What do firms know? What do they produce? A new look at the relationship between patenting profiles and patterns of product diversification

Dosi, Giovanni; Grazzi, Marco; Moschella, Daniele Working Paper What do firms know? What do they produce? A new look at the relationship between patenting profiles and patterns of product diversification Quaderni - Working Paper DSE, No. 1004 Provided in Cooperation with: University of Bologna, Department of Economics Suggested Citation: Dosi, Giovanni; Grazzi, Marco; Moschella, Daniele (2015) : What do firms know? What do they produce? A new look at the relationship between patenting profiles and patterns of product diversification, Quaderni - Working Paper DSE, No. 1004, Alma Mater Studiorum - Università di Bologna, Dipartimento di Scienze Economiche (DSE), Bologna, https://doi.org/10.6092/unibo/amsacta/4243 This Version is available at: https://hdl.handle.net/10419/159842 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc/3.0/ ISSN 2282-6483 What do firms know? What do they produce? A new look at the relationship between patenting profiles and patterns of product diversification Giovanni Dosi Marco Grazzi Daniele Moschella Quaderni - Working Paper DSE N°1004 What do firms know? What do they produce? A new look at the relationship between patenting profiles and patterns of product diversification∗ Giovanni Dosi†1, Marco Grazzi‡2, and Daniele Moschella§1,3 1Institute of Economics, Scuola Superiore Sant’Anna 2Department of Economics, University of Bologna 3IBIMET-CNR Abstract In this work we analyze the relationship between the patterns of firm diversification, if any, across product lines and across bodies of innovative knowledge, proxied by the patent classes where the firm is present. Putting it more emphatically we investigate the relationship between “what a firm does” and “what a firm knows”. Using a newly developed dataset matching information on patents and products at the firm level, we provide evidence concerning firms’ technological and product scope, their relationships, the size-scaling and coherence properties of diversification itself. Our analysis shows that typically firms are much more diversified in terms of products than in terms of technologies, with their main products more related to the exploitation of their innovative knowledge. The scaling properties show that the number of products and technologies increase log-linearly with firm size. And the directions of diversification themselves display coherence between neighboring activities also at relatively high degrees of diversification. These findings are well in tune with a capability-based theory of the firm. JEL codes: C81, D22, L20, L25, O31 Keywords: Diversification, Coherence, Patents, Products, Capabilities, Firm behaviour ∗We thank the participants at the Conference on Entrepreneurship, Innovation and Enterprise Dynamics organized by the OECD Working Party on Industry Analysis (WPIA) in Paris (2014). We also thank Giulio Bottazzi and Davide Pirino for having allowed to use the computing routines associated to their coherence measure. We gratefully acknowledge the research support by the IBIMET-CNR (grant CrisisLab-ProCoPe). Marco Grazzi gratefully acknowledges Fondazione Cassa dei Risparmi di Forl`ı for financial support. Without the unique support of the Italian Statistical Office, and in particular Roberto Monducci, this all endeavour would not have been possible. The usual disclaimer applies. †Piazza Martiri della Libert`a 33, 56127 Pisa, Italy, tel: +39 050883343, email: [email protected] ‡Piazza Scaravilli 2, 40126 Bologna, Italy. tel: +39 0512098130, email: [email protected] §Piazza Martiri della Libert`a 33, 56127 Pisa, Italy, tel: +39 050883343, email: d.m[email protected] 1 1 Introduction In this work we analyze the relationship between the patterns of firm diversification, if any, across product lines and across bodies of innovative knowledge, proxied by the patent classes where the firm is present. Putting it more emphatically we investigate the relationship between “what a firm does” and “what a firm knows”. Firms, and large companies in particular, tend to be active in more than one product. In this respect, on the one side the process of firm growth is related to the choice of the diversification pattern of the existing product portfolio and to the entry into new sub-markets (Bottazzi et al., 2001). On the other side, the successful accomplishment of the introduction of a new product crucially depends upon, and is constrained by, the existing stock of knowledge embedded in the organization (see among the others Nelson and Winter, 1982; Teece et al., 1994). The technological and product breadth of a firm co-evolve over time, with technological diversification generally anticipating the introduction of new products (Pavitt, 1998). Knowledge accumulation and innovation, the evidence shows, are indeed positively related to productivity and contribute to boost sales, both on domestic and international markets (for empirical evidence see, among others, Hall and Mairesse, 1995; Wakelin, 1998; Dosi et al., 2014). Product innovation is one of the main forms of such innovative efforts: introducing a new good is likely to benefit the innovating firm in terms of market shares and penetration in new markets. Moreover, by producing new goods, the firm can diversify its product portfolio and exploit economies of scope of various kind (Hirsch and Lev, 1971; Montgomery, 1994). Also the process of continuous feedback between internal research activities and the actual production of artefacts contributes to the accumulation of capabilities and skills (see, among the others, Dosi, 1988; Pavitt, 1998). In all that a crucial question concerns the precise relation between the expansion of the scope of technological knowledge on the one hand and of product portfolios on the other. Relatedly, a central issue regards the pattern and direction of diversification: how do firms choose to enter new technological fields or add products, conditional upon the existing stock of knowledge and product portfolio? Teece et al. (1994) investigates the diversification of US corporates larger than 20 employees into different industrial sectors and find that firms tend to diversify into related or neighboring industries. Patel and Pavitt (1997), who focus instead on the technological knowledge, report an analogous finding of path dependency in the accumulation of firm-specific technological competencies. Several studies have further contributed to support the relatedness hypothesis (Breschi et al., 2003; Valvano and Vannoni, 2003; Nesta and Saviotti, 2006; Bottazzi and Pirino, 2010). And more recent work of Piscitello (2000, 2004); Miller (2006) started to investigate the relation between the diversification pattern followed by firms and corporate performance, finding a positive relation between corporate coherence and firm performance. In terms of scope of diversification Patel and Pavitt (1997), looking at large diversified firms, find that “firms know more than they do”: the scope of knowledge diversification is wider than product diversification. Brusoni et al. (2001) confirm the finding in the case of the highly specialized industry of aircraft engine control systems and show that in a highly interdependent environment, in which a change in one firm within the network affects all the others, corporates “need to have knowledge in excess of what they need for what they make, to cope with imbalances caused by uneven rates of development in the technologies on which they rely and with unpredictable product-level interdependencies.” This work provides several contributions to this whole stream of literature. First, 2 we analyze the characteristics of the universe of Italian exporting firms and those which display innovative technological knowledge, as proxied by patents1. Patenting firms turn out to be bigger than non patenting firms, both in terms of total exports and in terms of product scope. Second, we investigate the properties of the diversification breadth of both technological knowledge and product portfolios, using detailed product information from custom data. In this respect, employing the algorithm recently developed by (Lybbert and Zolas, 2014), we match the technological field in which a firm holds a patent to the corresponding 4-digit industry in which the firm is active with at least one product. This matching, although with some caveats, enables us to shed light on relevance of specific technological knowledge in backing the product in international markets and also on the Pavittian question of “knowing” vs. “making”. In our Italian case we find an anti-Pavittian answer. The overwhelming majority of patenting firms appear to be more specialized in terms of innovative knowledge than products. Third, we study the scaling relation between the size of the firm and diversification of both technological knowledge and product scope. We do find that both scaling relations are essentially log-linear: this supports the idea that firm diversification unfolds as a branching process characterized by incremental accumulation of capabilities. Finally, we investigate to what extent firms are coherent in their technological and product diversification. Using a refined version of the Teece et al. (1994) measure of corporate coherence, developed in Bottazzi and Pirino (2010), we find that as firms introduce new products and develop new technologies, the coherence between neighboring activities is high for relatively low levels of diversification, but remain present also for sufficiently diversified firms. 2 Dataset description In order to investigate the characteristics of technological and product diversification, three different microdata have to be linked together. We take information on Italian firms’ patent applications from AMADEUS, a commercial database provided by Bureau van Dijk. The last update at our access (March 2015) covers around one million of Italian companies, of which more than 20,000 turn out to own at least one patent. For those firms, AMADEUS also provides some basic information on patents, including the IPC classification code, the application date, and whether the patent has been granted or not. In order to have a better proxy of technological capabilities, here we focus only on patents that have been applied to USPTO or EPO offices, and that have been effectively granted; this choice restricts the original dataset to a smaller one of 49,803 patents owned by 7,311 firms.2 Data on the production activities of Italian firms come from two firm-level datasets collected by the Italian statistical office (ISTAT), namely the Business Register known as “Archivio Statistico Imprese Attive” (ASIA) and Statistiche del Commercio Estero (COE). ASIA is the register of all active Italian businesses. It covers the period 19982006 and contains information on firms’ operations including the number of employees, 1On the limitations, but also the relevance of the use of patents as a proxy for innovation, see the critical review by Griliches (1990). 2Notice that a single patent may be owned by more than one firm. In case of co-patenting, our analysis credits the patent to each co-patentee, as it is usually done in the literature (see, for example, Breschi et al., 2003). 3 Table 1: Patents and firms, by period of application and patent office Total USPTO EPO Period patents firms patents firms patents firms 1949-1978 1,086 187 1,086 187 1979-1995 8,055 1,426 3,929 863 4,126 1,168 1996-2006 21,305 2,946 9,817 1,647 11,488 2,499 2007-2014 9,340 1,948 4,871 1,006 4,469 1,550 1949-2014 39,786 4,411 19,703 2,586 20,083 3,709 Note. Number of USPTO and EPO granted patents owned by Italian firms. The period refers to the application date. Data from AMADEUS, ASIA, and COE. total turnover,3and firm’s age, defined as the year of incorporation. We are able to link around 85% of firms in AMADEUS and 90% of their patent applications granted in USPTO and EPO to 2006 ASIA archive. The higher rate of matching for patent applications signals that the firms left out of the combined database are mostly small patentees (the median value of their patent stock is 1). This is partly due to the fact that ASIA archive refers to 2006, whereas AMADEUS also contains information on firms born later on. These young firms are most likely small patentees (see below). The COE dataset consists of all cross-border transactions performed by Italian firms and it covers the period 2000-2007. COE includes the annual value and quantity of export transactions by the firm at the product level. A product is defined as a six digit category in the Harmonized System (HS6). 4Using the unique identification code of the firm, we link the firm-level export data from COE 2006 to the ASIA-AMADEUS dataset. After this final link, we are able to maintain information for around 90% of patent applications (with respect to the ASIA-AMADEUS dataset) and 70% of firms. Such a great proportion of patenting firms that are present both in ASIA and in COE dataset is explained by the general high export propensity of patenting firms (see Dosi et al., 2014). Finally, notice that as COE reports information on exported goods, patenting firms that are active in non manufacturing sectors are less likely to be present in COE: in our case, around 70% of non matched firms are active in non manufacturing firms. Once again, the very high rate of matching for patent applications says that patenting firms that exports tend to be, on average, bigger (in terms of patents) of patenting firms that do not export. Table 1 reports the number of patents and firms present in our final dataset, broken down by date of application and patent office. The first thing to note is that the number of patent applications during the period 1996-2006 overtake all the patent applications that took place before (and after5). These figures are consistent with the patent explosion that have been observed both in US and in Europe starting from the mid of 90’s (see the discussion in Dosi et al., 2010). Secondly, we observe that the number of patents applied to USPTO by Italian firms is similar to patents applied to EPO, even if there are much more firms applying to EPO. This implies that small patentees are more likely to apply for a patent only in Europe. The joint analysis of USPTO and EPO patents allows us not to loose information about such small patentees while at the same time considering 3Information on total turnover is available only in 2000 and 2003. 4More on the dataset in Grazzi et al. (2013). 5Notice that figures relative to more recent years are less reliable due to the time lag between a patent application and its grant. 4 Table 2: Distribution of firms and patents by number of patents No. of patents #1 #2 #3 #4 #5 #6 #7 #8 #9 #10+ Obs. Share of firms 41.68 19.14 9.37 6.45 4.41 3.29 1.66 1.60 1.15 11.24 2,946 Share of patents 5.73 5.26 3.86 3.54 3.03 2.71 1.60 1.75 1.43 71.08 21,441 Source. Amadeus, ASIA, and COE, 1996-2006. only patents of higher average quality (with respect to the universe of patents). In the following analysis, we shall proxy firms’ technological capabilities in 2006 by considering only patents applied during the period 1996-2006 (in boldface in the table). This ten-year counting period seems a reasonable choice as it discards too “old” patents as a meaningful proxy of firms’ capabilities (for similar time periods, see Patel and Pavitt, 1997; Breschi et al., 2003). Table 2 reports the distribution of firms by number of patents (first row) and also the share of patents accounted for by those firms (second row). The distribution is, as expected, extremely skewed: the typical firm owns just one patent (41.68%), and more than 60% of firms do not have more than two patents. On the right tail of the distribution, there is about only a 10% of firms which have ten or more patents: these firms, however, account disproportionately for the whole stock of patents (71.08%). In the next sections, we shall analyze the patterns of technological diversification of patenting firms and their relationship with product diversification. In order to do so, we need a classification of technological fields which is consistent with standard industrial and product classification. In this respect, existing IPC-industry concordances are far from being satisfactory as they do not allow to go further than using a 2-digits (in some cases 3-digits) classification of industrial activities6. To overcome this limitation, we resort to the probabilistic algorithm recently developed by Lybbert and Zolas (2014) which allows to build a correspondence between technological and production activities of the firm at a more disaggregated level. In particular, we link IPC codes to 125 4-digits ISIC codes (Rev. 3), and SH6 codes to 145 4-digits ISIC codes (Rev. 3). Firms in our sample turn out to patent in 118 different technological fields and produce 138 different products.7 Before moving to the main analysis, let us flag some of the characteristics that distinguish patenting from non patenting firms. 3 Characteristics of patenting firms In this section we exploit our integrated dataset to compare patenting firms with the universe of exporters in 2006. Table 3 reports the differences between patenting and non-patenting firms in terms of size, age, and number of products. Columns (1), (2), and (3) report the β’s estimated from the following regression: Xi=α+βDP ATi+ǫi where DP ATiis a binary variable taking value one if the firm has a granted patent in 6See in particular the Schmoch et al. (2003) concordance table. 7In what follows, we will assume that “products” and “technological fields” map one-to-one into each other (a 4-digit ISIC Rev.3 sector). 5 Table 3: Size, age, and #products: patenting vs. non-patenting firms (1) (2) (3) (4) (5) Dep. var. sizeiageiagei#productsi#productsi DP ATi3.092∗∗∗ 0.320∗∗∗ 0.189∗∗∗ 1.003∗∗∗ 0.371∗∗∗ (0.047) (0.016) (0.017) (0.015) (0.012) sizei0.042∗∗∗ 0.205∗∗∗ (0.001) (0.001) N139,360 139,360 139,360 139,360 139,360 adj. R20.195 0.060 0.074 0.122 0.484 Sector dummies Yes Yes Yes Yes Yes Note. Standard errors in parentheses.∗p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01 the period 1996-2006, and zero otherwise, and Xis either size, measured as the (log) of trade volumes, or (log) age or (log) number of products. Columns (2) and (4) estimate the same equation also controlling for size. Size, age and number of products are measured in 2006. Patenting firms are both bigger (column 1) and older (columns 2 and 3) than non patenting firms, and they are also more diversified in terms of products, even when controlling for size (columns 4 and 5). We also investigate within each product-country pair the price (unit value) of patentees and non-patentees. For each product-country pair we get two price distributions, one for patentees and one for non-patentees. By employing non-parametric Fligner-Policello (FP) test we study to what extent one distribution dominate stochastically the other. Notice that to compute the statistics we need at least 20 observations both for patentees and non-patentees. We find that, when the two distributions statistically differ, in (around) 80% of the product-country pair, patentees have higher price. 4 Technological and product diversification: some stylized facts There is a remarkable degree of firm heterogeneity both in terms of technological and product diversification. A summary account of the diversification structure of firms in our database is provided in Figures 1,2,3, and 4. Figure 1 reports the distribution of firms according to the number of different technological fields in which they applied (and got granted) patents in the period 1996-2006. More than 65% of all firms are present in just one technological field, and around 20% in two. Thus, the distribution appears even more skewed than the one observed in Table 2: many firms do own more than one patent in the same technological field. Another implication of Figure 1 is that around 55% of diversified firms, i.e. firms which are active in more than one field, are active in just two fields. This figure is similar the one summarized in Breschi et al. (2003) who, however, consider a narrower classification of 30 technological fields. Very few firms are present in several technological fields: the most diversified firm is active in 35 fields (out of 125), with the second most diversified patenting in 23 fields and three firms in 20. These (few) very diversified firms are large patentees, as it is apparent from Figure 2, which reports the distribution of patents held by firms according to the number of tech6 0 10 20 30 40 50 60 70 (Percentage) 1 5 10 15 20 25 30 35 (No. of technological fields) Figure 1: Distribution of firms according to the number of technological fields in which they have a granted patent nological fields in which those firms applied for a patent in the period 1996-2006. The single firm patenting in 35 different fields owns around 15% of all patents; more in general, firms diversified in at least 10 technological fields account for more than 35% of all patent. On the other hand, firms which patent in less than 3 fields, and that represent around 85% of all firms, account for 30% of all patents. Figures 3 and 4 report similar distributions for products, with weights given by the export value of each category (Fig. 4). Figure 3 reveals that patenting firms are, on average, quite diversified in terms of products: around 40% of firms produce 10 or more products, and there is also a firm which is active in more than half of the total available products (77 over 145). As it is apparent from Figure 4, firms who diversify a lot across products account disproportionately for total export value, even if there is no single category which produces more than 6% of total export value. The relative higher frequency of multiproduct firms with respect to multitechnology firms suggests that firms, on average, diversify more across products than across technologies. This result stands in contrast to the evidence put forward by Patel and Pavitt (1997) and Brusoni et al. (2001), according to which firms’ technological reach is greater than their product reach. How do we interpret such different views? We try to shed some light on this issue by analyzing the joint distribution of firms across number of products and technological fields, reported in Table 4. Each row distributes the total number of firms active in a given number of technological fields according to the number of products they produce. Considering the first row, which contains all the firms active in just one technological field, we see that around 90% of them produce more than one product. In general, firms on the left of the main 7 The standard null hypothesis does not impose any constraint on the number of activities in which each firm is present: in principle, a firm could be present in all products or technological classes. As noted in Bottazzi and Pirino (2010), one implication of this assumption is that the implied distribution of firm scope converges to a binomial, in contrast with the Paretian shape that is usually observed in real data (and also in our dataset: see section 4, Figures 1 and 3). To overcome this drawback, Bottazzi and Pirino (2010) propose a different null hypothesis, according to which both firms scope and the number of firms per activities are fixed and correspond to the observed ones. We will refer to this hypothesis as the alternative null hypothesis. In the following analysis, we shall use both null hypotheses. The standard null hypothesis allows to compare our results with the ones usually found in the literature. The alternative null hypothesis tests the robustness of the results with respect to more restrictive assumptions. Notice that deriving the P-value under the alternative null hypothesis is, in general, not trivial. Here, we adopt the Monte Carlo algorithm developed and explained at greater length in Bottazzi and Pirino (2010). Equipped with our P-values, we can now calculate a modified version of two largely diffused measures of firm coherence. The first one is based on the weighted-average relatedness of activity ito all other activities within the firm and is defined as: W ARk(H) = 1 nX i Cik Pj6=ipij(H)wjk Pj6=iwjk !(4) where nis total number of products (technological fields) in which a firms is active and wjk the weight of product (technological field) jwith respect to firm k. We weight products with export share and technological fields with patent count. The second one is based on the weighted average relatedness of neighbors and is defined as: W ARNk(H) = 1 nX i Cik Pj6=ipij(H)mk ijwjk Pj6=imk ijwjk !(5) where mk ij = 1 if the pair ij is in the maximum spanning tree of firm k, defined as the graph with n−1 links such that the sum of the relatedness measures on each link is largest. Notice that W AR and W ARN capture two different aspects of firm diversification structure. W AR measures the inverse of the average distance from a firm activity to all other activities, whereas W ARN measures the inverse of the average distance from a firm activity to its neighbor activity. As firms diversify in more and more products and technological fields, one would expect the W AR measure, on average, to increase. However, if the competence-driven branching process is a reasonable account of firm diversification, the W ARN measure should not be affected by the degree of diversification, as firm should add products or technologies that are near to some of its previous products and technologies. 6.2 Results In Figure 7 (a) and (c), we plot the average W ARkof the firms belonging to different #products bins against the bin average #products under, respectively, the standard and the alternative null hypothesis. As expected, as firms increase their product scope, the coherence across all its activities decrease. The observed relationship seems to be linear 14 0.91 0.92 0.93 0.94 0.95 0.96 0.97 0.98 0.99 0 10 20 30 40 50 60 70 80 WAR #products (a) 0.988 0.99 0.992 0.994 0.996 0.998 1 0 10 20 30 40 50 60 70 80 WARN #products (b) 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0 10 20 30 40 50 60 70 80 WAR #products (c) 0.8 0.82 0.84 0.86 0.88 0.9 0.92 0.94 0.96 0.98 1 0 10 20 30 40 50 60 70 80 WARN #products (d) Figure 7: Binned relationship between product WAR (left) and WARN (right) as a function of #products. Standard null hypothesis (top) and alternative null hypothesis (bottom). Linear fit in (a) and log-linear fit in (c). under the standard null hypothesis, but log-linear under the alternative hypothesis. The interpretation is that if heterogeneity in firm scope is properly accounted for, then the addition of a new product to small firms reduce coherence much more than in large firm (see Bottazzi and Pirino, 2010). In Figure 7 (b) and (d), we plot the same binned relationship for the W ARNkmeasure. It is apparent that as firms introduce new products the coherence between neighboring activities slightly increase for relatively low levels of diversification, and stay constant for sufficiently diversified firms. In Figure 8 (a)-(d), we repeat the same exercise as before by plotting the binned relationship between #technological fields and coherence measures. The scantiness of observations related to diversified firms make it difficult to appreciate any real difference between (a) and (c): in general, no clear relationship seems to emerge between levels of diversification and average coherence across all technological fields. It is clear, however, from (b) and (d) that also in the case of technological fields, diversification is associated to constant levels of coherence, in agreement with our hypothesis of competence-driven diversification and with the original intuition of Teece et al. (1994). 15 0.74 0.76 0.78 0.8 0.82 0.84 0.86 0.88 0 5 10 15 20 25 30 35 WAR #tech. fields (a) 0 0.2 0.4 0.6 0.8 1 0 5 10 15 20 25 30 35 WARN #tech. fields (b) 0.6 0.62 0.64 0.66 0.68 0.7 0.72 0.74 0.76 0.78 0 5 10 15 20 25 30 35 WAR #tech. fields (c) 0 0.2 0.4 0.6 0.8 1 0 5 10 15 20 25 30 35 WARN #tech. fields (d) Figure 8: Binned relationship between technological WAR (left) and WARN (right) as a function of #technological fields. Standard null hypothesis (top) and alternative null hypothesis (bottom). Linear fit in (a) and log-linear fit in (c). 16 7 Conclusions The paper contributes to the analysis of the main patterns of firm diversification both in the technological and in the product space. Using a newly developed dataset matching information on patents and products at the firm level, we have provided evidence concerning firms’ technological and product scope, their relationships, and the size-scaling properties of diversification itself. Our analysis shows that typically firms are much more diversified in terms of products than in terms of technologies, even if their main products are related to the exploitation of their (seemingly narrow) innovative knowledge. Interestingly, “Pavitt” firms, who know more than they make, are quite rare, mostly limited to the few large patentees. At the risk of overtheorizing, the picture of the Italian national system of innovation and production appears to be one of complementarity between few relatively large members of international oligopolies - especially in relatively new technological paradigms - which “know more than they do”, and a large cloud of relatively narrowly specialized firms. At the same time, the analysis has also shown some similarities between the technological and the product diversification processes. In particular, the scaling properties of both show that the number of products and technologies increase log-linearly as firms grow. And the directions of diversification themselves display coherence between neighboring activities also at relatively high degrees of diversification (consistently with the conjecture put forward in Teece et al. (1994)). Both findings are well in tune with a capability-based theory of the firm (for a critical survey within a vast literature, see Dosi et al. (2008)). The patterns of diversification - both in technologies knowledge and in products - are consistent with a branching process whereby knowledge on production and innovation, so to speak, “spurs out” from what the firm already does and knows. And of course, the coherence in the directions of diversification reinforced the point: where a firm stands in term of pre-existing capabilities shapes to a good extent where it will go. References Bottazzi, G., G. Dosi, M. Lippi, F. Pammolli, and M. Riccaboni (2001): “Innovation and corporate growth in the evolution of the drug industry,” International Journal of Industrial Organization, 19, 1161–1187. Bottazzi, G. and D. Pirino (2010): “Measuring Industry Relatedness and Corporate Coherence,” LEM Papers Series 2010/10, Laboratory of Economics and Management (LEM), Sant’Anna School of Advanced Studies, Pisa, Italy. Bottazzi, G. and A. Secchi (2006): “Gibrat’s Law and diversification,” Industrial and Corporate Change, 15, 847–875. Breschi, S., F. Lissoni, and F. Malerba (2003): “Knowledge-relatedness in firm technological diversification,” Research Policy, 32, 69 – 87. Brusoni, S., A. Prencipe, and K. Pavitt (2001): “Knowledge Specialisation, Organizational Coupling and the Boundaries of the Firm: Why Firms Know More Than They Make?” Administrative Science Quarterly, 46, 597–621. Dosi, G. (1988): “Sources, Procedures, and Microeconomic Effects of Innovation,” Journal of Economic Literature, 26, 1120–71. 17 Dosi, G., M. Faillo, and L. Marengo (2008): “Organizational Capabilities, Patterns of Knowledge Accumulation and Governance Structures in Business Firms: An Introduction,” Organization Studies, 29, 1165–1185. Dosi, G., M. Grazzi, and D. Moschella (2014): “Technology and costs in international competitiveness: from countries and sectors to firms,” LEM Papers Series 2014/10, Laboratory of Economics and Management (LEM), Sant’Anna School of Advanced Studies, Pisa, Italy. Dosi, G., L. Marengo, and C. Pasquali (2010): “How much should society fuel the greed of innovators? On the relations between appropriability, opportunities and rates of innovation,” in The Capitalization of Knowledge, ed. by R. Viale and H. Etzkowitz, Edward Elgar. Dosi, G., R. R. Nelson, and S. Winter (2000): “Introduction,” in The nature and dynamics of organizational capabilities, ed. by G. Dosi, R. R. Nelson, and S. Winter, Oxford University Press: Oxford. Grazzi, M., R. Sanzo, A. Secchi, and A. Zeli (2013): “The building process of a new integrated system of business micro-data 1989-2004,” Journal of Economic and Social Measurement, 38, 291–324. Griliches, Z. (1990): “Patent Statistics as Economic Indicators: A Survey,” Journal of Economic Literature, 28, 1661–1707. Hall, B. H. and J. Mairesse (1995): “Exploring the relationship between R&D and productivity in French manufacturing firms,” Journal of Econometrics, 65, 263–293. Hirsch, S. and B. Lev (1971): “Sales Stabilization Through Export Diversification,” The Review of Economics and Statistics, 53, 270–77. Lybbert, T. J. and N. J. Zolas (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, 530–542. Miller, D. J. (2006): “Technological diversity, related diversification, and firm performance,” Strategic Management Journal, 27, 601–619. Montgomery, C. A. (1994): “Corporate diversification,” The Journal of Economic Perspectives, 8, 163–178. Nelson, R. R. and S. G. Winter (1982): An Evolutionary Theory of Economic Change, The Belknap Press of Harvard University Press: Cambridge, MA. Nesta, L. and P.-P. Saviotti (2006): “Firm knowledge and market value in biotechnology,” Industrial and Corporate Change, 15, 625–652. Patel, P. and K. Pavitt (1997): “The technological competencies of the world’s largest firms: Complex and path-dependent, but not much variety,” Research Policy, 26, 141–156. 18 Pavitt, K. (1998): “Technologies, Products and Organization in the Innovating Firm: What Adam Smith Tells Us and Joseph Schumpeter Doesn’t,” Industrial and Corporate Change, 7, 433–452. Piscitello, L. (2000): “Relatedness and coherence in technological and product diversification of the world’s largest firms,” Structural Change and Economic Dynamics, 11, 295–315. ——— (2004): “Corporate diversification, coherence and economic performance,” Industrial and Corporate Change, 13, 757–787. Schmoch, U., F. Laville, P. Patel, and R. Frietsch (2003): “Linking technology areas to industrial sectors,” Final Report to the European Commission, DG Research. Teece, D. J., R. Rumelt, G. Dosi, and S. Winter (1994): “Understanding corporate coherence: Theory and evidence,” Journal of Economic Behavior & Organization, 23, 1–30. Valvano, S. and D. Vannoni (2003): “Diversification Strategies and Corporate Coherence Evidence from Italian Leading Firms,” Review of Industrial Organization, 23, 25–41. Wakelin, K. (1998): “Innovation and export behaviour at the firm level,” Research policy, 26, 829–841. Winter, S. G. (2003): “Understanding dynamic capabilities,” Strategic management journal, 24, 991–995. 19 