683* 0.0674* 0.0689* 0.0670* (0.0286) (0.0295) (0.0313) (0.0292) Δ PP (first lag) 0.0830* 0.0894** 0.195*** (0.0328) (0.0327) (0.0448) Δ PP innovation propensity index (first lag) 0.0767* 0.0890* (0.0320) (0.0407) Import penetration on PP (first lag) 0.00136 (0.000911) Δ PP * import penetration (first lag) −0.00230** (0.000714) Δ PP innovation propensity index* import penetration (first lag) −0.000342 (0.00123) Observations 5385 5383 4795 4795 Note: Standard errors in parentheses. * P < 0.05, ** P < 0.01, *** P < 0.001. View Large Moving to the analysis of the moderating effect of import penetration, sign, and significance of the coefficient associated with the PP–import penetration interaction term (Column 4) points to a negative and significant moderating effect of import penetration in both groups of industries, though, again, this effect appears to be stronger in HT industries Finally, the importance of looking at industry heterogeneity is confirmed by the test on industries clustered according to the Peneder’s (2010) taxonomy as reported in Tables 6 and 7, which provide us with an even more defined picture The positive role of PP as innovation driver as well as the moderating effect of import penetration are confirmed when the medium HT industries cluster is considered—i.e., sectors belonging to the intermediate, intermediate-to-high, and high innovation intensity groups. Remarkably, the interaction between PP and the variable capturing the degree of import penetration on PP is negative and strongly statistically significant. Table 6. Fixed effects Poisson estimations—patents stock vs. PP, innovative PP, import penetration, interaction term, and controls (medium, med high, and HT industries) (1) (2) (3) (4) Export intensity (first lag) 0.00655*** 0.00580** 0.00528** 0.00456 (0.00187) (0.00180) (0.00184) (0.00252) Δ R&D expenditure (first lag) 0.0733 0.0671 0.191*** 0.177*** (0.0953) (0.0958) (0.0489) (0.0404) Δ PP (first lag) 0.174*** 0.168*** 0.378*** (0.0511) (0.0470) (0.0714) Δ PP innovation propensity index (first lag) 0.134*** 0.157** (0.0396) (0.0496) Import penetration on PP (first lag) 0.000806 (0.00134) Δ PP * import penetration (first lag) −0.00493*** (0.00113) PP Innovation propensity index* import penetration (first lag) −0.000520 (0.00118) Observations 4315 4315 3846 3846 (1) (2) (3) (4) Export intensity (first lag) 0.00655*** 0.00580** 0.00528** 0.00456 (0.00187) (0.00180) (0.00184) (0.00252) Δ R&D expenditure (first lag) 0.0733 0.0671 0.191*** 0.177*** (0.0953) (0.0958) (0.0489) (0.0404) Δ PP (first lag) 0.174*** 0.168*** 0.378*** (0.0511) (0.0470) (0.0714) Δ PP innovation propensity index (first lag) 0.134*** 0.157** (0.0396) (0.0496) Import penetration on PP (first lag) 0.000806 (0.00134) Δ PP * import penetration (first lag) −0.00493*** (0.00113) PP Innovation propensity index* import penetration (first lag) −0.000520 (0.00118) Observations 4315 4315 3846 3846 Note: Standard errors in parentheses. * P < 0.05, ** P < 0.01, *** P < 0.001. View Large Table 6. Fixed effects Poisson estimations—patents stock vs. PP, innovative PP, import penetration, interaction term, and controls (medium, med high, and HT industries) (1) (2) (3) (4) Export intensity (first lag) 0.00655*** 0.00580** 0.00528** 0.00456 (0.00187) (0.00180) (0.00184) (0.00252) Δ R&D expenditure (first lag) 0.0733 0.0671 0.191*** 0.177*** (0.0953) (0.0958) (0.0489) (0.0404) Δ PP (first lag) 0.174*** 0.168*** 0.378*** (0.0511) (0.0470) (0.0714) Δ PP innovation propensity index (first lag) 0.134*** 0.157** (0.0396) (0.0496) Import penetration on PP (first lag) 0.000806 (0.00134) Δ PP * import penetration (first lag) −0.00493*** (0.00113) PP Innovation propensity index* import penetration (first lag) −0.000520 (0.00118) Observations 4315 4315 3846 3846 (1) (2) (3) (4) Export intensity (first lag) 0.00655*** 0.00580** 0.00528** 0.00456 (0.00187) (0.00180) (0.00184) (0.00252) Δ R&D expenditure (first lag) 0.0733 0.0671 0.191*** 0.177*** (0.0953) (0.0958) (0.0489) (0.0404) Δ PP (first lag) 0.174*** 0.168*** 0.378*** (0.0511) (0.0470) (0.0714) Δ PP innovation propensity index (first lag) 0.134*** 0.157** (0.0396) (0.0496) Import penetration on PP (first lag) 0.000806 (0.00134) Δ PP * import penetration (first lag) −0.00493*** (0.00113) PP Innovation propensity index* import penetration (first lag) −0.000520 (0.00118) Observations 4315 4315 3846 3846 Note: Standard errors in parentheses. * P < 0.05, ** P < 0.01, *** P < 0.001. View Large Table 7. Fixed effects Poisson estimations—Patents stock vs. PP, innovative PP, import penetration, interaction term, and controls (medium-low and LT industries) (1) (2) (3) (4) Export intensity (first lag) 0.00993*** 0.00990*** 0.0102*** 0.00810** (0.00238) (0.00227) (0.00234) (0.00255) Δ R&D expenditure (first lag) 0.0386 0.0375 0.0367 0.0362 (0.0302) (0.0300) (0.0302) (0.0302) Δ PP (first lag) 0.0718 0.0787 0.159** (0.0399) (0.0415) (0.0567) Δ PP innovation propensity index (first lag) 0.0421 0.0475 (0.0367) (0.0472) Import penetration on PP (first lag) 0.00203 (0.00117) Δ PP * import penetration (first lag) −0.00130 (0.000811) PP Innovation propensity index* import penetration (first lag) −0.000275 (0.00127) Observations 2661 2659 2364 2364 (1) (2) (3) (4) Export intensity (first lag) 0.00993*** 0.00990*** 0.0102*** 0.00810** (0.00238) (0.00227) (0.00234) (0.00255) Δ R&D expenditure (first lag) 0.0386 0.0375 0.0367 0.0362 (0.0302) (0.0300) (0.0302) (0.0302) Δ PP (first lag) 0.0718 0.0787 0.159** (0.0399) (0.0415) (0.0567) Δ PP innovation propensity index (first lag) 0.0421 0.0475 (0.0367) (0.0472) Import penetration on PP (first lag) 0.00203 (0.00117) Δ PP * import penetration (first lag) −0.00130 (0.000811) PP Innovation propensity index* import penetration (first lag) −0.000275 (0.00127) Observations 2661 2659 2364 2364 Note: Standard errors in parentheses. * P< 0.05, ** P < 0.01, *** P < 0.001. View Large Table 7. Fixed effects Poisson estimations—Patents stock vs. PP, innovative PP, import penetration, interaction term, and controls (medium-low and LT industries) (1) (2) (3) (4) Export intensity (first lag) 0.00993*** 0.00990*** 0.0102*** 0.00810** (0.00238) (0.00227) (0.00234) (0.00255) Δ R&D expenditure (first lag) 0.0386 0.0375 0.0367 0.0362 (0.0302) (0.0300) (0.0302) (0.0302) Δ PP (first lag) 0.0718 0.0787 0.159** (0.0399) (0.0415) (0.0567) Δ PP innovation propensity index (first lag) 0.0421 0.0475 (0.0367) (0.0472) Import penetration on PP (first lag) 0.00203 (0.00117) Δ PP * import penetration (first lag) −0.00130 (0.000811) PP Innovation propensity index* import penetration (first lag) −0.000275 (0.00127) Observations 2661 2659 2364 2364 (1) (2) (3) (4) Export intensity (first lag) 0.00993*** 0.00990*** 0.0102*** 0.00810** (0.00238) (0.00227) (0.00234) (0.00255) Δ R&D expenditure (first lag) 0.0386 0.0375 0.0367 0.0362 (0.0302) (0.0300) (0.0302) (0.0302) Δ PP (first lag) 0.0718 0.0787 0.159** (0.0399) (0.0415) (0.0567) Δ PP innovation propensity index (first lag) 0.0421 0.0475 (0.0367) (0.0472) Import penetration on PP (first lag) 0.00203 (0.00117) Δ PP * import penetration (first lag) −0.00130 (0.000811) PP Innovation propensity index* import penetration (first lag) −0.000275 (0.00127) Observations 2661 2659 2364 2364 Note: Standard errors in parentheses. * P< 0.05, ** P < 0.01, *** P < 0.001. View Large The test on LT industries—i.e., sectors belonging to the medium-to-low and low-innovativeness Peneder’s (2010) categories—displays some interesting differences as compared to the medium HT cluster results. First, R&D loses its significance across all the estimations matching with the well-documented weak propensity toward formalized and intramural R&D activities in such sectors (see Peneder (2010) for a discussion on this point). Confirming the evidence reported for the industries belonging to the Pavitt’s SD and SI categories, in turn, PP seems to stimulate patenting activities even in LT sectors. Differently to what emerged in Table 6, in the LT case, no statistically significant association between the PP innovation propensity index and patenting is detected. Similarly, the moderating effect of import penetration detected before seems not to operate in this case. Overall, the pro-innovative effect of PP on industries’ patenting activities seems to be confirmed for both the HT and the LT clusters. This effect, however, seems to be stronger for industries belonging to medium and HT segments of the economy. On the other hand, the positive effect of PP on HT industries patenting activities seems to be significantly mitigated when the degree of import penetration on PP is high. Hence, the analysis of industry heterogeneity confirms the importance of looking at sectoral characteristics, as they might affect the mechanisms through which public demand and procurement strategies display their influence on innovation dynamics and economic performances. From an industrial policy point of view, this appears to be relevant in the choice between horizontal and selective measures related to PP strategies. In this respect, while the present study highlights the relevance of this issue, the structure of the data employed for the empirical analysis did not allow for a more detailed study of sectoral specificities, which however emerge as a crucial aspect to be further investigated. 5. Conclusions Nine years after the explosion of the 2008 crisis, industrial policy is coming back to the fore as a key action to promote growth and development. This revival interrupts 30 years or radical neglect, at least if industrial policy is intended as direct and selective interventions aimed at creating and steering new productions and markets (Chang and Andreoni, 2016; Mazzucato, 2016). A set of “old-style” industrial policy instruments is now back in the policy makers’ toolbox to spur firms, industries, and, ultimately, countries long-term growth. Among these instruments, a pivotal role is played by PP, largely recognized as a powerful driver of technological upgrading and industrial renewal (Edquist, 2015). This work fills a gap in the empirical literature by analyzing the impact of PP on industries’ patenting activities for 24 countries over the period 1995–2011. By enriching a largely qualitative literature, this analysis provides a quantitative account of the role of PP as an innovation driver. In addition, the relationship between PP and innovation is studied taking into consideration the potentially moderating effect of import penetration of PP. Finally, to capture the role of both institutional and technological heterogeneity, the analysis is replicated on different subsamples. By limiting the estimation to European industries, we first test whether institutional differences—i.e., in particular, differences concerning the degree of openness of domestic PP to foreign producers—dividing Europe and the other countries included in the sample—i.e., Japan, Korea, Russia, Turkey, and the United States—affect the magnitude and the shape of the investigated relations. Second, building on two distinct classifications, we separately analyze different groups of sectors so to verify if the effect of PP on patents changes depending on industries technological characteristics. The main results can be summarized as follows. First of all, PP emerges as a strong innovation driver. All across the estimations, a positive and strongly significant association between PP and industries’ patenting activities is detected. This is true for both the “regular” PP indicator as well as for the PP innovation propensity index—i.e., the index that weights industries’ PP by the country-level share of public expenditure devoted to R&D activities. Such association is robust and significant when controlling for other factors such as R&D expenditure and export propensity. The second finding regards the role of import penetration. Confirming the expectations, industries displaying a relatively stronger import penetration are characterized by a significant reduction in the pro-innovative effect of PP, as testified, in all the estimations, by the negative sign of the coefficient of the PP–import penetration interaction term. The test on European industries provides some additional insights. The positive impact of PP on patents emerging from the baseline estimation is confirmed. Such positive effect is detected irrespectively from the variable used to measure PP. The key difference, however, regards import penetration. In the European case, the moderating effect of import penetration operates for both the regular and for the “innovation oriented” PP differently from the baseline case where only the regular PP–import penetration interaction resulted negative and significant. This result appears to confirm the limited ability of European countries to take advantage of the preferential treatment toward domestic industries that can be attributed in the case of pre-commercial PP. From this point of view, these findings seem to support the recent President Macron’s proposal regarding the adoption of a “Buy Europe” Act analogous to the one launched in the United States after the 2008 crisis.20 Such proposal, in fact, recognizes that to maximize PP-related economic and technological benefits EU economies should properly calibrate procurement’s degree of openness. Finally, the relationship between PP and patenting activities has been scrutinized by looking at sectoral heterogeneity. Two distinct tests have been performed using the revised Pavitt taxonomy proposed by Bogliacino and Pianta (2016) and the classification developed by Peneder (2010). Both tests evidenced that the innovation enhancing effect of PP operates in all examined groups of industries, confirming the great potential of this instrument as an industrial policy tool. Interestingly, this result reveals the PP’s capacity to stimulate a process of technological upgrading even in sectors traditionally less prone to the introduction of innovations. Moreover, interesting differences among groups of industries emerge in the analysis, suggesting the importance of looking in more detail at industry specificities in future analyses. In terms of policy implications, this work provides a strong claim in favor of PP as a crucial instrument to spur innovativeness and to favor industrial renewal and technological upgrading. In this sense, this study strengthens the arguments put forth by the qualitative literature emphasizing the pro-innovative stance of PP. Moreover, the findings on the moderating effect of import penetration suggest that decisions concerning the degree of openness of PP should be carefully undertaken to avoid to harm the pro-innovative effects of PP. Finally, the results on sectoral heterogeneity suggest that the choices on the adoption of horizontal or selective strategies associated to PP should take into account the way through which sectoral specificities shape the relationships between public demand and innovation performance. Footnotes 1 Source: OECD—Government at a glance 2015 database (http://www.oecd.org/gov/government-at-a-glance-2015-database.htm) 2 Austria, Belgium, Czech Republic, Germany, Estonia, Finland, France, Greece, Hungary, Ireland, Italy, Japan, Korea, The Netherlands, Poland, Portugal, Romania, Russia, Sweden, Slovenia, Slovak Republic, Spain, Turkey, and the United States. 3 Input/Output data represent the best available approximation in standard statistics to obtain a sector-based breakdown of government procurement (Appelt and Galindo-Rueda, 2016). 4 According to Geroski and Walters (1995), the economic explanation of this phenomenon is twofold. First, markets have a limited ability to absorb new products in a given period. When demand expands, this capacity tends to grow, making the introduction of innovation more profitable. Second, appropriability problems are involved in innovative activities, and the time firms have to profit from innovation is often limited. Thus, innovations are more likely to appear in periods characterized by a growing demand trend. 5 The 1991’s US Congress Office of Technology report (1991) documents how discriminatory PP has been crucial to support the Japanese national supercomputer’s industry at the expenses of the US one. In this phase, strong discrimination in favor of local producers provided a substantial help to companies as Hitachi, Non electrical components (NEC), and Fujitsu in developing productive capacity and technological capabilities. 6 Patent data refer to European Patent Office (EPO) applications carried out by the considered countries between 1995 and 2009. 7 As Lybbert and Zolas (2014) point out, “While patent data often serve as useful proxies for technological change and diffusion, fully exploiting patent data in economic analyses would require that patents be linked to economic activity at a level of disaggregation that allows for different technological, industrial and spatial patterns. Such a detailed link between technological and economic activity would further improve our assessment of policies that aim to promote innovation, as well as assess the relationship between technological change and economic development” (pp. 530). In light of this, further attempts for industry-level linkages that associate patents and economic data based on the domain of goods and services they represent are strongly required for enrich the economic analysis 8 The Lybbert and Zolas’s (2014) imputation procedure is based on an ALP applied to patents’ content descriptions. Focusing on the correspondence between patent contents and sectors characteristics, this procedure allows matching, with a significant level of precision, patents identified according to the International Patent Classification (IPC) classification, and two-digit industries. The distribution of patents across industries obtained through this procedure is reported in the last column of Table A4. 9 For instance, the share of import penetration of the Italian motor vehicle industry is equal to the share of government purchases of motor vehicles produced abroad in the total PP directed toward the motor vehicle industry. The same holds for all the manufacturing sectors of each country. 10 The time horizon coincides with the one used for the econometric analysis. In the descriptive analysis we include all EU27 economies, while in the econometric one some EU countries are missing due to the lack of data on patents. The EU3 aggregate includes Germany, France, and Italy and is already used in Messerlin and Mirodout (2012) to investigate the dynamics of PP and import penetration on PP in Europe. 11 The Japanese economy is characterized by a relatively lower propensity toward production outsourcing and offshoring as compared to the EU and the United States. This element tends to relatively reduce the purchase of goods and services from abroad to meet public demand. 12 This is the natural starting point for an analysis of count data, but it may be biased by an excess in zeros and an overdispersion problem. In many applications, the model underestimates the probability of a zero count and low counts in general. In addition, the equidispersion assumption of the Poisson model, the equality of the conditional mean, and the conditional variance are commonly violated. Real variables are often overdispersed, that is, the variance exceeds the mean. The major disadvantage in the presence of overdispersion is that estimates are inefficient with the standard errors biased downward, resulting in spuriously large z-values and small P-values (Cameron and Trivedi, 1998). In these cases, the NBRM, which addresses the failure of the PRM by introducing unobserved heterogeneity across the Poisson means, could be used. In our case, however, the reduced number of zeros among the dependent variables observations of the dependent variable and the results of tests conducted to this purpose are in favor of the use of the PRM. 13 The Poisson fixed-effects estimator conditions the probability of the counts for each group on the sum of the counts for the group. The maximum likelihood method is used to estimate the model parameters. 14 We thank an anonymous referee for suggesting these additional tests. 15 For a discussion on the VIF test and on thresholds to be used to check covariates mutlicollinearity, see Wooldridge (2013: 95–98). In tune with the findings in Table A1, the highest VIF values are registered for export intensity and degree of import penetration that are both closely related to industries’ degree of openness. 16 The path-dependent nature of R&D is linked to the development of knowledge and the unfolding of technological change. Both processes are shaped by the paradigm and trajectory of technological change, making the process of search eminently localized (Atkinson and Stiglitz, 1969; Dosi, 1982; Nelson and Winter, 1982). The path-dependent nature of technological change has been pointed out by a recent set contributions focusing on innovation persistence (Antonelli et al., 2012; Triguero-Cano and Corcoles-Gonzales, 2013). 17 In this way, we are able to account for potential R&D–PP relationships obtaining a cleaner R&D measure. 18 The countries included in this subsample are Austria (AT), Belgium (BE), Czech Republic (CZE), Germany (GER), Spain (SP), Finland (FIN), France (FR), the United Kingdom (UK), Greece (GRC), Hungary (HUN), Ireland (IRE), Italy (IT), The Netherlands (NE), Poland (PO), Portugal (PT), Romania (RO), Slovakia (SK), Slovenia (SV), and Sweden (SWE) covering almost all the EU27 economy. 19 Using both the Bogliacino and Pianta (2016) revised Pavitt taxonomy and the Peneder’s (2010) classification allows exploring in depth the role of technological heterogeneity accounting for its multidimensional nature. The Bogliacino and Pianta’s (2016) revised Pavitt taxonomy characterizes industries focusing on importance and relative intensity of R&D activities. The Peneder’s (2010) classification, in turn, considers both firm-level heterogeneities as well as sectoral contingencies building upon a larger set of factors: technological opportunities, appropriability conditions and the cumulativeness of knowledge. We thank an anonymous referee for providing suggestions on this point. The two taxonomies are illustrated in the Appendix. 20 For a synthetic description of the Macron’s Buy Europe proposal, see the Financial Times article by Cahssany (2017). References Andreoni A. ( 2016 ), ‘Varieties of industrial policy: models, packages and transformation cycles,’ in Noman A. , Stiglitz J. (eds), Efficiency, Finance and Varieties of Industrial Policy: Guiding Resources, Learning, and Technology for Sustained Growth . Columbia University Press : New York . Google Preview WorldCat COPAC Andreoni A. , Chang H. J. ( 2016 ), ‘ Industrial policy and the future of manufacturing ,’ Economia e Politica Industriale: Journal of Industrial and Business Economics , 43 ( 4 ), 491 – 502 . Google Scholar Crossref Search ADS WorldCat Andreoni A. , Scazzieri R. ( 2014 ), ‘ Triggers of change: structural trajectories and production dynamics ,’ Cambridge Journal of Economics , 38 ( 6 ), 1391 – 1408 . Google Scholar Crossref Search ADS WorldCat Antonelli C. ( 1999 ), The Microdynamics of Technological Change , Routledge : London . Google Preview WorldCat COPAC Antonelli C. , Crespi F. , Scellato G. ( 2012 ), ‘ Inside innovation persistence: new evidence from Italian micro-data ,’ Structural Change and Economic Dynamics , 23 ( 4 ), 341 – 353 . Google Scholar Crossref Search ADS WorldCat Appelt S. , Galindo-Rueda F. ( 2016 ), ‘Measuring the Link between Public Procurement and Innovation,’ OECD Science, Technology and Industry Working Papers, 2016/03. OECD Publishing: Paris. http://dx.doi.org/10.1787/5jlvc7sl1w7h-en Archibugi D. , Pianta M. ( 1992 ), The Technological Specialization of Advanced Countries: A Report to the EEC on International Science and Technology Activities . Springer Science & Business Media : Germany . Google Preview WorldCat COPAC Arrow K. ( 1962 ), ‘ The economic implications of learning by doing ,’ The Review of Economic Studies , 29 ( 3 ), 155 – 173 . Google Scholar Crossref Search ADS WorldCat Atkinson A. , Stiglitz J. ( 1969 ), ‘ A new view of technological change ,’ The Economic Journal , 79 ( 315 ), 573 – 578 . Google Scholar Crossref Search ADS WorldCat Beviglia-Zampetti A. ( 1997 ), The UNCITRAL Model Law on Procurement of Goods, Construction and Services . Hoekman and Mavroidis . Google Preview WorldCat COPAC Block F. , Keller M. R. . ( 2009 ), State of Innovation . Paradigm : Boulder . Google Preview WorldCat COPAC Bogliacino F. , Pianta M. ( 2012 ), ‘ Profits, R&D, and innovation—a model and a test ,’ Industrial and Corporate Change , 22 ( 3 ), 649 – 678 . Google Scholar Crossref Search ADS WorldCat Bogliacino F. , Pianta M. ( 2016 ), ‘ The Pavitt Taxonomy, revisited: patterns of innovation in manufacturing and services ,’ Economia Politica , 33 ( 2 ), 153 – 180 . Google Scholar Crossref Search ADS WorldCat Bresnahan T. , Greenstein S. ( 2001 ), ‘ The economic contribution of information technology: towards comparative and user studies ,’ Journal of Evolutionary Economics , 11 ( 1 ), 95 – 118 . Google Scholar Crossref Search ADS WorldCat Brouwer E. , Kleinknecht A. ( 1999 ), ‘ Keynes-plus? Effective demand and changes in firm-level R&D ,’ Cambridge Journal of Economics , 23 ( 3 ), 385 – 391 . Google Scholar Crossref Search ADS WorldCat Brulhart M. , Trionfetti F. ( 2004 ), ‘ Public expenditure, international specialisation and agglomeration ,’ European Economic Review , 48 ( 4 ), 851 – 881 . Google Scholar Crossref Search ADS WorldCat Cabral L. , Cozzi G. , Denicolo V. , Spagnolo G. , Zanza M. ( 2006 ), ‘Procuring innovations,’ in Dimitri N. , Piga G. , Spagnolo G. (eds) Handbook of Procurement , Cambridge University Press : Cambridge, MA , pp. 483 – 530 . Google Preview WorldCat COPAC Cameron A. C. , Trivedi P. K. ( 1998 ), Regression Analysis of Count Data. Econometric Society Monograph No.30 , Cambridge University Press : Cambridge, MA . Google Preview WorldCat COPAC Cave J. , Frinking E. ( 2003 ), ‘Public procurement and R&D: short analysis of the potential and practices,’ in Gavigan J. P. (ed.) Public Procurement and R&D: A JRC/IPTS-ESTO Fast Track Working Paper . European Commission Joint Research Centre–Institute for Prospective Technological Studies, European Science and Technology Observatory : Paris , pp. 11 – 44 . Google Preview WorldCat COPAC Chang H.-J. ( 1994 ), The Political Economy of Industrial Policy . Macmillan : London and Basingstoke . Google Preview WorldCat COPAC Chang H.-J. , Andreoni A. ( 2016 ), ‘Industrial policy in a changing world: basic principles, neglected issues and new challenges,’ in Cambridge Journal of Economics of 40years Conference. Springer International Publishing: Cambridge. Cahssany A. S. ( 2017 ), ‘Macron wants tougher EU on trade and foreign investment,’ Financial Times, May, 11, https://www.ft.com/content/38e98f94-359b-11e7-99bd-13beb0903fa3 (15 June 2017, date last accessed). Cirillo V. , Guarascio D. ( 2015 ), ‘ Jobs and competitiveness in a polarised Europe ,’ Intereconomics , 50 ( 3 ), 156 – 160 . Google Scholar Crossref Search ADS WorldCat Cohen E. ( 2007 ), ‘ Industrial policies in France: the old and the new ,’ Journal of Industry, Competition and Trade , 7 ( 3–4 ), 213 – 227 . Google Scholar Crossref Search ADS WorldCat Crespi F. , Pianta M. ( 2007 ), ‘ Innovation and demand in European industries ,’ Economia Politica-Journal of Analytical and Institutional Economics , 24 ( 1 ), 79 – 112 . WorldCat Crespi F. , Quatraro F. ( 2013 ), ‘ Systemic technology policies: issues and instruments ,’ Technological Forecasting and Social Change , 80 , 1447 – 1449 . Google Scholar Crossref Search ADS WorldCat Dawid H. , Pellegrino G. , Vivarelli M. ( 2017 ), ‘Demand and innovation: theory and evidence, No 92,’ GLO Discussion Paper Series, Global Labor Organization (GLO). Dosi G. ( 1982 ), ‘ Technological paradigms and technological trajectories: a suggested interpretations of the determinants and directions of technical change ,’ Research Policy , 11 , 147 – 162 . Google Scholar Crossref Search ADS WorldCat Dosi G. , Fagiolo G. , Roventini A. ( 2010 ), ‘ Schumpeter meeting Keynes: a policy-friendly model of endogenous growth and business cycles ,’ Journal of Economic Dynamics and Control , 34 ( 9 ), 1748 – 1767 . Google Scholar Crossref Search ADS WorldCat Edler J. , Georghiou L. ( 2007 ), ‘ Public procurement and innovation - resurrecting the demand side ,’ Research Policy , 36 ( 7 ), 949 – 963 . Google Scholar Crossref Search ADS WorldCat Edquist C. , Hommen L. ( 2000 ), ‘Public technology procurement and innovation theory,’ in Edquist C. , Hommen L. , e Tsipouri L. (eds), Public Technology Procurement and Innovation . Kluwer Academic : The Netherlands . Google Preview WorldCat COPAC Edquist C. ( 2015 ), ‘Innovation-related public procurement as a demand-oriented innovation policy instrument,’ Circle Papers in Innovation Studies, Lund University, n. 2015/28. Edquist C. , Vonortas N. S. , Zabala-Iturriagagoitia J. M. , Edler J. (eds) ( 2015 ), Public Procurement for Innovation . Edward Elgar Publishing : Cheltenham . Google Preview WorldCat COPAC ECWG (European Commission Working Group) ( 2006 ), ‘Pre-commercial procurement of innovation: a missing link in the European innovation cycle,’ Working Group Report, http://ec.europa.eu/information_society/research/key_docs/documents/procurement.pdf (16 December 2007, date last accessed). European Commission ( 2011 ), Commission staff working paper: impact and effectiveness of EU public procurement legislation . European Commission : Bruxelles . WorldCat COPAC Geroski P. ( 1990 ), ‘ Procurement policy as a tool of industrial policy ,’ International Review of Applied Economics , 4 ( 2 ), 182 – 198 . Google Scholar Crossref Search ADS WorldCat Geroski P. A. , Walters C. F. ( 1995 ), ‘ Innovative activity over the business cycle ,’ The Economic Journal , 105 ( 431 ), 916 – 928 . Google Scholar Crossref Search ADS WorldCat Georghiou L. , Edler J. , Uyarra E. , Yeow J. ( 2014 ), ‘ Policy instruments for public procurement of innovation: choice, design and assessment, ’ Technological Forecasting and Social Change , 86 , 1 – 12 . Google Scholar Crossref Search ADS WorldCat Griliches Z. ( 1990 ), ‘ Patent statistics as economic indicators: a survey ,’ Journal of Economic Literature , 28 , 1661 – 1707 . WorldCat Gereffi G. ( 1999 ), ‘ International trade and industrial upgrading in the apparel commodity chain ,’ Journal of International Economics , 48 ( 1 ), 37 – 70 . Google Scholar Crossref Search ADS WorldCat Gereffi G. , Humphrey J. , Kaplinsky R. , Sturgeon T. J. ( 2001 ), ‘ The value of value chains ,’ IDS Bulletin Special , 32 ( 3 ), 1. Google Scholar Crossref Search ADS WorldCat Guarascio D. , Pianta M. , Lucchese M. , Bogliacino F. ( 2015 ), ‘ Business cycles, technology and exports ,’ Economia Politica , 32 ( 2 ), 167 – 200 . Google Scholar Crossref Search ADS WorldCat Guarascio D. , Pianta M. , Bogliacino F. ( 2016 ), ‘ Export, R&D and new products. A model and a test on European industries ,’ Journal of Evolutionary Economics , 26 ( 4 ), 869 – 905 . Google Scholar Crossref Search ADS WorldCat Guarascio D. , Pianta M. ( 2017 ), ‘ The gains from technology: new products, exports and profits ,’ Economics of Innovation and New Technology , 26 ( 8 ), 779 – 804 . Google Scholar Crossref Search ADS WorldCat Jaffe A. B. , Trajtenberg M. ( 2004 ), ‘ Patents, citations, and innovations: a window on the knowledge economy ,’ Journal of Economic Literature , 42 , 1158 – 1160 . WorldCat Kaldor N. ( 1957 ), ‘ A model of economic growth ,’ The economic journal , 67 ( 268 ), 591 – 624 . Google Scholar Crossref Search ADS WorldCat Kaldor N. ( 1966 ), Causes of the slow rate of economic growth of the United Kingdom: an inaugural lecture . Cambridge University Press : Cambridge . Google Preview WorldCat COPAC Kaldor N. ( 1981 ), ‘ The role of increasing returns, technical progress and cumulative causation in the theory of international trade and economic growth ,’ Economie Appliquee , 34 , 593 – 617 . WorldCat Kattel R. , Lember V. ( 2010 ), ‘ Public Procurement as an industrial policy tool: an Option for De-veloping countries? ’ Journal of Public Procurement , 10 ( 3 ), 368 – 404 . Google Scholar Crossref Search ADS WorldCat Kleinknecht A. , Verspagen B. ( 1990 ), ‘ Demand and innovation: schmookler re-examined ,’ Research Policy , 19 ( 4 ), 387 – 394 . Google Scholar Crossref Search ADS WorldCat Kleinknecht A. (ed.) ( 1996 ), Determinants of Innovation . Springer : London . Google Preview WorldCat COPAC Lowinger T. C. ( 1976 ), ‘ Discrimination in government procurement of foreign goods in the US and Western Europe ,’ Southern Economic Journal , 42 ( 3 ), 451 – 460 . Google Scholar Crossref Search ADS WorldCat Lucchese M. , Nascia L. , Pianta M. ( 2016 ), ‘ Industrial policy and technology in Italy ,’ Economia e Politica Industriale: Journal of Industrial and Business Economics , 43 ( 3 ), 233 – 260 . Google Scholar Crossref Search ADS WorldCat Lybbert T. J. , Zolas N. J. ( 2014 ), ‘ Getting patents and economic data to speak to each other: an algorithmic links with probabilities approach for joint analyses of patenting and economic activity ,’ Research Policy , 43 ( 3 ), 530 - 542 . Google Scholar Crossref Search ADS WorldCat Malerba F. ( 2007 ), ‘ Innovation and the dynamics and evolution of industries: progress and challenges ,’ International Journal of Industrial Organization , 25 ( 4 ), 675 – 699 . Google Scholar Crossref Search ADS WorldCat Mazzucato M. ( 2013 ), The Entrepreneurial State: Debunking the Public vs. Private Myth in Risk and Innovation . Anthem : London . Google Preview WorldCat COPAC Mazzucato M. , Cimoli M. , Dosi G. , Stiglitz J. , Landesmann M. , Pianta M. , Walz R. , Page T. ( 2015 ), ‘ Which industrial policy does Europe need? ’ Intereconomics: Review of European Economic Policy , 50 ( 3 ), 120 – 155 . Google Scholar Crossref Search ADS WorldCat Mazzucato M. ( 2016 ), ‘ From market fixing to market-creating: a new framework for innovation policy ,’ Industry and Innovation , 23 ( 2 ), 140 – 156 . Google Scholar Crossref Search ADS WorldCat Messerlin P. , Mirodout S. ( 2012 ), ‘EU public procurement markets: how open are they?,’ SciencesPo, Groupe de Economie Mondiale, Policy Brief, August, 2012. Miyagiwa K. ( 1991 ), ‘ Oligopoly and discriminatory government procurement policy ,’ The American Economic Review , 81 ( 5 ), 1320 – 1328 . WorldCat Mowery D. , Rosenberg N. ( 1979 ), ‘ The influence of market demand upon innovation: a critical review of some recent empirical studies ,’ Research Policy , 8 ( 2 ), 102 – 153 . Google Scholar Crossref Search ADS WorldCat Nelson R. , Winter S. ( 1982 ), An Evolutionary Theory of Economic Change . The Belknap Press of Harvard University Press : Cambridge . Google Preview WorldCat COPAC OECD ( 2005 ), Oslo Manual: Guidelines for Collecting and Interpreting Innovation Data , 3 rd edn. OECD Publishing : Paris . WorldCat COPAC OECD ( 2011 ), Demand-Side Innovation Policies . OECD Publishing : Paris . WorldCat COPAC Okimoto D. I. ( 1989 ), Between MITI and the Market: Japanese Industrial Policy for High Technology . Stanford University Press : California . Google Preview WorldCat COPAC Pasinetti L. ( 1981 ), Structural Change and Economic Growth . Cambridge University Press : Cambridge . Google Preview WorldCat COPAC Peneder M. ( 2010 ), ‘ Technological regimes and the variety of innovation behaviour: creating integrated taxonomies of firms and sectors ,’ Research Policy , 39 ( 3 ), 323 – 334 . Google Scholar Crossref Search ADS WorldCat Peneder M. ( 2017 ), ‘ Competitiveness and industrial policy: from rationalities of failure towards the ability to evolve ,’ Cambridge Journal of Economics , 41 ( 3 ), 829 – 858 . WorldCat Pianta M. ( 2015 ), ‘ What is to be produced? The case for industrial policy ,’ Intereconomics , 50 ( 3 ), 140 – 146 . WorldCat Piva M. , Vivarelli M. ( 2007 ), ‘ Is demand-pulled innovation equally important in different groups of firms? ’ Cambridge Journal of Economics , 31 ( 5 ), 691 – 710 . Google Scholar Crossref Search ADS WorldCat Porter M. E. ( 1990 ), The Competitive Advantage of Nations . Free Press : New York . Google Preview WorldCat COPAC Rickard S. J. , Kono D. Y. ( 2014 ), ‘ Think globally, buy locally: international agreements and government procurement ,’ The Review of International Organizations , 9 ( 3 ), 333 – 352 . Google Scholar Crossref Search ADS WorldCat Rothwell R. ( 1983 ), ‘Creating a regional innovation-oriented infrastructure: the role of public procurement. Paper Submitted to the Conference on Public Procurement and Regional Policy. University of Neuchâtel (31.19-1.11.1983). Rothwell R. , Zegveld W. ( 1981 ), Industrial Innovation and Public Policy: Preparing for the 1980s and the 1990s (No. 42) . Greenwood Pub Group : California . Google Preview WorldCat COPAC Ruttan V. W. ( 2006 ), Is War Necessary for Economic Growth? Military Procurement and Technology Development . Oxford University Press : New York . Google Preview WorldCat COPAC Schmookler J. ( 1966 ), Invention and Economic Growth . Harvard University Press : Cambridge, MA . Google Preview WorldCat COPAC Simonazzi A. , Ginzburg A. , Nocella G. ( 2013 ), ‘ Economic relations between Germany and Southern Europe ,’ Cambridge Journal of Economics , 37 ( 3 ), 653 – 675 . Google Scholar Crossref Search ADS WorldCat Singh A. ( 2002 ), ‘Competition and competition policy in emerging markets: international and developmental dimensions,’ in Arestis P. , Thirlwall A. P. (eds), Growth and Economic Development: Essays in Honour of A. P. Thirlwall . Edward Elgar : London . Google Preview WorldCat COPAC Stiglitz J. E. , Greenwald B. C. ( 2014 ), Creating a Learning Society: A New Approach to Growth, Development, and Social Progress . Columbia University Press : New York . Google Preview WorldCat COPAC Timmer M. P. , Dietzenbacher E. , Los B. , Stehrer R. , de Vries G. J. .( 2015 ), ‘ An illustrated user guide to the World Input–Output Database: the case of global automotive production ,’ Review of International Economics , 23 ( 3 ), 575 – 605 . Google Scholar Crossref Search ADS WorldCat Triguero-Cano A. , Corcoles-Gonzales D. ( 2013 ), ‘ Understanding the innovation: an analysis of persistence for Spanish manufacturing firms ,’ Research Policy , 42 , 340 – 352 . Google Scholar Crossref Search ADS WorldCat Trionfetti F. ( 2000 ), ‘ Discriminatory public procurement and international trade ,’ World Economy , 23 ( 1 ), 57 – 76 . Google Scholar Crossref Search ADS WorldCat United States Congress Office of Technology Assessment ( 1991 ), Competing Economies: America, Europe, and the Pacific Rim , Vol. 22. US Government Printing Office : Washington . WorldCat COPAC Uyarra E. , Flanagan K. ( 2010 ), ‘ Understanding the innovation impacts of public procurement ,’ European Planning Studies , 18 ( 1 ), 123 – 143 . Google Scholar Crossref Search ADS WorldCat Von Hippel E. ( 1986 ), ‘ Lead users: a source of novel product concepts ,’ Management Science , 32 ( 7 ), 791 – 805 . Google Scholar Crossref Search ADS WorldCat Wade R. ( 1990 ), Governing the Market: Economic Theory and the Role of Government in East Asia’s Industrialization . Princeton University Press : Princeton, NJ . Google Preview WorldCat COPAC Wade R. ( 2017 ), ‘ The American paradox: ideology of free markets and the hidden practice of directional thrust ,’ Cambridge Journal of Economics , 41 ( 3 ), 859 – 880 . Google Scholar Crossref Search ADS WorldCat Weiss L. , Thurbon E. ( 2006 ), ‘ The business of buying American: public procurement as trade strategy in the USA ,’ Review of International Political Economy , 13 ( 5 ), 701 – 724 . Google Scholar Crossref Search ADS WorldCat Wooldridge J. M. ( 2013 ), Introductory Econometrics: A Modern Approach . 5 th edn, South-Western : Mason, OH . Google Preview WorldCat COPAC Appendix Matching patent data and industrial classifications The concordance system proposed by Lybbert and Zolas (2014) rely on data mining methods and probabilistic matching allowing to create direct linkages between patent data and a variety of economic classification schemes. Using the Algorithmic Links with Probabilities (ALP) approach, country-level information on patents is examined identifying relevant key words extracted from industry classification descriptions. The search is performed using patent titles and abstracts. Then, patents are tabulated by IPC code, and frequency matches between the industry-level classification—i.e., NACE—and IPC classifications are performed. By analyzing these frequencies, IPC and industrial classification are finally matched. The patents assigned for each year to the specific sector/country combination are used to compute the patent stock as a measure of the installed technological capability at the sector level. The “depreciation” of the stock is then taken into account by applying a standard decay rate. Formally, the procedure adopted to build the patent stock can be expressed as follows (4): KPATi,t=∑s=1tPATi,se-μt-s (4) where PATi,s represents the number of patents applied in country i in year s where s represents an index of years up to and including year t, whereas μ is the decay rate, here assumed as a standard 15% value. Robustness check Table A1. Baseline model—test on multicollinearity of R&D and exports (1) (2) Export intensity (first lag) 0.00795** (0.00270) Δ R&D expenditure (first lag) 0.102** (0.0323) Δ PP (first lag) 0.270*** 0.276*** (0.0541) (0.0572) Δ PP innovation propensity index (first lag) 0.130** 0.125** (0.0426) (0.0414) Import penetration on PP (first lag) −0.000118 0.00274*** (0.00125) (0.000738) Δ PP * import penetration (first lag) −0.00322*** −0.00288*** (0.000796) (0.000854) PP innovation propensity index* import penetration (first lag) −0.00111 −0.000172 (0.000962) (0.00120) Observations 8768 8768 (1) (2) Export intensity (first lag) 0.00795** (0.00270) Δ R&D expenditure (first lag) 0.102** (0.0323) Δ PP (first lag) 0.270*** 0.276*** (0.0541) (0.0572) Δ PP innovation propensity index (first lag) 0.130** 0.125** (0.0426) (0.0414) Import penetration on PP (first lag) −0.000118 0.00274*** (0.00125) (0.000738) Δ PP * import penetration (first lag) −0.00322*** −0.00288*** (0.000796) (0.000854) PP innovation propensity index* import penetration (first lag) −0.00111 −0.000172 (0.000962) (0.00120) Observations 8768 8768 Note: Standard errors in parentheses. * P < 0.05, ** P < 0.01, *** P < 0.001. View Large Table A1. Baseline model—test on multicollinearity of R&D and exports (1) (2) Export intensity (first lag) 0.00795** (0.00270) Δ R&D expenditure (first lag) 0.102** (0.0323) Δ PP (first lag) 0.270*** 0.276*** (0.0541) (0.0572) Δ PP innovation propensity index (first lag) 0.130** 0.125** (0.0426) (0.0414) Import penetration on PP (first lag) −0.000118 0.00274*** (0.00125) (0.000738) Δ PP * import penetration (first lag) −0.00322*** −0.00288*** (0.000796) (0.000854) PP innovation propensity index* import penetration (first lag) −0.00111 −0.000172 (0.000962) (0.00120) Observations 8768 8768 (1) (2) Export intensity (first lag) 0.00795** (0.00270) Δ R&D expenditure (first lag) 0.102** (0.0323) Δ PP (first lag) 0.270*** 0.276*** (0.0541) (0.0572) Δ PP innovation propensity index (first lag) 0.130** 0.125** (0.0426) (0.0414) Import penetration on PP (first lag) −0.000118 0.00274*** (0.00125) (0.000738) Δ PP * import penetration (first lag) −0.00322*** −0.00288*** (0.000796) (0.000854) PP innovation propensity index* import penetration (first lag) −0.00111 −0.000172 (0.000962) (0.00120) Observations 8768 8768 Note: Standard errors in parentheses. * P < 0.05, ** P < 0.01, *** P < 0.001. View Large Table A2. VIF—test on baseline model Variable VIF PP 1.03 Innovation propensity index 1.05 Exports 4.5 R&D expenditure 1.01 Import penetration share 4.7 Variable VIF PP 1.03 Innovation propensity index 1.05 Exports 4.5 R&D expenditure 1.01 Import penetration share 4.7 View Large Table A2. VIF—test on baseline model Variable VIF PP 1.03 Innovation propensity index 1.05 Exports 4.5 R&D expenditure 1.01 Import penetration share 4.7 Variable VIF PP 1.03 Innovation propensity index 1.05 Exports 4.5 R&D expenditure 1.01 Import penetration share 4.7 View Large Table A3. 2SLS estimation (first step)—robustness check on R&D (change in R&D vs. lagged R&D and controls) (1) Δ R&D expenditure (first lag) 0.151*** (0.00928) Export intensity (first lag) 0.000429*** (0.000113) Δ PP (first lag) 0.0418*** (0.0114) Δ PP innovation propensity index (first lag) 0.0298** (0.0110) Import penetration on PP (first lag) −0.000124 (0.000114) Δ PP * import penetration (first lag) −0.000355 (0.000265) PP innovation propensity index* import penetration (first lag) −0.000472 (0.000262) Observations 6210 (1) Δ R&D expenditure (first lag) 0.151*** (0.00928) Export intensity (first lag) 0.000429*** (0.000113) Δ PP (first lag) 0.0418*** (0.0114) Δ PP innovation propensity index (first lag) 0.0298** (0.0110) Import penetration on PP (first lag) −0.000124 (0.000114) Δ PP * import penetration (first lag) −0.000355 (0.000265) PP innovation propensity index* import penetration (first lag) −0.000472 (0.000262) Observations 6210 Standard errors in parentheses. * P < 0.05, ** P < 0.01, *** P < 0.001. View Large Table A3. 2SLS estimation (first step)—robustness check on R&D (change in R&D vs. lagged R&D and controls) (1) Δ R&D expenditure (first lag) 0.151*** (0.00928) Export intensity (first lag) 0.000429*** (0.000113) Δ PP (first lag) 0.0418*** (0.0114) Δ PP innovation propensity index (first lag) 0.0298** (0.0110) Import penetration on PP (first lag) −0.000124 (0.000114) Δ PP * import penetration (first lag) −0.000355 (0.000265) PP innovation propensity index* import penetration (first lag) −0.000472 (0.000262) Observations 6210 (1) Δ R&D expenditure (first lag) 0.151*** (0.00928) Export intensity (first lag) 0.000429*** (0.000113) Δ PP (first lag) 0.0418*** (0.0114) Δ PP innovation propensity index (first lag) 0.0298** (0.0110) Import penetration on PP (first lag) −0.000124 (0.000114) Δ PP * import penetration (first lag) −0.000355 (0.000265) PP innovation propensity index* import penetration (first lag) −0.000472 (0.000262) Observations 6210 Standard errors in parentheses. * P < 0.05, ** P < 0.01, *** P < 0.001. View Large Table A4. 2SLS estimation (second step)—robustness check on R&D (1) (2) (3) (4) Δ R&D expenditure (predicted) 0.836** 0.565* 0.580* 0.725*** (0.255) (0.274) (0.278) (0.218) Export intensity (first lag) 0.00764*** 0.00719*** 0.00664*** 0.00533** (0.00167) (0.00164) (0.00164) (0.00204) Δ PP (first lag) 0.109** 0.113** 0.252*** (0.0364) (0.0347) (0.0542) Δ PP innovation propensity index (first lag) 0.0879** 0.109** (0.0295) (0.0389) Import penetration on PP (first lag) 0.00123 (0.000941) Δ PP * import penetration (first lag) −0.00309*** (0.000772) PP innovation propensity index* import penetration (first lag) −0.000498 (0.00105) Observations 6210 6210 6210 6210 (1) (2) (3) (4) Δ R&D expenditure (predicted) 0.836** 0.565* 0.580* 0.725*** (0.255) (0.274) (0.278) (0.218) Export intensity (first lag) 0.00764*** 0.00719*** 0.00664*** 0.00533** (0.00167) (0.00164) (0.00164) (0.00204) Δ PP (first lag) 0.109** 0.113** 0.252*** (0.0364) (0.0347) (0.0542) Δ PP innovation propensity index (first lag) 0.0879** 0.109** (0.0295) (0.0389) Import penetration on PP (first lag) 0.00123 (0.000941) Δ PP * import penetration (first lag) −0.00309*** (0.000772) PP innovation propensity index* import penetration (first lag) −0.000498 (0.00105) Observations 6210 6210 6210 6210 Note: Standard errors in parentheses. * P < 0.05, ** P < 0.01, *** P < 0.001. View Large Table A4. 2SLS estimation (second step)—robustness check on R&D (1) (2) (3) (4) Δ R&D expenditure (predicted) 0.836** 0.565* 0.580* 0.725*** (0.255) (0.274) (0.278) (0.218) Export intensity (first lag) 0.00764*** 0.00719*** 0.00664*** 0.00533** (0.00167) (0.00164) (0.00164) (0.00204) Δ PP (first lag) 0.109** 0.113** 0.252*** (0.0364) (0.0347) (0.0542) Δ PP innovation propensity index (first lag) 0.0879** 0.109** (0.0295) (0.0389) Import penetration on PP (first lag) 0.00123 (0.000941) Δ PP * import penetration (first lag) −0.00309*** (0.000772) PP innovation propensity index* import penetration (first lag) −0.000498 (0.00105) Observations 6210 6210 6210 6210 (1) (2) (3) (4) Δ R&D expenditure (predicted) 0.836** 0.565* 0.580* 0.725*** (0.255) (0.274) (0.278) (0.218) Export intensity (first lag) 0.00764*** 0.00719*** 0.00664*** 0.00533** (0.00167) (0.00164) (0.00164) (0.00204) Δ PP (first lag) 0.109** 0.113** 0.252*** (0.0364) (0.0347) (0.0542) Δ PP innovation propensity index (first lag) 0.0879** 0.109** (0.0295) (0.0389) Import penetration on PP (first lag) 0.00123 (0.000941) Δ PP * import penetration (first lag) −0.00309*** (0.000772) PP innovation propensity index* import penetration (first lag) −0.000498 (0.00105) Observations 6210 6210 6210 6210 Note: Standard errors in parentheses. * P < 0.05, ** P < 0.01, *** P < 0.001. View Large The taxonomies used to analyze the role of industry-level technological heterogeneity The separate test on high- and low-tech (LT) sectors is performed relying on two distinct taxonomies: the revised Pavitt taxonomy (Bogliacino and Pianta, 2016) and the classification proposed by Peneder (2010). The following table reports the list of sectors included in the analysis and the detail of each taxonomy. More specifically, we define as high-tech (HT) those industries belonging to: the Science Based (SB) and Specialized Suppliers revised Pavitt’s categories; and the Peneder’s (2010) medium, medium-high, and HT groups. On the other hand, we define as low-tech (LT) those industries belonging to: the Supplier Dominated (SD) and Scale Intensive (SI) revised Pavitt’s categories; and the Peneder’s (2010) medium, med low, and LT groups. In spite of a partial overlapping, these taxonomies build upon different assumptions to distinguish industries according to their technological characteristics. The revised Pavitt taxonomy rank sectors focusing, in particular, on their relative R&D intensity. The Peneder’s (2010) classification, in turn, takes into account both R&D and additional variables—such as the share of firms pursuing innovation opportunities through the acquisition of new technology, the appropriability conditions, and the role of internally accumulated knowledge—allowing to capture technological heterogeneities even among sectors where formalized R&D activities are relatively less diffused. Therefore, replicating the analysis of the PP–innovation nexus relying separately on each taxonomy helps capturing more precisely the role of industry-level technological heterogeneities in shaping the relation under scrutiny. Table A5. The classification of sectors adopted for the HT-LT test Nr. Sectors (NACE Rev.1) NACE codes Pavitt classification Peneder classification Patents (% over the sample total) Manufacturing 1 Food products, beverages, and tobacco 15–16 SD Med low-tech 2.42 2 Textiles 17 SD Med high-tech 5.02 3 Wearing apparel, dressing and dyeing 18 SD Low-tech 5.04 4 Leather, leather products, and footwear 19 SD Low-tech 0.20 5 Wood and products of wood and cork 20 SD Medium 0.79 6 Pulp, paper, and paper products 21 SI Medium 2.34 7 Printing and publishing 22 SI Med low-tech 1.26 8 Chemical and chemical products 24 SB Med high-tech 12.84 9 Rubber and plastic products 25 SI Med high-tech 3.71 10 Other nonmetallic mineral products 26 SI Med high-tech 4.84 11 Basic metals 27 SI Med high-tech 7.21 12 Fabricated metal products (except machinery and equipment) 28 SD Medium 7.25 13 Machinery and equipment, NEC 29 SB High-tech 15.31 14 Office, accounting, and computing machinery 30 SB High-tech 8.06 15 Electrical machinery and apparatus, NEC 31 SS High-tech 8.01 16 Radio, television, and communication equipment 32 SB High-tech 8.03 17 Medical precision and optical instruments 33 SB High-tech 7.09 18 Motor vehicles, trailers, and semitrailers 34 SI Medium-high 2.17 19 Other transport equipment 35 SS Medium-high 1.13 20 Manufacturing NEC and recycling 36 SD Low-tech 2.19 Nr. Sectors (NACE Rev.1) NACE codes Pavitt classification Peneder classification Patents (% over the sample total) Manufacturing 1 Food products, beverages, and tobacco 15–16 SD Med low-tech 2.42 2 Textiles 17 SD Med high-tech 5.02 3 Wearing apparel, dressing and dyeing 18 SD Low-tech 5.04 4 Leather, leather products, and footwear 19 SD Low-tech 0.20 5 Wood and products of wood and cork 20 SD Medium 0.79 6 Pulp, paper, and paper products 21 SI Medium 2.34 7 Printing and publishing 22 SI Med low-tech 1.26 8 Chemical and chemical products 24 SB Med high-tech 12.84 9 Rubber and plastic products 25 SI Med high-tech 3.71 10 Other nonmetallic mineral products 26 SI Med high-tech 4.84 11 Basic metals 27 SI Med high-tech 7.21 12 Fabricated metal products (except machinery and equipment) 28 SD Medium 7.25 13 Machinery and equipment, NEC 29 SB High-tech 15.31 14 Office, accounting, and computing machinery 30 SB High-tech 8.06 15 Electrical machinery and apparatus, NEC 31 SS High-tech 8.01 16 Radio, television, and communication equipment 32 SB High-tech 8.03 17 Medical precision and optical instruments 33 SB High-tech 7.09 18 Motor vehicles, trailers, and semitrailers 34 SI Medium-high 2.17 19 Other transport equipment 35 SS Medium-high 1.13 20 Manufacturing NEC and recycling 36 SD Low-tech 2.19 Note: We consider LT group also the following service sectors: inland transport (61), water transport (62), air transport (63) and transport auxiliary activities (64); as well as the agricultural and haunting (AtB), mining (C), electricity and gas (E), and the construction (F) sectors. On the contrary, ICT, R&D, and other business services are included in the HT group. The patents shares by sector are computed as the ratio between industry i’s patent stock and the total amount of patents in our sample. NEC, Non electrical components. View Large Table A5. The classification of sectors adopted for the HT-LT test Nr. Sectors (NACE Rev.1) NACE codes Pavitt classification Peneder classification Patents (% over the sample total) Manufacturing 1 Food products, beverages, and tobacco 15–16 SD Med low-tech 2.42 2 Textiles 17 SD Med high-tech 5.02 3 Wearing apparel, dressing and dyeing 18 SD Low-tech 5.04 4 Leather, leather products, and footwear 19 SD Low-tech 0.20 5 Wood and products of wood and cork 20 SD Medium 0.79 6 Pulp, paper, and paper products 21 SI Medium 2.34 7 Printing and publishing 22 SI Med low-tech 1.26 8 Chemical and chemical products 24 SB Med high-tech 12.84 9 Rubber and plastic products 25 SI Med high-tech 3.71 10 Other nonmetallic mineral products 26 SI Med high-tech 4.84 11 Basic metals 27 SI Med high-tech 7.21 12 Fabricated metal products (except machinery and equipment) 28 SD Medium 7.25 13 Machinery and equipment, NEC 29 SB High-tech 15.31 14 Office, accounting, and computing machinery 30 SB High-tech 8.06 15 Electrical machinery and apparatus, NEC 31 SS High-tech 8.01 16 Radio, television, and communication equipment 32 SB High-tech 8.03 17 Medical precision and optical instruments 33 SB High-tech 7.09 18 Motor vehicles, trailers, and semitrailers 34 SI Medium-high 2.17 19 Other transport equipment 35 SS Medium-high 1.13 20 Manufacturing NEC and recycling 36 SD Low-tech 2.19 Nr. Sectors (NACE Rev.1) NACE codes Pavitt classification Peneder classification Patents (% over the sample total) Manufacturing 1 Food products, beverages, and tobacco 15–16 SD Med low-tech 2.42 2 Textiles 17 SD Med high-tech 5.02 3 Wearing apparel, dressing and dyeing 18 SD Low-tech 5.04 4 Leather, leather products, and footwear 19 SD Low-tech 0.20 5 Wood and products of wood and cork 20 SD Medium 0.79 6 Pulp, paper, and paper products 21 SI Medium 2.34 7 Printing and publishing 22 SI Med low-tech 1.26 8 Chemical and chemical products 24 SB Med high-tech 12.84 9 Rubber and plastic products 25 SI Med high-tech 3.71 10 Other nonmetallic mineral products 26 SI Med high-tech 4.84 11 Basic metals 27 SI Med high-tech 7.21 12 Fabricated metal products (except machinery and equipment) 28 SD Medium 7.25 13 Machinery and equipment, NEC 29 SB High-tech 15.31 14 Office, accounting, and computing machinery 30 SB High-tech 8.06 15 Electrical machinery and apparatus, NEC 31 SS High-tech 8.01 16 Radio, television, and communication equipment 32 SB High-tech 8.03 17 Medical precision and optical instruments 33 SB High-tech 7.09 18 Motor vehicles, trailers, and semitrailers 34 SI Medium-high 2.17 19 Other transport equipment 35 SS Medium-high 1.13 20 Manufacturing NEC and recycling 36 SD Low-tech 2.19 Note: We consider LT group also the following service sectors: inland transport (61), water transport (62), air transport (63) and transport auxiliary activities (64); as well as the agricultural and haunting (AtB), mining (C), electricity and gas (E), and the construction (F) sectors. On the contrary, ICT, R&D, and other business services are included in the HT group. The patents shares by sector are computed as the ratio between industry i’s patent stock and the total amount of patents in our sample. NEC, Non electrical components. View Large © The Author(s) 2018. Published by Oxford University Press on behalf of Associazione ICC. All rights reserved. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)
TI - The demand-pull effect of public procurement on innovation and industrial renewal
JF - Industrial and Corporate Change
DO - 10.1093/icc/dty055
DA - 2019-08-01
UR - https://www.deepdyve.com/lp/oxford-university-press/the-demand-pull-effect-of-public-procurement-on-innovation-and-tU91ok5kJ4
SP - 793
VL - 28
IS - 4
DP - DeepDyve
ER -