Knowledge spillovers from renewable energy technologies: Lessons from patent citations

G Model EIST-219; No. of Pages 14

ARTICLE IN PRESS Environmental Innovation and Societal Transitions xxx (2016) xxx–xxx

Contents lists available at ScienceDirect

Environmental Innovation and Societal Transitions journal homepage: www.elsevier.com/locate/eist

Knowledge spillovers from renewable energy technologies: Lessons from patent citations夽 Joëlle Noailly a,b,∗ , Victoria Shestalova a,c a b c

CPB Netherlands Bureau for Economic Policy Analysis, The Hague, The Netherlands Centre for International Environmental Studies, Graduate Institute of International and Development Studies, Geneva, Switzerland NZa (Dutch Healthcare Authority), Utrecht, The Netherlands

a r t i c l e

i n f o

Article history: Received 10 November 2014 Received in revised form 22 July 2016 Accepted 22 July 2016 Available online xxx Keywords: Renewable energy Innovation Patents Knowledge spillovers Technology policy

a b s t r a c t This paper studies the knowledge spillovers generated by renewable energy technologies, unraveling the technological fields that benefit from knowledge developed in storage, solar, wind, marine, hydropower, geothermal, waste and biomass energy technologies. Using citation data of patents in renewable technologies filed at 18 European patent offices over the 1978–2006 period, the analysis examines the importance of knowledge flows within the same specific technological field (intra-technology spillovers), to other technologies in the field of power-generation (inter-technology spillovers), and to technologies unrelated to power-generation (external-technology spillovers). The results show significant differences across various technologies. Overall, patents in wind, storage and solar technologies tend to be more frequently cited than other technologies. While wind technologies mainly find applications within their own field, a large share of innovations in solar energy and storage technologies find applications outside the field of power generation. The paper discusses the implications of these results for policymaking. © 2016 Elsevier B.V. All rights reserved.

1. Introduction Climate change mitigation will require the increasing development of renewable energy technologies in the power generating sector. In Europe, renewable energy sources, such as solar, wind, geothermal, marine, hydropower, waste and biomass energy, represent about 24% of electricity production against 48% for fossil-fuels1 (European Energy Agency (EEA), 2012). Increasing the share of electricity produced by renewable sources could thus greatly reduce the levels of greenhouse gas emissions from the power generation sector, currently responsible for about 30% of carbon emissions in Europe. Although over the last few years, renewable energy production costs have shown notable decreases for several renewable technologies,

夽 This research was initiated when both authors were still working at the CPB Netherlands Bureau for Economic Policy Analysis. This study is part of the research project “Directed technical change and the environment” initiated by the Dutch Ministry of Economic Affairs. We are grateful to the representatives of Dutch ministries: Victor Joosen, Frans Duijnhouwer, Jochem van der Waals, Bert Knoester, Matthijs Neerbos; and our colleagues Rob Aalbers, George Gelauff, Free Huizinga, Bas Straathof, and Henry van der Wiel for their feedback during the project. We also would like to thank Ivan Haˇscˇ iˇc from the OECD for providing us with the most updated classification codes on technologies for the analysis. Finally, we are grateful to two anonymous reviewers for useful comments and suggestions. ∗ Corresponding author at: CIES, Graduate Institute of International and Development Studies, Chemin Eugène−Rigot 2, CP 136, 1211 Geneva 21, Switzerland. E-mail addresses: [email protected] (J. Noailly), [email protected] (V. Shestalova). 1 The rest being nuclear energy. Renewable energy is mostly hydropower. http://dx.doi.org/10.1016/j.eist.2016.07.004 2210-4224/© 2016 Elsevier B.V. All rights reserved.

Please cite this article in press as: Noailly, J., Shestalova, V., Knowledge spillovers from renewable energy technologies: Lessons from patent citations. Environ. Innovation Soc. Transitions (2016), http://dx.doi.org/10.1016/j.eist.2016.07.004

G Model EIST-219; No. of Pages 14

2

ARTICLE IN PRESS J. Noailly, V. Shestalova / Environmental Innovation and Societal Transitions xxx (2016) xxx–xxx

notably onshore wind and solar PV, some other renewable energy alternatives are still too expensive to compete with fossil fuel technologies2 (International Energy Agency (IEA), 2015). Technological innovation is thus key to lower the costs of renewable energy technologies. Public policies play an important role in stimulating innovation in this sector, since private firms have too weak incentives to invest in clean technologies (Jaffe et al., 2005). This occurs mainly because the consequence of pollution is not borne by the firm itself but by third parties (the so-called ‘environmental externality’) and because innovating firms cannot prevent other firms from benefiting from their new knowledge (the ‘knowledge externality’). Additional market failures and barriers (e.g. capital market constraints, information asymmetries, national security externalities; see Gillingham and Sweeney (2012), for a review) as well as characteristics inherent to the process of technological change (e.g. knowledge feedbacks, learning externalities, path-dependency, lock-ins; Arthur, 1989; Nelson and Winter, 1982; Rosenberg, 1994) further justify government intervention in renewable energy technological fields. Yet, many questions remain open regarding how such policy support should be designed. Some of the recurrent questions that have emerged within policy circles deal with the issue of which renewable technological field should receive most policy support. Another set of related questions examine whether R&D policies targeted at renewable energy should encourage more specific or diverse technological trajectories. Parts of the answers require a better understanding of how various pieces of knowledge are combined to enable the development of new technologies and a more careful investigation of how specific knowledge flows within or across technological fields. Consider, for instance, wind energy: if inventors in wind energy technologies mostly learn from prior art within the same technological field, these intra-technology spillovers will tend to reinforce the existing technological trajectory, so that R&D subsidies specifically targeted at wind energy will be particularly effective at encouraging further developments in this technological field. If instead technological developments in wind energy are driven by knowledge from various technological domains, then more generic policy measures targeted at developing inter-technology spillovers may be more beneficial. To provide some first answers to these questions, this study aims to present evidence on the extent and the direction of knowledge spillovers generated by renewable energy technologies. Our main research questions are: (1) which renewable energy technological fields generate the most knowledge spillovers? and (2) where do knowledge spillovers generated by renewable technological fields flow to? To address this second research question, we make a distinction between intratechnology knowledge spillovers (knowledge flows within the same field of renewable energy technology), inter-technology spillovers (knowledge flows to other power generation technologies) and external technology spillovers (knowledge flows to technologies outside the power generation field3 ). Our empirical analysis uses citations of patents in eight renewable energy technologies filed at 18 European patent offices over the 1978–2006 period. Results from negative binomial estimations show that wind, storage and solar patents tend to be the most frequently cited patents, suggesting that these fields are particularly important and valuable for society. Regarding the direction of knowledge spillovers, we find significant differences across the various technological fields. While wind technologies mainly find applications within their own technological field, a large share of innovations in solar energy and storage technological fields find applications outside the field of power generation. We provide a detailed description of the technological fields that benefit the most from knowledge in renewable energy and discuss the implications of our analysis for policymaking. The study is organized as follows. Section 2 describes the literature on knowledge spillovers, in particular related to energy. Section 3 discusses the patent data used in the analysis and the empirical methodology. Section 4 presents the results on knowledge spillovers. Section 5 discusses the implications for policies. Section 6 concludes. 2. Knowledge spillovers and energy patents The notion that technological innovation is the result of the combination of existing components is deeply rooted in the literature on the history of technological change (Usher, 1954). Nelson and Winter (1982) describe innovation as consisting “to a substantial extent of a recombination of conceptual and physical materials that were previously in existence” (1982, p.130) and emphasize the role that firms play in combining technical, organizational and market knowledge. The inherent combinatorial characteristic of innovation has led scholars to focus on the question of how new technologies build on prior art and on how inventors combine and transfer knowledge across technological domains. Since knowledge is a public good, part of an inventor’s original idea necessarily spills over to other firms, other sectors and other technological fields, generating positive externalities (the so-called ‘knowledge spillovers’) for the economy. Previous work on knowledge spillovers has exploited the comprehensive information provided by patent data to examine how knowledge flows from one inventor to the other. The analysis of R&D manager surveys by Jaffe et al. (2000) shows that patent citations provide a reasonably good indication of communication between inventors – a ‘learning trail’ – in the knowledge transfer process. As a result, a large body of literature has used patent citations to proxy the importance of

2 According to International Energy Agency (IEA) (2015), indicative global average onshore wind generation costs for new plants fell in the period 2010–2015 by an estimated 30% on average while that for new utility-scale solar PV declined by two-thirds, and additional declines are forecasted for the next five-year period. However, technologies such as offshore wind, solar thermal electricity and some bioenergy are still at the beginning of their learning curve. 3 Strictly speaking, we define external citations as citations to technological fields that are not included in our list of renewable and fossil fuel technological fields shown in Appendix A. The list covers a lion share of power generation and electricity storage fields. All the other technologies in the field of energy, e.g. energy transmission, are allocated to the external technology field.

Please cite this article in press as: Noailly, J., Shestalova, V., Knowledge spillovers from renewable energy technologies: Lessons from patent citations. Environ. Innovation Soc. Transitions (2016), http://dx.doi.org/10.1016/j.eist.2016.07.004

G Model EIST-219; No. of Pages 14

ARTICLE IN PRESS J. Noailly, V. Shestalova / Environmental Innovation and Societal Transitions xxx (2016) xxx–xxx

3

inventions (Jaffe and de Rassenfosse, 2016). Lanjouw and Schankerman (2004) argue that more frequently cited patents bring more value to society, as they provide building blocks to a larger number of future innovations. Schoenmakers and Duysters (2010) also assume that a large number of forward citations reflects the technological importance of the invention for future technological advances and use it as a criterion for distinguishing radical patents from non-radical ones In recent years, a few papers have aimed to measure knowledge spillovers in the field of energy technologies. Popp and Newell (2012) use patent citations to address the question of the social value of energy R&D in comparison to non-energy technologies. After correcting for factors that affect the likelihood of citations, they find that energy patents have more chance to be cited than other patents. Dechezleprêtre et al. (2013) analyze patent citations in four technological fields, namely, energy production, automobiles, fuel, and lighting, and find that within these fields clean inventions generate substantially more knowledge spillovers than dirty inventions. On average, clean patented inventions receive 43% more citations than patents from dirty technologies. These findings justify the fact that research activities targeted at clean technologies should receive higher R&D support than dirty technologies and suggest that the benefits from higher spillovers from clean inventions might exceed the costs of environmental policies. While these studies tend to compare the knowledge spillovers of clean versus dirty technological fields, there is little empirical evidence comparing the importance of inventions among renewable energy fields using patent citations. Hence, our main hypothesis regarding our first research question (“which renewable energy technological fields generate the most knowledge spillovers?”) is that some technological fields generate higher levels of knowledge spillovers than others. Depending on inherent characteristics, certain technological fields will generate more spillovers, and thus be relatively more important, than other fields. Based on the above-cited literature comparing clean and dirty technologies and reporting higher spillover effects from clean technologies, an additional hypothesis is that specific renewable energy technological fields covered by our study generate more spillovers than fossil-fuel power generation fields. Besides measuring the extent of knowledge spillovers, the innovation literature has also studied questions regarding the direction of knowledge flows, such as: where does knowledge contributing to the development of a technology come from? The literature generally makes a distinction between specialized (‘local’ or close) or diversified (‘distant’4 ;) knowledge. On the one hand, some authors have argued that relying on prior knowledge which is relatively more specific to a particular technological field may generate higher knowledge spillovers. This is because local search and local learning may benefit from the incremental cumulative character of knowledge within a technological trajectory (Dosi, 1982; Cohen and Levinthal, 1990). Using patent citations of US patents between 1976 and 2006, Nemet and Johnson (2012) find for instance that integrating technological ‘near’ prior art is more important for knowledge spillovers than ‘distant’ prior art. This is because the integration of external knowledge may be more costly, time consuming and risky than that of more proximate knowledge. On the other hand, the evolutionary perspective on technology emphasizes that under certain circumstances – e.g. due to learning by doing, increasing returns to adoption, economies of scale in production, technological interrelatedness – local search may become too highly localized and incremental in nature, leading to technological lock-ins and a too high level of path-dependency in innovation (Arthur, 1989; David, 1985; Unruh, 2002).5 To limit technological lock-ins, diversified knowledge originating from various technological fields may be particularly valuable in contributing to technological progress. Several authors have emphasized the fact that most important inventions tend to use concepts from diverse scientific specialties and have implications across various industries (Rosenberg, 1994; Mowery and Rosenberg, 1998). Hence, combining knowledge from more diverse and distant prior art may generate larger knowledge spillovers and may also lead to more radical innovations (van den Bergh, 2008). In the field of energy technologies, several papers have aimed to identify whether the knowledge from clean energy spills over to a broad or narrow set of technological fields. Using patent citations, Popp and Newell (2012) find that patents in alternative energy are more ‘general’ than other patents (i.e., they contribute to a broader set of patent classes so that the knowledge from alternative energy patents is used by a broad set of technologies). Similarly, Dechezleprêtre et al. (2013) find that the knowledge from clean energy has a higher degree of ‘generality’ than the knowledge of dirty technologies, suggesting that the knowledge from clean energy technologies tends to spill over to many sectors with a broader set of applications than dirty energy technologies. Yet, these studies do not describe which specific technological fields benefit from knowledge developed in specific clean energy fields. Other papers have relied on a knowledge production function framework (Griliches, 1979) rather than patent citations, to identify whether knowledge from a specific energy technological field flows to the same technological field (intra-technology spillovers), to related technological fields (inter-technology spillovers) or to more distant technological fields (external spillovers). Looking at patents in eleven different energy technologies, Popp (2002) finds clear evidence for significant intra-technology knowledge spillovers. Johnstone and Haˇscˇ iˇc (2010) find evidence for intertechnology spillovers among some energy technologies, as they find that past knowledge accumulated in storage has a

4 Note that distance and diversity, although closely related concepts, are typically not defined in the same way. Jaffe and de Rassenfosse (2016) define ‘distance’ as how technologically different are the patents connected by a citation link and ‘diversity’ as whether the patents (independent of whether they are connected by a citation link) are bunched together in technology space. Also, Trajtenberg et al. (1997) measure technological distance using 3-digit patent classification classes: (e.g. when two patents are in the same technology class, distance is set to 0). Instead, they measure technological diversity as 1 minus the Herfindhal-Hirschman Index of concentration of citations across patent classes (e.g. diversity is zero if all citations are in the same class and it approaches 1 as the citations are more widely spread across all classes). 5 The idea of technological path-dependency has also been modelled in standard economic growth models with the environment as in Acemoglu et al. (2012).

Please cite this article in press as: Noailly, J., Shestalova, V., Knowledge spillovers from renewable energy technologies: Lessons from patent citations. Environ. Innovation Soc. Transitions (2016), http://dx.doi.org/10.1016/j.eist.2016.07.004

G Model EIST-219; No. of Pages 14

ARTICLE IN PRESS J. Noailly, V. Shestalova / Environmental Innovation and Societal Transitions xxx (2016) xxx–xxx

4

positive impact on innovation in other clean energy fields, especially in the field of intermittent technologies. Noailly and Smeets (2015) also find evidence for both intra- (within renewable technologies) and inter-technology (from fossil-fuels to renewable technologies) spillovers in the R&D portfolio of large patenting firms. Braun et al. (2010) find that solar and wind innovation greatly benefit from intra-technology spillovers. Yet, only wind seems to be affected by inter-technology spillovers (mainly from the field of energy machinery). Compared to these studies, we aim to provide novel evidence on the direction of spillovers (either intra-, inter-technology or external) from renewable energy patents using the learning trail embedded in patent citations. What kind of knowledge – either specialized or diversified – matters for technological advances may actually depend on the intrinsic characteristics of each technological field. In recent work, Battke et al. (2016) test a number of hypotheses on the factors affecting the direction of knowledge flows in the field of battery technologies. Their theoretical argument is that the diversity of prior art and the degree of technological centrality of knowledge affect the subsequent direction of this knowledge within and across technological field. Examining patent citations in the field of batteries, they find that innovation that is based on relatively less (more) diverse prior knowledge is more (less) likely to generate intra-technology (inter-technology) spillovers. They measure the diversity of prior art by looking at the count of backward citations to other battery patents (technologically ‘near’) and to non-battery patents (technological ‘distant’). In addition, they measure the level of centrality of knowledge by identifying various levels of product architecture via experts’ interviews. They conclude that specialized and core knowledge is more likely to flow within the same technological field, while diversified and peripheral knowledge is more likely to flow across technological fields. Hence, to address our second research question (“where do knowledge spillovers generated by renewable technological fields flow to?”) and given the mixed evidence from the literature, our hypothesis is that the direction of spillovers generated by renewable technologies differs across the various technological fields. Again, the difference across technologies is likely to be related to inherent characteristics of the technological fields.6 3. Methodology Section 3.1 introduces definitions that we use in the analysis of patent citations. Section 3.2 explains the estimation procedure and empirical methodology to uncover which technological fields benefit from knowledge in renewable energy technologies. 3.1. Definitions We measure knowledge spillovers by means of patent citations that are included in patent applications. There is a wellestablished literature arguing that patent citations represent a form of learning trail or knowledge flow from one technology to the other. Since we are interested in where knowledge from renewable energy flows to, we focus on forward citations, i.e. the citations by subsequent patents over time; reflecting the knowledge spillover from this patent to follow-on inventions. According to Jaffe et al. (1993), forward citations can measure “knowledge spillovers” under the assumption that “a citation of Patent X by Patent Y means that X represents a piece of previously existing knowledge upon which Y builds”. Since highlyvaluable patents tend to be cited more often, the number of forward citations also characterizes the value of the inventions (Trajtenberg, 1990). We focus on innovation aimed at improving the generation and storage of renewable energy, to which we will refer as REN-technological fields. The analysis covers eight REN-technological fields: wind, solar, hydropower, marine, biomass, geothermal, waste,7 and electricity storage8 technological fields. The patent applications in these technologies were selected using the relevant International Patent Classification (IPC) codes for each technology9 as borrowed from the earlier work by Johnstone et al. (2010) for renewable energy technologies, and Johnstone and Haˇscˇ iˇc (2010) for storage technologies. Along our analysis, we will investigate the extent of inter-technology spillovers, i.e. whether the knowledge from renewable technologies spills over to other power-generation technological fields. To this end, we make a distinction between inter-technology spillovers to renewable (inter-REN) and fossil-fuel (inter-to-FF) power-generation technological fields. We identify fossil-fuel energy patents using the International Patent Classification codes provided in Lanzi et al. (2011) and Haˇscˇ iˇc et al. (2009). Our set of FF energy patents pools patents in the production of fuel gases by carbureting air, steam engines plants, gas turbines plants, hot-gas or combustion-product positive displacement engine, steam generation, combustion

6 The scant literature examining the differences across renewable energy technological fields using patent citations confirms that there are important intrinsic characteristic to each technology. Huenteler et al. (2016) compare the technology life-cycles of wind and solar energy, using patent-citation networks, and find that wind technologies exhibit a more complex product architecture than solar PV technology, which seem to follow more the life-cycle pattern of mass-produced goods. In a descriptive analysis, Noailly and Shestalova (2013) show that solar and waste energy patents tend to cite patents across a larger number of technological fields than wind patents, suggesting that these two technologies may built on more diversified prior art than others. 7 We can question whether waste-to-energy technology can be categorized as renewable energy as burning household waste to produce heat for power generation present many similarities with burners and furnaces technologies used to burn fossil-fuels. Nonetheless, we classify waste-to-energy as a REN technology as is standard in the economic literature (Johnstone et al., 2010). 8 IEA (2014) stresses the valuable contribution of storage technologies to the development of energy systems; including actions on battery storage to facilitate renewables integration and frequency regulation. Johnstone and Haˇscˇ iˇc (2010) stress the importance of storage technologies for the development of intermittent renewable generation technologies. Note that pumped hydro storage falls into the hydropower category. See Table A1 in Appendix A. 9 Details on the IPC codes are given in Appendix A. We thank Ivan Haˇscˇ iˇc from the OECD for providing us with the most updated classification codes.

Please cite this article in press as: Noailly, J., Shestalova, V., Knowledge spillovers from renewable energy technologies: Lessons from patent citations. Environ. Innovation Soc. Transitions (2016), http://dx.doi.org/10.1016/j.eist.2016.07.004

G Model EIST-219; No. of Pages 14

ARTICLE IN PRESS J. Noailly, V. Shestalova / Environmental Innovation and Societal Transitions xxx (2016) xxx–xxx

5

apparatus, furnaces and improved compressed-ignition engines. In the empirical analysis, we will also provide additional results in which we compare the likelihood of citations between the various renewable technological fields and an average FF patent. All the other technologies, which are not covered by the definition of REN or FF, will be treated as external technologies. We select patent applications filed at the European Patent Office and 17 national European patent offices (EU-15, Norway, Switzerland) over the 1978–2006 period.10 The patent invention data are extracted from the EPO/OECD World Patent Statistical Database (PATSTAT). For each patent application, the database provides information on the year of application, the technological field of invention given by the IPC code, and the citations, i.e. the references to prior art used by this patent. For each patent, we count the number of forward citations by subsequent patents (if any). The resulting dataset includes all patents with their citations records, including patents that have zero citations. Since European patents also contribute to the knowledge developing outside Europe, we also consider citations by patents filed at the US Patent Office and at the Japanese Patent Office, as these two countries are the largest contributors to the worlds’ patents. There are several caveats to be aware of when working with patent citations. First, it is important to realize that not all citations included in a patent are added by inventors. In some countries, many references to prior art are added by patent attorneys and examiners; and there is evidence that examiners often add citations that were actually not known to the inventor. As examiner-added citations do not carry correct information on knowledge spillovers, this might affect our analysis of forward citations. Yet, the regression results are not vulnerable to bias, as long as the examiners are not biased towards a particular field and simply include more citations in all the fields.11 Second, some citations take place within the same family of patents, a patent family being a group of equivalent patents which have been granted in several different countries for the same invention. We thus do not include intra-family citations, for which both cited and citing patents refer to the same invention. The share of patents including intra-family citations, however, is negligible (about 1%) and leaving them in the dataset would not significantly affect the result. Third, we also exclude self-citations from the analysis. Presumably, citations to patents that belong to the same assignee represent transfers of knowledge that are mostly internalized, whereas citations to patents of other inventors are closer to the pure notion of spillovers.12 Furthermore, firms may include self-citations for strategic reasons. The share of self-citations is about 7% (about 2% of all citation records in total).13 In addition to this, it is important to acknowledge that our estimation results characterize spillover effects in the period covered by the dataset, which ends in 2006. Since some technologies, especially solar PV and batteries, have been rapidly developing in the period after 2006, limiting the dataset to an earlier period may introduce bias in our interpretation. For example, if more recent patents turn out to be more (or less) frequently cited, then our results overstate (or understate) spillover effects from the respective technological fields. At last, there are truncation issues for forward citations as the dataset cannot possibly include all patents that will be granted in the future. Earlier patents tend to be cited more often since they exist for a longer time period, thus, having more opportunities to be cited. To minimize truncation issues, we count the number of citations within a 5-year window after the year of application. This is to avoid that older patents, which had a longer period to be cited, are overweighed in estimation, so that each patent is taken with the same weight (Nemet and Johnson, 2012). One caveat of using a 5-year citation window is that our sample size is reduced, so that our estimation period ends in 2001 rather than 2006. 3.2. Empirical model Our dependent variable is the number of forward citations of a given REN patent by subsequent patents within 5 years after application. Since we want to investigate the differences across categories of spillovers, we will also consider several models with alternative dependent variables. We make a distinction between four types of spillovers defined as follows (see Fig. 1 for an illustration): • the number of forward citations of REN patents to patents within the same technological field: intra-technology spillovers

10 Since our analysis exploits citations of European patents, the findings are restricted to Europe. Given the significant shares of non-European patent filings in REN technological fields, the inclusion of other leading patenting locations (such as Japan or the US) specialized in particular energy fields may lead to a higher share of intra-technology citations in the case where these patents where not filed in Europe afterwards (Lee et al., 2009). 11 Based on the patent literature, the relative shares of examiner and inventor citations differ over patent offices. In particular, different stringency of the requirement on describing the state of the art leads to a smaller share of inventor citations in patents filed at the EPO in comparison to those filed at the USPTO. For example, in the analysis of the sample of US patents by Alcacer and Gittelman (2006), the inventor citation share is 63%, while it is only 9% in the analysis on EPO patents by Criscuolo and Verspagen (2008). Analyzing differences between inventor- and examiner-added citations, Criscuolo and Verspagen (2008) find that inventor citations are more localized than examiner citations both geographically and cognitively; however, cognitive distance appears less important in comparison to the geographic distance. More precisely: “Inventors are also more likely to include citations to patents in the same 4-digit IPC class than examiners, but the difference in means, though statistically significant, is not very large: for the within Europe (US) sample 71% (65%) of inventor citations are to patents in the same technology class, while the corresponding proportion for examiner citations is 68% (62%).”(p.1901). Therefore, the inclusion of examiner citations is likely to cause only a small upward bias of external citation share in our results. 12 Hall et al. (2001) find that on average self-citations represent about 11% of all citations to US patents. For the US patents falling into the energy field, Nemet (2012) reports that 9.8% of records were self-citation pairs. 13 These are the numbers on backward citations. In addition, 10% of patents will receive a forward citation by the same applicant.

Please cite this article in press as: Noailly, J., Shestalova, V., Knowledge spillovers from renewable energy technologies: Lessons from patent citations. Environ. Innovation Soc. Transitions (2016), http://dx.doi.org/10.1016/j.eist.2016.07.004

G Model

ARTICLE IN PRESS

EIST-219; No. of Pages 14

J. Noailly, V. Shestalova / Environmental Innovation and Societal Transitions xxx (2016) xxx–xxx

6

Power generaon technologies considered

FF

Inter-to-FF

Intra-REN

Inter-REN

REN

REN X

REN Y

External

External

Fig. 1. Knowledge spillovers from a renewable technological field, REN X.

• the number of forward citations to patents related to power-generation either in REN or FF energy technological fields – inter-REN or inter-to-FF technology spillovers – but excluding the same technological field • the number of forward citations to patents in all other technological fields that are not covered by our definitions of REN and FF technologies: external technology spillovers Since the number of citations is a count variable and because many patents are never cited, the model will be estimated by a negative binomial regression.

 CITATIONSsi = exp

ˇ0 + ˇ1 APYEARs +



 ˇi TECHFIELDi + εsi

(1)

i ∈ REN

where CITATIONSsi denotes the number of forward citations of patent s from technological field i within a 5-year window, APYEARs is the year of application of the patent which controls for the year fixed effects,TECHFIELDi is a dummy variable which is equal to one if the patent falls into the technological field i, and ␧si is the error term. The inclusion of the year dummies corrects for the differences related to the year of application. The technological field dummies correct for the differences that may affect the specific technologies, such as differences in terms of propensity to patent or in terms of policy support across technological fields.14 Since we exclude one technological field dummy in the estimation, the estimation results can be interpreted as the likelihood of citation relative to a base case patent in the excluded technological field. 4. Results In this section we describe our data and estimation results. Section 4.1 includes some descriptive statistics on patent citations, providing a first picture of knowledge spillovers generated by REN technological fields. Section 4.2 gives the estimation results. 4.1. Descriptive results The dataset that we constructed for our analysis includes 32,099 REN patents filed in Europe in 1978–2001, together with the respective citation numbers of these patents. Among REN-patents included in the dataset, the largest share are solar (37%), storage (27%) and wind (18%) patents, while the other technological fields account for much smaller shares of patents. Table 1 provides the descriptive statistics and gives the average number of REN-patent citations in each of the

14 While the technology dummies may capture the fact that in Europe solar energy received more policy support than marine energy over the period, they will not capture changes in policy support over time. Our estimates may thus be biased by the fact that we do not control for changes in specific renewable policies. Note, however, that the year dummies will capture specific events, such as a rise or drop in energy prices, which would affect all technological fields.

Please cite this article in press as: Noailly, J., Shestalova, V., Knowledge spillovers from renewable energy technologies: Lessons from patent citations. Environ. Innovation Soc. Transitions (2016), http://dx.doi.org/10.1016/j.eist.2016.07.004

G Model

ARTICLE IN PRESS

EIST-219; No. of Pages 14

J. Noailly, V. Shestalova / Environmental Innovation and Societal Transitions xxx (2016) xxx–xxx

7

Table 1 Descriptive statistics on REN patents. Variable

Obs

Number of forward citations within a 5-year window: Total citations 32099 Intra-technology citations 32099 Inter-REN technology citations 32099 Inter-to-FF technology citations 32099 External technology citations 32099 Application year 32099

Mean

Std. Dev.

Min

Max

0.61 0.41 0.01 0.02 0.19 1989

1.74 1.30 0.12 0.30 0.83 7.9

0 0 0 0 0 1978

45 38 6 21 30 2001

Table 2 Descriptive statistics per REN technological field. REN technological field

Total number of patents

Total number of citations

Citations per patent

Solar Storage Wind Waste Marine Hydro Biomass Geothermal FF technological fields

27807 19698 16921 3534 3897 2373 1173 937 102080

18469 12366 11040 1942 2055 935 610 518 48998

0.66 0.62 0.65 0.54 0.52 0.39 0.52 0.55 0.48

100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% solar

stor

wind marine hydro

intra

inter-REN

waste biomass geo

inter-to-FF

external

Fig. 2. Forward citation categories of citing patents per REN technological fields.

four spillovers categories. The average number of REN-patent citations within a five-year window is 0.61.15 About 70% of patents in our initial sample have not received subsequent citations, while there are also patents with multiple citations. Furthermore, when characterizing the allocation of REN-patent citations by type, we observe that almost two thirds are intra-technology citations, and almost one third are external citations, while only a very small share accounts for both inter-REN and inter-to-FF citations. Table 2 shows the distribution of patents and citations across technological fields. Solar, wind and storage energy patents produce the highest shares of citations (above 60% of patents are cited on average), suggesting that these technological fields seem to generate important spillovers. By contrast, other fields such as hydropower or fossil-fuel energy in general generate relatively less citations over the whole sample. Please cite this article in press as: Noailly, J., Shestalova, V., Knowledge spillovers from renewable energy technologies: Lessons from patent citations. Environ. Innovation Soc. Transitions (2016), http://dx.doi.org/10.1016/j.eist.2016.07.004

G Model

ARTICLE IN PRESS

EIST-219; No. of Pages 14

J. Noailly, V. Shestalova / Environmental Innovation and Societal Transitions xxx (2016) xxx–xxx

8

Table 3 Distribution of forward citations per technological field. Destination

Origin Solar Storage Wind Waste Marine Hydro Biomass Geo

Solar

Storage

Wind

Waste

Marine

Hydro

Biomass

Geo

FF

Other

62 0 2 0 1 1 0 5

0 59 0 0 0 0 0 0

1 0 82 0 11 13 0 3

0 0 0 28 0 0 13 0

0 0 3 0 75 13 0 1

0 0 2 0 7 37 0 0

0 0 0 4 0 0 48 0

0 0 0 0 0 0 0 52

2 0 1 59 1 3 16 7

36 41 15 29 15 43 37 37

Source: Own computations, based on the sample of patents with at least one forward citation. The central part of the table gives the percentage of forward citations from each origin technology (row) to each destination technology (column). Each row sums up to 100%. The total number of citations corresponding to 100% is reported in Table 2.

Regarding the direction of knowledge spillovers, Fig. 2 provides some first insights on the distribution of forward citations of REN patents to other technological fields, shedding the light on notable differences between various REN technological fields. Overall, most forward citations are found in the same technological field, indicating that REN patents often find applications within the same fields. As shown in Fig. 2, the share of intra-technology spillovers is high for wind (80%) and marine patents (70%), medium for solar and storage patents (around 60%) and lowest for waste patents (25%).16 Hence, past innovation in wind, marine, solar and storage contributes a great deal to current innovation in these specific technological fields, possibly indicating some form of path-dependency in knowledge creation. Looking at the share of inter-technology spillovers – i.e., spillovers to other related technological fields of power generation: either REN (inter-REN) or FF (inter-to-FF), we find that hydro and marine spill over mainly to other REN technological fields: about 10% of their forward citations are in other REN energy patents. Instead, biomass and waste mainly spill over to FF power-generation technological fields. By contrast, solar, wind and storage generate only very limited inter-technology spillovers, either to REN or FF technological fields. The last notable result that emerges from Fig. 2 is the relatively high share of external technology spillovers in particular for hydropower, storage, solar, and geothermal technological fields (about 40%). By contrast, the share of forward citations to external technology is the lowest for wind and marine technological fields where the share is below 20%. Table 3 gives some more detailed insights on inter-technology citations for each REN technological field. As an illustration, the first row of Table 3 shows which percentage of forward citations of solar patents goes to which field: 62% of these citations are allocated to solar patents; 1% of citations to wind patents; 2% to patents in FF technological fields and 36% to patents in other technological fields. Table 3 illustrates that indeed 11% and 7% of forward citations of marine patents are found in wind and hydropower patents respectively. Similarly, wind patents are cited by patents in hydropower and marine fields, suggesting that these three technological fields are intertwined as they seem to partly rely on the same knowledge base. In addition, 59% of the citations of waste patents are found in FF power-generation fields. This suggests that, as we expected, waste-to-energy and FF fields rely on the same type of knowledge because technologies developed to burn one type of fuel (such as coal) may also be used to burn another type of fuel (namely waste or biomass). This has for instance led to the development of co-firing techniques, using biomass and waste as supplementary fuel in coal and gas electricity generators and boilers (e.g., Maciejewska et al., 2006). Our descriptive results also show that REN technological fields generate external-technology spillovers, outside the field of power-generation technologies. In order to investigate which fields benefit from the knowledge in specific REN technological fields, we classify the external citations into technological fields according to the WIPO Technology Concordance Table linking the International Patent Classification symbols with 35 fields (Schmoch, 2008). Table 4 illustrates the results. We find that solar patents are mainly cited by other patents in the field of semiconductors, thermal processes and apparatus and civil engineering. Wind patents, but also marine and hydropower patents, are mainly cited by other patents in the field of electrical machinery, engines, pumps and turbines, mechanical elements, and transport; while storage patents are mainly cited by inventions in electrical machinery. Finally, waste and biomass patents find applications into the fields of basic materials chemistry, chemical engineering and environmental technology patents. Links across technological fields can be explained by the fact that technologies combining knowledge from one or more technological fields are by definition ‘technically close’ to these fields. According to Breschi et al. (2003), the concept of

15 We do not include descriptive statistics on FF patents since our analysis focuses mainly on REN patents. Yet, we also constructed a dataset on FF technological fields, which includes 102,080 FF patents filed over the 1978–2001 period. The average number of FF-patent citations is 0.48. In Section 4.2, we provide estimation results where we compare the citations patterns of REN technological fields with an average FF-patent. 16 Braun et al. (2010) also finds that intra-technology spillovers play a greater role for innovation in wind than in solar technologies.

Please cite this article in press as: Noailly, J., Shestalova, V., Knowledge spillovers from renewable energy technologies: Lessons from patent citations. Environ. Innovation Soc. Transitions (2016), http://dx.doi.org/10.1016/j.eist.2016.07.004

G Model

ARTICLE IN PRESS

EIST-219; No. of Pages 14

J. Noailly, V. Shestalova / Environmental Innovation and Societal Transitions xxx (2016) xxx–xxx

9

Table 4 External technological fields receiving the highest (>10%) spillovers from REN technologies. % external spillovers Solar Semiconductors Civil engineering Thermal processes and apparatus

13 12 11

Wind Electrical machinery Transport Engines, pumps, turbines Mechanical elements

24 20 14 10

Storage Electrical machinery

61

Marine Engines, pumps and turbines Electrical machinery Transport

40 21 12

% external spillovers Hydro Engines, pumps, turbines Civil engineering Electrical machinery Mechanical elements

25 18 16 13

Biomass Basic materials chemistry Environmental technology Chemical engineering Machine tools

34 11 10 10

Waste Basic materials chemistry Environmental technology Chemical engineering

21 19 18

Source: Own computation based on the set of citations included in REN patents.

“technological relatedness” can be defined by the observation that “certain groups of technologies share a common or complementary knowledge base, rely on common scientific principles or have similar heuristics of search” (Breschi et al., 2003; p.69). Solar technologies for instance are technically related to semiconductor technologies and, therefore, there are interactions between the two knowledge bases (Nemet, 2012). Also, technologies gasifying waste to produce electricity are directly related to FF engine and gas turbine technologies. Second, knowledge spillovers across various technological fields are more likely to arise within large firms seeking to exploit scope economies. Noailly and Smeets (2015) show for instance that innovating firms which patent in both fossil-fuel and renewable energy tend to be larger than firms specialized in only one type of technology. Hence, large technology firms which have specialized in fossil-fuel technologies in the past may want to invest in waste technologies, as the two types of technologies present complementarities. Although illustrative, this initial descriptive analysis presents several limitations. First, the discussion of the direction of spillover effects focused on allocation of citations in the set of citing-cited patent pairs, and therefore did not account for non-citing patents. Second, it did not control for the effect of time on the development of different technologies and for the fact that more recent patents had less years to be cited within the period studied. Hence, in the next section we turn to regression analysis to correct for these issues and compare the spillovers across REN technological fields, all other things being equal. 4.2. Estimation results Table 5 shows the results of our estimations for Eq. (1). Each column in Table 5 corresponds to a different model based on each specific dependent variable, characterizing the number of citations in a particular category within a five-year citation window after a patent application. The estimation is conducted at the patent level for the complete sample of patents, including patents without any forward citations. In Panel A, we conduct our estimations on the sample of REN patents only. The estimation excludes the dummy variable for solar patents, so that the results are interpreted in comparison with an average solar patent. This panel provides detailed analysis of spillover effects from different REN technological fields, following the descriptive analysis in Fig. 1. In Panel B, we provide additional results where we interpret the results in comparison with an average FF patent. This panel covers two types of citations: total citations and external citations, providing some additional insights into differences between REN and FF patents. All the results are presented in the exponential form: a coefficient above one means that the patents of this technological field generate more forward citations compared to the average base case patent. For example, the exponentiated estimated coefficient of the wind technology dummy reported in column (1) – equal to 1.099 – means that wind patents are 9.9% more likely to be cited than solar patents. The coefficient value is obtained as exp(0.094), where 0.094 is the original non-exponentiated estimate of the model coefficient < MML : MSUB > ␤wind < /MML : MSUB > in Eq. (1). The results presented in column (1) in panel A help us to address our first research question (“which renewable energy technological fields generate the most knowledge spillovers?”). We find that wind patents are on average about 10% more likely to be cited than solar patents all other things being equal. Storage patents in particular are 40% more likely to be cited than an average solar patent. Instead, all other REN patents are less likely to be cited than solar patents. In panel B, we provide some additional insights by comparing our set of REN patents with patents in FF technologies. Column (6) shows that some REN technological fields are more likely to be cited than an average FF patent. The significant coefficients above Please cite this article in press as: Noailly, J., Shestalova, V., Knowledge spillovers from renewable energy technologies: Lessons from patent citations. Environ. Innovation Soc. Transitions (2016), http://dx.doi.org/10.1016/j.eist.2016.07.004

G Model

ARTICLE IN PRESS

EIST-219; No. of Pages 14

J. Noailly, V. Shestalova / Environmental Innovation and Societal Transitions xxx (2016) xxx–xxx

10 Table 5 Estimation results.

PANEL A: Comparison to an average solar patent

(1)

PANEL B: Comparison to an average FF patent

(2)

(3)

(4)

(5)

(6)

(7)

Intra-technology citations

Inter-REN technology citations

Inter-FF technology citations

External technology citations

Total citations

External technology citations

1.099** (0.045) 1.403*** (0.053) 0.520*** (0.039) 0.413*** (0.039) 0.799*** (0.057) 0.593*** (0.067) 0.440*** (0.068)

1.481*** (0.066) 1.376*** (0.059) 0.582*** (0.047) 0.271*** (0.032) 0.869 (0.075) 0.472*** (0.068) 0.348*** (0.076)

4.500*** (0.809) 0.192*** (0.111) 6.699*** (1.456) 5.193*** (1.348) 1.726* (0.553) 5.787*** (1.847) 1.200 (0.940)

0.451*** (0.118) 0.018*** (0.013) 0.312** (0.154) 0.297** (0.170) 22.730*** (3.333) 2.139*** (0.626) 1.949 (1.001)

0.454*** (0.034) 1.517*** (0.082) 0.314*** (0.044) 0.598*** (0.086) 0.585*** (0.060) 0.771 (0.124) 0.637** (0.135)

Constant

0.510*** (0.026)

0.383*** (0.021)

0.003*** (0.001)

0.006*** (0.002)

0.115*** (0.012)

1.378*** (0.052) 1.706*** (0.053) 0.614*** (0.045) 0.477*** (0.044) 0.929 (0.062) 0.673*** (0.076) 0.509*** (0.077) 1.198*** (0.028) 0.408*** (0.013)

0.508*** (0.034) 1.686*** (0.071) 0.339*** (0.046) 0.633*** (0.088) 0.635*** (0.062) 0.801 (0.125) 0.675* (0.138) 1.080* (0.042) 0.123*** (0.007)

Alpha

5.075*** (0.095) −29933 32099 1.87

6.127*** (0.133) −23536 32099 1.47

30.432*** (5.686) −1387 32099 0.09

15.509*** (2.283) −2134 32099 0.13

9.181*** (0.295) −14225 32099 0.89

5.876*** (0.058) −111948 132198 1.69

12.263*** (0.197) −56286 132198 0.85

Dependent Total citations variable: number of forward citations Wind Storage Marine Hydro Waste Biomass Geothermal Solar

log likelihood Observations AIC/N

Citations: 5 years windows. The results of the negative binomial regression are presented as exp(ˇ) and are interpreted as the likelihood of receiving a citation, compared to the base case of a solar patent in Panel A and an FF-patent in panel B. Robust standard errors in parentheses: *p < 0.10 **p < 0.05 ***p < 0.01. To gauge and compare the models’ performance, the final row reports the Akaike’s Information Criterion (AIC) divided by the number of observations. The models for total citations, intra- and external-technology citations perform better than the models for inter-technology citations.

1 for wind, storage and solar fields show that these patents are more cited – and thus more important – than an average FF patent. To address our second research question (“where do knowledge spillovers generated by renewable technological fields flow to?”), in column (2) we only consider the forward citations in the same technological fields, i.e. the intra-technology spillovers. Here again, compared to solar patents, wind and storage are about 40% more likely to be cited within the same technological field, while other REN technologies such as marine, hydro, waste, biomass and geothermal do not lead to much intra-technology spillovers. Columns (3) and (4) give the likelihood of being cited in other REN or FF power-generation technological fields, respectively. In column (3) only storage patents do not lead to more inter-REN spillovers than solar.17 Knowledge from other technological fields, such as wind, marine, biomass and hydropower, is particularly beneficial to other REN-technological fields. In column (4), we find that waste patents are much more likely to be cited by patents in fossil-fuel energy generation technological fields, than solar patents. This is also the case to a lower extent for patents in biomass. Overall, the results on inter-technology spillovers confirm some of the insights learnt from Fig. 2, namely that the knowledge from waste, biomass, hydropower and marine technological fields is beneficial for other related technological fields in power generation. Finally, column (5) gives the results for external technology citations. Compared to solar patents, storage patents are about 50% more likely to be cited by patents of external technological fields. Patents from the other technological fields will be less likely to be cited outside the REN and FF fields than solar patents. The estimation results suggest that innovation in hydropower and geothermal technological fields leads to much less external spillovers than what we could first infer by looking at Fig. 2. Hence, after correcting for the application year and truncation effects, hydropower and geothermal technology patents do not appear to be more cited externally than an average solar patent. Finally, the results in column (7) of panel B show that, during the period studied, only solar and storage patents generate substantial spillovers to external fields, exceeding the external spillovers of an average FF patent. We can summarize our estimation results on the knowledge spillovers from REN technologies as follows:

17 Note that some of coefficients in columns (3) and (4) are very large, reflecting the fact that solar patents do not generate much inter-technology spillovers within the field of energy generation and storage technologies.

Please cite this article in press as: Noailly, J., Shestalova, V., Knowledge spillovers from renewable energy technologies: Lessons from patent citations. Environ. Innovation Soc. Transitions (2016), http://dx.doi.org/10.1016/j.eist.2016.07.004

G Model EIST-219; No. of Pages 14

ARTICLE IN PRESS J. Noailly, V. Shestalova / Environmental Innovation and Societal Transitions xxx (2016) xxx–xxx

11

(1) Wind, storage and solar patents tend to be frequently cited (compared to other renewable and fossil-fuel patents), suggesting that inventions in these technological fields are particularly important for society. (2) Wind and storage technological fields are characterized by important intra-technology spillovers, more than solar or other renewable fields. Past technological advances in these specific technological fields have been particularly useful to develop current inventions in these domains. (3) Solar and storage technological fields exhibit large external-technology spillovers compared to other fields. Knowledge from these technologies finds applications in technological fields outside the field of power generation as these patents exhibit a high likelihood of citations by patents from external technological fields. (4) The knowledge base from wind, hydropower, marine and biomass seem to be intertwined with other REN technological fields. They, however, deliver less contribution to technological fields outside the field of power-generation, in comparison to an average solar patent or an average FF patent. (5) Waste and biomass mainly find applications in fossil-fuel power-generation. Past advances in waste and biomass technological fields have been useful in developing recent knowledge in FF technological fields. However, knowledge from these technological fields does not find broad application outside the field of power-generation, less than for an average solar patent and an average FF patent. 5. Discussion and policy implications Policymakers have a diverse set of tools to design public policies to stimulate innovation in renewable energy technologies. Technological change can be triggered either by the demand side (“demand-pull policies”, such as feed-in tariffs) or the supply side (“technology-push policies” such as R&D support). Both types of policies interact to stimulate the rate and direction of technological change and scholars now acknowledge that both types of instruments are necessary to induce energy technological change (Jaffe et al., 2005). In recent work, Acemoglu et al. (2012) recommend to combine a carbon tax with an increase in R&D-support for clean innovation, in a strategy to increase both the demand and supply of clean energy innovation. Yet, policymakers are often confronted with the more specific question on how to allocate funds to individual technologies. Our results make it possible to differentiate among the various REN technological fields as our findings show that the extent of knowledge spillovers varies across technological fields. Wind, storage and solar fields tend to generate more spillovers, as patents in these technological fields are more likely to be cited, than other renewable (or fossil-fuel) technologies. Our study show that inventions in solar and storage energy tend to be particularly important: they receive a large number of citations and find applications in a large set of diverse fields – combining two characteristics of highly valuable innovations (Lanjouw and Schankerman, 2004; Popp and Newell, 2012), which might warrant larger policy support for these technological fields. Our results on the direction of knowledge flows also help to inform policymakers on how the R&D support for renewable technology can be designed. Aalbers et al. (2013) argue that the strength of the argument for technology-specific R&D support depends (among other things) on the size of spillover effects between REN and FF technological fields. In particular, the justification for R&D support is weaker for renewable technological fields that are characterized by larger knowledge spillovers to fossil-fuel technological fields. In contrast, technological fields with larger contribution to the REN knowledge base may be eligible for public R&D. Therefore, the empirical evidence regarding the direction of various knowledge spillover effects of different REN technologies reported in our study provides a concrete tool for the design of innovation policies in the power-generation sector. The magnitude of intra-technology spillovers tells us how powerful the innovation machine is for each specific technological field (Veugelers et al., 2009). For wind, once the stock of wind inventions is large enough, specific innovation subsidies may no longer be needed since the technological field will benefit from large intra-technology spillovers, ensuring that these technologies will continue to develop fast. Solar and storage technological fields might instead – ceteris paribus – need longer policy support as intra-technology spillovers are less strong than for wind. Marine energy benefits from a well-functioning innovation machine as this technological field is also building upon the past knowledge stocks of wind, marine and hydropower. Only specific temporary policy support will probably be needed for these technological fields. Waste and biomass technological fields present characteristics that are very different from wind, storage or solar fields and more similar to most FF technological fields. Indeed, we find that the knowledge from waste mainly flows to other FF technological fields, and not much to other REN or to technological fields unrelated to power generation. New inventions in these fields find applications mainly in FF technological fields. Hence, policy support for waste and biomass may contribute to further increases of the FF knowledge base and the productivity gap between FF and other REN technologies.18 Also, the results on citations suggest that knowledge in waste and biomass is on average less important – ceteris paribus – than knowledge from other REN technological fields and their contribution to technological fields outside the energy fields considered is less than that of other energy technologies. If the policy goal is purely to stimulate a transition away from FF power generation, then public policy does not need to be directed to these technological fields. Even generic innovation

18

This reasoning is supported by the theoretical model by Acemoglu et al. (2012).

Please cite this article in press as: Noailly, J., Shestalova, V., Knowledge spillovers from renewable energy technologies: Lessons from patent citations. Environ. Innovation Soc. Transitions (2016), http://dx.doi.org/10.1016/j.eist.2016.07.004

G Model

ARTICLE IN PRESS

EIST-219; No. of Pages 14

12

J. Noailly, V. Shestalova / Environmental Innovation and Societal Transitions xxx (2016) xxx–xxx

policy would simply encourage future developments of FF technological fields, rather than REN or other technological fields useful to society. While our focus in this study lies on innovations contributing to knowledge spillovers in the context of climate change mitigation, other policy considerations (such as security of supply or risk of crowding out between energy and non-energy patents) may need to be taken into account in the integral policy framework. 6. Conclusions This paper unravels the knowledge spillovers accompanying the development of renewable energy technological fields by analyzing several technological fields and four types of spillovers (intra-, inter-REN, inter-to-FF, and external-technology spillovers). Using an empirical analysis of citations of renewable patents filed in Europe, we show that knowledge from renewable energy technological fields generate important knowledge spillovers. Patents in wind, storage and solar fields tend to be frequently cited, suggesting that these technological fields are particularly important and valuable for society. Regarding the direction of knowledge flows, we find that on average about 60% of forward citations of renewable patents are from patents in the same technological field. Yet, the size of intra-technology spillovers differs significantly between technologies. The three main technological fields in renewable energy, namely solar, wind and storage fields, exhibit high levels of intra-technology spillovers. Our results also show that innovations in solar and storage hardly contribute to the knowledge base of other power-generation technological fields (inter-technology spillovers), while having large spillovers to technological fields outside of power-generation (external spillovers). Instead, knowledge from wind, hydropower and marine technological fields appear to be intertwined as these fields benefit a lot from each other knowledge base. Finally, the analysis shows that waste technological field contributes mainly to fossil-fuel technological fields. Our results lend support to the idea of differentiating policy support per type of renewable technological field. Regarding the specific design of policies, our findings can help policymakers in identifying for which technological fields it may be more appropriate to support the development of the technology specific knowledge stock (to leverage the incremental nature of technological change in technologies for which intra-technology spillovers are important) and for which fields it may be more appropriate to maintain diversity in the technological pool. Future research could check the robustness of our results by reproducing the analysis for more recent years and additional countries outside Europe. In addition, extending the analysis to account for policy developments in the period studied would help refine conclusions with respect to policy design to support the development of renewable electricity technologies. Appendix A.

Table A1 Classification into technology classes for Renewable Energy Generation Technologies. Technology

Description

IPC classes

Wind power Solar energy

Wind motors Devices for producing mechanical power from solar energy Use of solar heat, e.g. solar heat collectors Drying solid materials or objects by processes involving the application of heat by radiation − e.g. from the sun Devices consisting of a plurality of semiconductor components sensitive to infra-red radiation, light − specially adapted for the conversion of the energy of such radiation into electrical energy Semiconductor devices sensitive to infra-red radiation, light, electromagnetic radiation of shorter wavelength, or corpuscular radiation, specially adapted as devices for the conversion of the energy of such radiation into electrical energy, including a panel or array of photoelectric cells, e.g. solar cells Generators in which light radiation is directly converted into electrical energy Devices for producing mechanical power from geothermal energy Production or use of heat, not derived from combustion − using geothermal heat Tide or wave power plants Submerged units incorporating electric generators or motors characterized by using wave or tide energy Ocean thermal energy conversion Water-power plants; Layout, construction or equipment, methods of, or apparatus for; and not Tide or wave power plants Machines or engines for liquids of reaction type; Water wheels; Power stations or aggregates of water-storage type; Machine or engine aggregates in dams or the like; Controlling machines or engines for liquids; and NOT Submerged units incorporating electric generators or motors characterized by using wave or tide energy

F03D F03G6 F24J2 F26B3/28

Geothermal energy Marine (ocean) energy

Hydro power

H01L27/142

H01L31/042-058

H02N6 F03G4 F24J3/08 E02B9/08 F03B13/10–26 F03G7/05 E02B9; and not E02B9/08 [F03B3 or F03B7 or F03B13/06-08 or F03B15] and not F03B13/10-26

Please cite this article in press as: Noailly, J., Shestalova, V., Knowledge spillovers from renewable energy technologies: Lessons from patent citations. Environ. Innovation Soc. Transitions (2016), http://dx.doi.org/10.1016/j.eist.2016.07.004

G Model

ARTICLE IN PRESS

EIST-219; No. of Pages 14

J. Noailly, V. Shestalova / Environmental Innovation and Societal Transitions xxx (2016) xxx–xxx

13

Table A1 (Continued) Technology

Description

IPC classes

Biomass energy

Solid fuels based on materials of non-mineral origin – animal or vegetable substances Engines or plants operating on gaseous fuels from solid fuel – e.g. wood Solid fuels based on materials of non-material origin – sewage, town, or house refuse; industrial residues or waste materials Incineration of waste – recuperation of heat Incinerators or other apparatus consuming waste – field organic waste Liquid carbonaceous fuels; Gaseous fuels; Solid fuels; and Dumping solid waste; Destroying solid waste or transforming solid waste into something useful or harmless; Incineration of waste; Incinerator Plants for converting heat or fluid energy into mechanical energy – use of waste heat; Profiting from waste heat of combustion engines; Machines, plant, or systems, using particular sources of energy – using waste heat. And Incineration of waste; Incinerator constructions; Incinerators or other apparatus specially adapted for consuming specific waste or low grade fuels. Lead-acid accumulators gastight accumulators Alkaline accumulators Gastight accumulators Other types of accumulators not provided for elsewhere

C10L5/42–44 F02B43/08 C10L5/46–48

Waste-to-energy

Storage

F23G5/46 F23G7/10 [C10L1 or C10L3 or C10L5] and [B09B1 or B09B3 or F23G5 or F23G7] [F01K27 or F02G5 or F25B27/02] and [F23G5 or F23G7]

H01M10/06-18 H01M10/24–32 H01M10/34 H01M10/36–40

Sources: Johnstone et al. (2010) and Johnstone and Haˇscˇ iˇc (2010) for storage technologies. Table A2 Classification into IPC classes for Fossil-Fuel Energy Generation Technologies. Technology

Description

IPC classes

Coal Engines

Production of fuel gases by carburetting air or other gases without pyrolysis Steam engine plants; steam accumulators; engine plants not otherwise provided for; engines using special working fluids or cycles Gas-turbine plants; air intakes for jet-propulsion plants; controlling fuel supply in air-breathing jet-propulsion plants Hot-gas or combustion-product positive-displacement engine; Use of waste heat of combustion engines, not otherwise provided for Steam generation Combustion apparatus; combustion processes Furnaces; kilns; ovens; retorts [Classes listed below excluding combinations with B60, B68, F24, F27] Engines characterised by fuel-air mixture compression ignition Engines characterised by air compression and subsequent fuel addition; with compression ignition Engines characterised by the fuel-air charge being ignited by compression ignition of an additional fuel Engines characterised by both fuel-air mixture compression and air compression, or characterised by both positive ignition and compression ignition, e.g. in different cylinders Engines characterised by the introduction of liquid fuel into cylinders by use of auxiliary fluid; Compression ignition engines using air or gas for blowing fuel into compressed air in cylinder Methods of operating air-compressing compression-ignition engines involving introduction of small quantities of fuel in the form of a fine mist into the air in the engine’s intake.

C10J F01K

Turbines Hotgas Steam Burners Furnaces Ignition

F02C F02G F22 F23 F27 F02B1/12–14 F02B3/06-10 F02B7 F02B11

F02B13/02-04

F02B49

Source: Lanzi et al. (2011) and Haˇscˇ iˇc et al. (2009). We thank Ivan Haˇscˇ iˇc for providing us the last updated version of fossil-fuels IPC codes.

References Aalbers, R., Shestalova, V., Kocsis, V., 2013. Innovation policy for directing technical change in the power sector. Energy Policy 63, 1240–1250. Acemoglu, D., Aghion, P., Bursztyn, L., Hemous, D., 2012. The environment and directed technical change. Am. Econ. Rev. 102 (1), 131–166. Alcacer, J., Gittelman, M., 2006. Patent citations as a measure of knowledge flows: the influence of examiner citations. Rev. Econ. Stat. 88, 774–779. Arthur, W.B., 1989. Competing technologies, increasing returns and lock-in by historical events. Econ. J. 99, 116–131. Battke, B., Schmidt, T.S., Stollenwerk, S., Hoffman, V.H., 2016. Internal or external spillovers—which kind of knowledge is more likely to flow within or across technologies? Res. Policy 45, 27–41. Braun, F.G., Schmidt-Ehmcke, J., Zloczysti, P., 2010. Innovative Activity in Wind and Solar Technology: Empirical Evidence on Knowledge Spillovers Using Patent Data, Discussion Paper 993. German Institute for Economic Research, Germany. Breschi, S., Lissoni, F., Malerba, F., 2003. Knowledge-relatedness in firm technological diversification. Res. Policy 32, 69–87. Cohen, W.M., Levinthal, D.A., 1990. Absorptive capacity: a new perspective on learning and innovation. Adm. Sci. Q. 35, 128–152. Criscuolo, P., Verspagen, B., 2008. Does it matter where patent citations come from? Inventor vs. examiner citations in European patents. Res. Policy 37, 1892–1908. David, P.A., 1985. Clio and the economics of QWERTY. Am. Econ. Rev., 332–337. Dechezleprêtre, A., Martin, R., Mohnen, M., 2013. Knowledge Spillovers from Clean and Dirty Technologies: A Patent Citation Analysis. Working Paper, LSE. Dosi, G., 1982. Technological paradigms and technological trajectories: a suggested interpretation of the determinants and directions of technical change. Res. Policy 11 (3), 147–162. European Energy Agency (EEA), 2012. Overview of the electricity production and use in Europe.

Please cite this article in press as: Noailly, J., Shestalova, V., Knowledge spillovers from renewable energy technologies: Lessons from patent citations. Environ. Innovation Soc. Transitions (2016), http://dx.doi.org/10.1016/j.eist.2016.07.004

G Model EIST-219; No. of Pages 14

14

ARTICLE IN PRESS J. Noailly, V. Shestalova / Environmental Innovation and Societal Transitions xxx (2016) xxx–xxx

Gillingham, K., Sweeney, J., 2012. Barriers to implementing low carbon technologies. Clim. Change Econ. 3 (04). Griliches, Z., 1979. Issues in assessing the contribution of research and development to productivity growth. Bell J. Econ. 10, 92–116. Haˇscˇ iˇc, I., Johnstone, N., Lanzi, E., 2009. The determinants of innovations in electricity generation technologies, working paper OECD. Hall, B.H., Jaffe, A.B., Trajtenberg, M., 2001. The NBER Patent Citation Data File: Lessons, Insights and Methodological Tools. NBER Working Paper 8498. Huenteler, J., Schmidt, T.S., Ossenbrink, J., Hoffmann, V.H., 2016. Technology life-cycles in the energy sector—Technological characteristics and the role of deployment for innovation. Technol. Forecasting Social Change 104, 102–121. International Energy Agency (IEA), 2014, Technology Roadmap Energy storage, 2014 Edition, Paris. International Energy Agency (IEA), 2015. Renewable energy Medium-Term Market Report 2015: Market Analysis and Forecasts to 2020, 2015 Edition, Paris. Jaffe, A.B., de Rassenfosse, G., 2016. Patent citation data in social science research: Overview and best practices, NBER Working Paper, No 21868. Jaffe, A.B., Trajtenberg, M., Henderson, R., 1993. Geographic localization of knowledge spillovers as evidenced by patent citations. Q. J. Econ. 108 (3), 577–598. Jaffe, A.B., Trajtenberg, M., Forgarty, M., 2000. Knowledge spillovers and patent citations: evidence from a survey of inventors. Am. Econ. Rev. 90 (2), 215–218. Jaffe, A.B., Newell, R.G., Stavins, R.N., 2005. A tale of two market failures: technology and environmental policy. Ecol. Econ. 54, 164–174. Johnstone, N., Haˇscˇ iˇc, I., 2010. Directing Technological Change while Reducing the Risk of (not) Picking Winners: The Case of Renewable Energy, OECD working paper. Johnstone, N., Haˇscˇ iˇc, I., Popp, D., 2010. Renewable energy policies and technological innovation: evidence based on patent counts. Environ. Resour. Econ. 45 (1), 133–155. Lanjouw, J.O., Schankerman, M., 2004. Patent quality and research productivity: measuring innovation with multiple indicators*. Economic Journal 114 (495), 441–465. Lanzi, E., Verdolini, E., Haˇscic, I., 2011. Efficiency Improving Fossil Fuel Technologies for Electricity Generation: Data Selection and Trends, Working Papers 2011.10, Fondazione Eni Enrico Mattei. Lee, B., Iliev, I., Preston, F., 2009. Who Owns Our Low Carbon Future? Intellectual Property and Energy Technologies. A Chatham House Report. Maciejewska, A., Veringa, H., Sanders, J., Peteves, S.D., 2006. Co-firing of biomass with coal: constraints and role of biomass pre-treatment, EUR 22461 EN, report by DG JRC and Institute for Energy (Petten, The Netherlands). Mowery, D.C., Rosenberg, N., 1998. Paths of Innovation: Technological Change in 20th-century America. Cambridge University Press, Cambridge. Nelson, R.R., Winter, S.G., 1982. An Evolutionary Theory of Economic Change. The Belknap Press of Harvard University Press, Cambridge, MA. Nemet, G., Johnson, E., 2012. Do important inventions benefit from knowledge originating in other technological domains? Res. Policy 41 (1), 190–200. Nemet, G., 2012. Inter-technology knowledge spillovers for energy technologies. Energy Econ. 34 (5), 1259–1270. Noailly, J., Shestalova, V., 2013. On which technologies do renewable energy innovations build on? An analysis using backward patent citations, CPB Background Document. Noailly, J., Smeets, R., 2015. Directing technical change from fossil-fuel to renewable energy innovation: an empirical investigation using patent data. J. Environ. Econ. Manag. 72 (C), 15–37. Popp, D., Newell, R.G., 2012. Where does energy R&D come from? Examining crowding out from environmentally-friendly R&D. Energy Econ. 34 (4), 980–991. Popp, D., 2002. Induced innovation and energy prices. Am. Econ. Rev. 92, 160–180. Rosenberg, N., 1994. Exploring the Black Box: Technology, Economics and History. Cambridge niversity Press, Cambridge. Schmoch, U., 2008. Concept of a Technology Classification for Country Comparisons, final report to the World Intellectual Property Organization (WIPO). Schoenmakers, W., Duysters, G., 2010. The technological origins of radical inventions. Res. Policy 39, 1051–1059. Trajtenberg, M., Henderson, R., Jaffe, A., 1997. University versus corporate patents: a window on the basicness of invention. Econ. Innov. New Technol. 5 (1), 19–50. Trajtenberg, M., 1990. A penny for your quotes: patent citations and the value of innovations. Rand J. Econ. 21 (1), 172–187. Unruh, G.C., 2002. Escaping carbon lock-in. Energy Policy 30, 317–325. Usher, A.P., 1954. A History of Mechanical Inventions, Revised Edition. Harvard University Press, Cambridge, MA. van den Bergh, J.C.J.M., 2008. Optimal diversity: increasing returns versus recombinant innovation. J. Econ. Behav. Org. 68, 565–580. Veugelers, R., Aghion, P., Hemous, D., 2009. Kick-starting the Green Innovation Machine. Voyeur (December 2009).

Please cite this article in press as: Noailly, J., Shestalova, V., Knowledge spillovers from renewable energy technologies: Lessons from patent citations. Environ. Innovation Soc. Transitions (2016), http://dx.doi.org/10.1016/j.eist.2016.07.004