Policy mixes for sustainability transitions: New approaches and insights through bridging innovation and policy studies

There has been an increasing interest in science, technology and innovation policy studies in the topic of policy mixes. While earlier studies conceptualised policy mixes mainly in terms of combinations of instruments to support innovation, more recent literature extends the focus to how policy mixes can foster sustainability transitions. For this, broader policy mix conceptualisations have emerged which also include considerations of policy goals and policy strategies; policy mix characteristics such as consistency, coherence, credibility and comprehensiveness; as well as policy making and implementation processes. It is these ‘policy mixes for sustainability transitions’ which are the subject of a special issue introduced in this article. With this special issue we aim at supporting the emergence of a new strand of interdisciplinary social science research on policy mixes which combines approaches, methods and insights from innovation and policy studies to further such broader policy mix research fostering sustainability transitions. Our introduction presents a bibliometric analysis of the literature on policy mixes in both fields and their emerging connections. We also introduce five major themes in the policy mix literature and summarise the contributions made by the articles in the special issue to these: methodological advances; policy making and implementation; actors and agency; evaluating policy mixes; and the co-evolution of policy mixes and socio-technical systems. We conclude by summarising key insights for policy making.


Introduction
It has long been acknowledged that a combination of technology push and demand pull instruments is required for stimulating innovation (Di Stefano et al., 2012).In addition, systemic instruments have been proposed to complement more traditional innovation policies (Borrás and Edquist, 2013;Smits and Kuhlmann, 2005;Wieczorek and Hekkert, 2012).
Indeed, the topic of policy mixes has received increasing interest in innovation studies over the last five years, with several contributions published in Research Policy (Flanagan et al., 2011;Kivimaa and Kern, 2016;Rogge and Reichardt, 2016) and other innovation studies journals (Borrás and Edquist, 2013).However, there is no uniform definition of what constitutes a policy mix for innovation (Rogge and Reichardt, 2016), with innovation scholars initially focusing on "[t]he combination of policy instruments, which interact to influence the quantity and quality of R&D investments in public and private sectors (Nauwelaers et al., 2009, p. 3).This emerging literature on policy mixes for innovation showcases a remarkable diversity.First, several empirical areas have been studied, such as R&D subsidies and tax credits (Neicu et al., 2016), research tax credits (Liu, 2013), energy efficiency technologies (Costantini et al., 2017;Rosenow et al., 2017), renewable energy technologies (Quitzow, 2015;Reichardt et al., 2016), regional innovation systems (Magro and Wilson, 2013) or small business innovation research programs (Lanahan and Feldman, 2015).Second, studies have utilized a variety of different methodologies, including company innovation surveys (Rogge and Schleich, 2018), patent and network analysis (Cantner et al., 2016), case studies (Reichardt and Rogge, 2016;Uyarra et al., 2016), mapping of policy mixes (Kivimaa and Kern, 2016), and tracing changes in mixes over time (Kern et al., 2017).Finally, the focus of research has varied, ranging from questions of instrument interactions and additionality (Neicu et al., 2016), to policy mix characteristics such as balance and comprehensiveness (Costantini et al., 2017), and the multilevel nature of policy mixes (Magro and Wilson, 2013).This academic interest is mirrored by an increasing recognition of policy makers that it is useful to view innovation policy through the lens of policy mixes.One example is a study commissioned by the European Commission to investigate which mixes of policies are most effective to increase the quantity and performance of research investments (Nauwelaers et al., 2009).Another example is the OECD Science, Technology and Innovation Outlook 2016 which has a chapter on 'Policy mix for business R&D and innovation' (OECD, 2016).The report understands policy mixes as being comprised of policy rationales, arrangements and instruments implemented in a specific policy domain, and sees interactions between instruments as central.A third example is the report of the International Energy Agency (2017) on 'Real-world policy packages for sustainable energy transitions' which acknowledges that a variety of different instruments need to be combined into coherent policy packages, while pointing out that there is no one "right" policy package as national objectives and constraints differ.In addition, strategic sequencing of policies is seen as important, and layering of policy incentives to ensure achievement of a certain target is provided as one reason to justify using multiple policies in a situation where constraints do not permit a single policy.This interest represents a real advance over earlier policy discussions about the optimal selection of (individual) policy instruments based on criteria such as economic efficiency or costeffectiveness (e.g.Goulder and Parry, 2008;Sterner and Coria, 2012), but also raises many new questions for further academic work to which there may not be simple or immediate answers.Within these emerging academic and policy discussions, it was recognized that more research is needed on how policy mixes aimed at promoting innovation should look like.This is particularly important considering the need for policy makers to address sustainability transitionswith the decarbonisation of the energy system and the whole economy being a prime example (Markard et al., 2012).For such sustainability transitions, it has been argued that policy mixes are required in order to address not only traditional market failures such as underinvestment in R&D or negative environmental externalities of greenhouse gas emissions, but also structural and transformational system failures, such as institutional failures or failures regarding the direction of a transformation process (Weber and Rohracher, 2012).The importance of policy mixes in addressing this array of failures in order to address societal challenges such as the climate crisis, have also been recognized by the OECD in its report on system innovation (OECD, 2015).Also Schot andSteinmueller (2018, p. 1563) argue that transformative change towards sustainability requires a mix of policies and that "any new policy attempt must navigate pre-existing policies and find ways to create a productive layering of existing and new policies".
Policy mixes aimed at stimulating sustainability transitions are particularly challenging for a number of reasons: they normally go beyond single policy domains (e.g. they cannot be addressed by innovation policy alone but also need changes in other policy fields such as market regulations or tax rules); there is large uncertainty about future developments (e.g.technical, political, cultural); and the desired direction of change and the complexity of the change process is significant as transitions go beyond technologies (e.g. also require changes in infrastructures, social practices, and market arrangements).
In this special issue we follow calls by Flanagan et al (2011) for reconceptualising policy mixes for innovation, including paying attention to the politics and policy processes, and by Rogge and Reichardt (2016) for adopting a broad perspective on policy mixes for sustainability transitions.Flanagan et al (2011, p. 702) advocate a multi-actor, multi-level analytical approach to policy mixes which incorporates a "dynamic understanding of the processes by which policies emerge, interact and have effects".To develop such an approach they draw on insights from evolutionary economics as well as the policy studies literature.Similarly, Rogge and Reichardt (2016) conceptualise policy mixes for sustainability transitions as not only being constituted of interacting instruments and corresponding policy strategies with their long-term targets, but they also explicitly include policy processes which shape the policy mix.We follow this line of argument and broadly define policy processes as "political problem-solving processes among constrained social actors in the search for solutions to societal problems" (Rogge andReichardt, 2016, p. 1625).
Given the increased interest in the processes underlying policy mix developments, we argue that research on policy mixes within the innovation studies field would benefit from more crossfertilization with related work in the policy studies literature (e.g.Howlett et al., 2015;Howlett and Rayner, 2007).In this literature, policy mixes are typically understood as "complex arrangements of multiple goals and means which, in many cases, have developed incrementally over many years" (Kern and Howlett, 2009, p. 395).More specifically, this literature focusses on assessing policy mixes against certain characteristics such as coherence and consistency as proxies for their potential success, which has already been taken up in the context of low-carbon innovation (Costantini et al., 2017;Kern et al., 2017; e.g.see Kern and Howlett, 2009;Rogge and Schleich, 2018).The potential 'fit' of proposed new policy programmes or instruments with their governance context is also important (Howlett and Rayner, 2013).One of the key arguments in this literature is that new policy developments are hoconstrained by previous policy choices.For example, Howlett and Rayner (2007) argue that the implementation of new policy programmes and governance arrangements depends on a number of well-understood processes such as increasing returns and other kinds of positive feedbacks, sunk costs, and incremental policy learning which are all very familiar to innovation scholars.
We suggest that drawing on this line of research can help the innovation studies community to better conceptualize and analyse through which kinds of processes policy mixes develop over time and what the likely outputs and innovation impacts of such processes are.Given the potential value of the conceptual frameworks developed in the policy studies literature for generating a better understanding of policy mixes for sustainability transitions, this special issue brings together contributions from scholars in innovation studies as well as policy studies in order to facilitate cross-fertilisation of ideas and the further advancements of research on policy mixes.Specifically, the special issue has three aims: 1.To promote conceptual novelty in the way innovation scholars understand and study policy mixes by bringing together a collection of authors from innovation studies and policy studies; 2. To increase the visibility of the topic of policy mixes for sustainability transitions within the field of innovation studies by showcasing high quality conceptual and empirical studies of policy mixes; 3. To synthesize relevant policy insights derived from the emerging academic work on policy mixes for sustainability transitions, thereby responding to the recent interest in policy mixes by organisations such as the OECD, the European Commission or the International Energy Agency.
The ambition of the SI is to support the emergence of a new strand of interdisciplinary social science research on policy mixes specifically in the context of sustainability transitions which actively combines approaches, methods and insights from innovation and policy studies.
In order to elicit contributions for this special issue we published an open call for papers through relevant mailing lists and our own networks, targeting both innovation as well as policy studies scholars, in response to which we received 65 extended abstracts.Through independent reviews by the three guest-editors we selected 17 manuscripts.The selected authors were invited to a 2day workshop at which authors received feedback from the other contributors as well as the guest editors on their first full drafts.The revised contributions were then submitted to Research Policy and went through the standard double-blind review process.Out of this process nine papers were accepted and will be introduced in more detail in section 3 of this introduction.
The remainder of this introductory article proceeds as follows: The next section presents a bibliometric analysis of the literature on policy mixes, starting with a general overview and then zooming in on contributions of the fields of innovation studies and policy studies, as well as highlighting some emerging linkages between these two strands.Section 3 introduces the key research themes of the special issue and summarises the contributions made to these themes by the papers in the special issue.Section 4 concludes by offering some critical reflections on the state of the field and deriving lessons learned for policy makers and scholars alike.

Bibliometric analysis: Policy mixes in innovation and policy studies
Over the past two decades, research on policy mixes has experienced increasing interest, with studies on innovation and transitions being an important contributor to this trend since 2014 (see Figure 1).In order to obtain a better overview of this expanding field we conducted a bibliometric analysis of academic articles discussing policy mixes, instrument mixes, policy portfolios and policy packages, thereby recognizing the variety of terminology used in policy mix studies.Through a topic search (TS) we first searched for any paper utilizing these terms in its title, abstract or keywords (see section 2.1), before zooming in into two subsets: policy mix articles published in innovation studies (see section 2.2) and those published in policy studies (see section 2.3).Table 1 provides an overview of the corresponding search terms and resulting overall publication figures up to 2017, with 1993 being the first year with two policy mix publications on record.These search terms are not free from limitations: some studies may be included which are not relevant for the field, while others may be overlooked if they do not fully meet the search criteria.For example, one of our own relevant and highly cited papers (Kern and Howlett, 2009) published in Policy Sciences does not come up in the search because it does not fulfil these relatively narrow search criteria (it contains the policy mix keyword but does not use any of the policy studies related keywords).This is a general limitation of this kind of research and could only be mitigated by extensive manual checks which were beyond the scope of this work.However, we believe that overall these search terms yield a good overview of policy mix studies research.
Source: Own, resulting number of studies based on data from Web of Science (as of December 21, 2018) For our bibliometric analysis we draw on data on academic articles from the Web of Science Core Collection (WoS), from which we downloaded the full record and cited references on December 21, 2018 for the period between 1900 and 2017.This data was then used in the software CitNetExplorer to generate citation trees for obtaining a 'big picture' of all policy mix studies, and for policy mix publications in 'innovation studies' and 'policy studies' more specifically ( van Eck & Waltman, 2014).We also analysed the publication data with the software VosViewer to conduct co-word analysis within the 'innovation' and 'policy studies' groups as well as co-authorship analysis within these two fields ( van Eck & Waltman, 2013).

Overview of policy mix studies
Research investigating policy mixes has seen an increasing trend, with most articles having been published over the last 20 years, namely 833 of the total 894 publications up to 2017.As early as in 2002 research on policy mixes had already seen more than 20 academic articles per year.By 2008 the number of annual publications had climbed to over 40 publications, and by 2017 it had reached a peak of over 150 publications (see Figure 1).Of these, the large majority uses the term 'policy mix' (502, or 56.2% of all articles), followed by 'policy package' (275, or 30.8%), whereas the terms 'policy portfolio' (86, or 9.6%) or 'instrument mix' (51, or 5.7%) are used much less frequently.These are not clear-cut boundaries, however, as several articles use multiple terms.As we are interested in illuminating the development of thinking around the notion of policy mixes in the widest sense we include all of these terms in our further analysis.
Figure 2 shows the citation tree for ~70 of the most cited publications on policy mixes. 2 The pioneering publication on policy mixes was authored by Gunningham and Sinclair (1999) on policy mixes for environmental protection.Another particularly influential article was published by Sorrell and Sijm on carbon trading in the policy mix (2003), and a study by Flanagan et al. (2011) on reconceptualizing policy mixes for innovation.For the purposes of the special issue, we are particularly interested in two subsets of policy mix articles: those within innovation studies and those within policy studies.Therefore, in the following we zoom into these two subsets to obtain more detailed insights into how these two fields have developed.The citation tree for the 200 connected components of the total 894 publications displays ~70 publications which were selected by CitNetExplorer based on citation scores (van Eck and Waltman, 2017).This results in an underrepresentation of more recent publications.For example, in 2017, 155 articles on the theme were published compared to 52 articles in 2013.However, Figure 2 gives the impression that there were significantly more publications in 2013 than in 2017.

3
Note that only the last name of the first author of each publication is displayed.

Policy mixes in innovation studies
We zoomed in on the 'innovation' group of articles to better understand the policy mix literature focussed on innovation, socio-technical transitions and sustainability transitions.As shown in Figure 1 and Table 1 until 2017 the 'innovation' group has so far produced 141 academic articles (the first in 1994), with a particular strong growth over the last four years.
Overall, the UK is particularly well placed in this field, with four universities each having at least published four policy mix articles and most overall articles.Our analysis also reveals that publications on policy mixes within the 'innovation' group have largely originated in Europe, but that the US and Canada have also made significant contributions to the field, though without a concentrated 'organisational home'.The top 3 authors in the 'innovation studies' group are Karoline Rogge (with eight publications on policy mixes), Elivra Uyarra (6) and Florian Kern   , country, author and journal (1994 -2017) Note: The following cut off points apply regarding the minimum number of publications: institutions (at least 4), country (at least 7), author (at least 3), and journal (at least 3).
Source: Own, based on data from Web of Science (as of December 21, 2018) When taking a closer look at the links between the identified 141 publications in the 'innovation' group we find that 79 publications are connected by direct citation.Figure 3 shows the resulting citation tree which highlights the early influence of two EU funded research projects for the development of the field: one project in the field of innovation policy (linked with Nauwelaers and Wintjes, 2002) and one in the field of climate policy (linked with Sorrell and Sijm, 2003).The citation tree also shows that the divide between innovation and climate policy has started to be overcome from 2014/2015 onwards, with increasing connections between these.The visualisation also shows the vastly growing number of well-connected outputs from 2016 onwards.For a list of the most cited papers in this group, which is led by an article by Flanagan et al (2011), see Annex A.  Source: Own, based on data from Web of Science (as of December 21, 2018)

Policy mixes in policy studies
We now zoom in on policy mix research focussing on policy design, policy implementation and policy processes to better understand the policy mix literature within the fields of policy studies, public administration and political science.As shown by the dotted grey line in Figure 1 and the numbers in Table 1 the 'policy studies' group is the smaller sub-set of the two groups of interest here.By the end of 2017 it had produced 72 academic articles, with the first two published in 1997.However, it also needs to be pointed out that a significant share of the studies captured within this 'policy studies' group also belong to the 'innovation' group (in total 23 publications).Also, the rate of growth in the 'policy studies' group has been smaller than in the 'innovation' group and the even faster growing overall policy mix literature.
When looking at the organisations with most academic outputs in the field our analysis reveals that their ranking differs significantly to the 'innovation' group, with the National University of Singapore and the Simon Fraser University in Canada leading the field, with 5 publications each (see Table 3).Overall, a more international picture emerges in the 'policy studies' group, which is led by the UK (with 15 publications and two universities in the top group), closely followed by the USA (with 14 publications).These top countries are followed by Germany   , country, author and journal (1997 -2017) Note: The following cut off points apply regarding the minimum number of publications: organisation (at least 3 publications), country (at least 3), author (at least 2), and journal (at least 2).
Source: Own, based on data from Web of Science (as of December 21, 2018) The dispersed nature of the 'policy studies' group is also illustrated by Figure 5 showing those publications that are connected by direct citation, which is the case for only 23 out of the 72 articles.We find that two publications are of central relevance in the citation tree: Doremus ( 2003) and Flanagan et al. (2011).Indeed, Kieron Flanagan and his co-authors Elvira Uyarra and Manuel Laranja with their explicit emphasis on policy processes in the context of policy mixes is the most cited paper in this group (see Annex B), just as was already the case in the 'innovation' group.Overall, however, the 'policy studies' group is not only smaller, but also much less connected than the 'innovation' group, suggesting that there may only be a weak community of practice involved in policy mix studies within the 'policy studies' group.Yet, in the past few years the connections between policy mix publications in this group have increased.Part of this development is driven by articles on innovation, as highlighted by the overlap in publications appearing in both groups (23).Our analysis of key words further illustrates the dispersed nature of the 'policy studies' group.
Again, we performed a co-word analysis with VosViewer, using the key words provided by the authors of the 72 publications in this group.unites studies on governance, coordination, implementation and politics more generally.
However, the largest cluster is the red one (on the right) which revolves around discussing policy design in a policy mix setting, with empirical studies largely focusing on environmental and climate policy, conservation, ecosystem services and energy efficiency, typically approached from a systems perspective.

Emerging links between innovation and policy studies groups
One of the key aims of this special issue is the bridging of conceptual, methodological and empirical developments across innovation and policy studies in order to be better able to investigate policy mixes for sustainability transitions.To get a sense of the starting point for such an endeavour we merged the 141 policy mix publications from the 'innovation' group with the 72 publications from the 'policy studies' group.This yields 190 publications in the combined 'innovation and policy studies' group (due to an overlap of 23 articles).
For these publications, Table 4 shows that the top 2 institutions in terms of publishing activities are the University of Sussex in the UK, followed by the Fraunhofer Society in Germany. 4  Overall, the UK is clearly dominating the ranking, with four universities in the top 9 list of organisations (with a total of 34 publications), followed by Germany, with two research associations in the top 9 list (with a total of 17 publications).Source: Own, based on data from Web of Science (as of December 21, 2018) Overall, the co-authorship analysis shows that there are authors who have started to collaborate across innovation and policy studiesprimarily addressing the topic of sustainability transitions and thus have started to build the bridges we want to strengthen with this special issue.
However, 38 out of 464 authors is a rather small set of people which means that there is much potential for more and deeper connections.

Overview of the contributions to the special issue
Building on the bibliometric analysis in section 2 as well as our reading of the relevant literature, we identified five important themes within policy mix research for sustainability transitions which would particularly benefit from a joining of insights from innovation and policy studies: methodological advances, policy making and implementation, actors and agency, evaluating policy mixes and co-evolution of policy mixes and socio-technical systems.These five themes are introduced in the subsections below, followed by a summary of the contributions that the papers in this special issue make to them.

Methodological advances in policy mix research
Analysing policy mixes is a challenging endeavour.While it may seem relatively easy to develop abstract conceptual frameworks and typologies of policy mixes and their characteristics, it is difficult to develop and systematically apply appropriate methodologies for analysing them.One important methodological challenge concerns the definition of the boundary of a policy mix.Studies of individual policy instruments find it straightforward to define the object of analysis, but for policy mixes this endeavour is much more complex.The definition of the policy mix itself of course provides some initial boundaries.For example, some policy mix concepts only include policy instruments (or Borrás and Edquist, 2013;such as Veugelers, 2012), others include policy goals and policy instruments (Kern and Howlett, 2009) while Rogge and Reichardt (2016) argue for an extended concept of policy mixes, which includes policy elements (policy strategies and instrument mixes), characteristics of policy mixes (such as consistency, coherence, credibility and comprehensiveness) as well as policy making and implementation processes.However, even in the narrowest definition of the policy mix concept, which only includes interacting policy instruments, it is by no means straightforward to decide which instruments are included in the mix (and which are not), with the challenge becoming more pronounced with broader policy mix conceptualisations.
A second core methodological challenge is how to characterise policy mixes given that they often have a significant number of elements to be considered.Much of the existing work has therefore focused on characterising policy mixes by categorising individual policy instruments e.g.into technology push or demand pull policies (Costantini et al., 2017), into carrots, sticks and sermons (Mavrot et al, this issue) or into contributing to the creation of new or the destruction of old regimes (Kivimaa and Kern, 2016).Mapping and counting the number of instruments in each category is often used as a proxy for assessing the balance or comprehensiveness of the mix.However, given that such categorisations are necessarily quite crude and much research shows that the specific design of individual instruments is important for their effectiveness (e.g.Kemp and Pontoglio, 2011) as well as how they interact with other instruments in the mix (del Río and Cerdá, 2017), this is a rather unsatisfactory approach.
The challenge of delineating policy mixes is addressed by Jan Ossenbrink, Sveinbjoern Finnsson, Catharina R. Bening, and Volker H. Hoffmann (this issue) who argue that in order to build a consistent research programme on policy mixes, it is important that there are accepted and shared heuristics for delineating policy mixes.The literature review conducted by Ossenbrink et al shows that there are principally two ways in which relevant elements of a policy mix can be identified, top-down or bottom-up, but that many studies seem not to consciously chose between the two options and provide little detail on how the mix is identified.
The authors clarify that the top-down approach builds on the idea that the elements of a policy mix originate from an overarching strategic policy intent, which is thus taken as the starting point and then all relevant instruments used to implement the strategy and its objectives are identified.By contrast, the authors highlight that the bottom-up approach starts from the definition of a focal impact domain that is of core interest for the respective research question (e.g. a specific technology or sector) and then searches all policy instruments which affect decision making by actors in this domain.Importantly Ossenbrink et al develop a systematic analytical procedure on how to go about either of the two approaches and demonstrate how these methodological procedures can be applied in practice by conducting an exemplary analysis of the policy mix in the energy storage domain as part of California's energy transition.
Their results show, maybe surprisingly, that there is very little overlap between the policy mixes identified through the two different approaches.This supports their argument that transparency and consistency is needed in order to increase the potential for replicating studies.Their study also demonstrates that both approaches have specific advantages and disadvantages and the authors suggest that they can be combined to complement each other, even though they acknowledge this is challenging in terms of resource implications.Overall, this paper establishes a new standard for delineating policy mixes transparently and systematically, which will hopefully provide useful guidance for future studies and make it more difficult to fall behind such standards.
The methodological challenge of how to characterise a policy mix in more sophisticated ways is addressed by Tobias Schmidt and Sebastian Sewerin (this issue) who conduct an analysis of renewable energy policy mixes in nine countries.The authors develop a conceptualization and measurement of policy mix balance across instrument types as well as two policy mix design features (intensity and technology specificity).While assessing the balance across instrument types is a common approach to characterise policy mixes (Costantini et al., 2017; OECD, 2016), Schmidt and Sewerin propose a novel measurement in terms of dispersion of instruments through the Gini-Simpson Index.Also, by assessing the policy mix design features the paper provides a richer analysis than simply counting the number of instruments.This is accomplished by first assessing the intensity of each individual instrument, which is then aggregated into an overall intensity of the mix.Building on Schaffrin et al. ( 2015), Schmidt and Sewerin conceptualise the intensity of an instrument as a combined score against a number of criteria, including its objectives, scope, integration, budget, implementation, and monitoring.
They operationalise it through a systematic qualitative coding procedure.In addition, the authors provide an analysis of the temporal dynamics of policy mixes which is a prominent theme in the policy sciences literature on policy mixes (e.g.Howlett and Rayner, 2007).In contrast, many studies in the innovation studies field often provide analyses of policy mixes at one point in time, which especially in the context of long-term sustainability transitions is an important weakness.The temporal analysis is used to answer the question how temporal dynamics of policy mixes differ between countries regarding their balance and design features.
Schmidt and Sewerin find that countries' policy mix dynamics vary strongly on some dimensions (e.g.intensity) but not so much on others (e.g.balance).In sum, this paper enriches ongoing discussions about how to measure characteristics of policy mixes (Costantini et al., 2017;Rogge and Dütschke, 2018;Rogge and Schleich, 2018) and hopefully will inspire others to further advance the systematic measurement of policy mix characteristics.

Processes of policy mix formulation and implementation
One of the features of much policy mix research is the focus on the content of the policy mix, while much less attention is paid to the policy making and especially implementation processes.
This is problematic if one is interested in the outcomes of policy since the policy studies literature suggests that the impacts of policies do not only depend on their design, but also on how they are being implemented (see e.g.McLaughlin, 1987 for a summary of different generations of implementation studies).Similarly, Borrás and Edquist (2013) pointed to the importance of policy implementation in contributing to differences in policy outcomes.In addition, Flanagan et al (2011) argued that institutional contexts within which instruments operate are crucial in determining their effects.Moreover, the policy making style and the coherence of policy processes were identified as influencing factors for low-carbon innovation (Reichardt et al., 2017(Reichardt et al., , 2016)).Therefore, Rogge and Reichardt (2016) argued that it is important to also consider policy making as well as policy implementation processes in the context of analysing policy mixes for sustainability transitions for which scholars can draw on various theories of the policy process (Kern and Rogge, 2018).Two of the contributions in this special issue engage with these challenges.
The paper by Gijs Dierks (this issue) provides a rich case study of an important international organisation, the Organisation for Economic Cooperation and Development (OECD), and examines how the OECD is shaping policy mixes for sustainability transitions.The role of the OECD is particularly interesting as it has in the past significantly shaped innovation policy in many of its member states (e.g. through the Frascati and Oslo manuals) and also because it has recently adopted the concept of system innovation (i.e.transitions, see OECD, 2015) and has taken up thinking about innovation policy in terms of policy mixes (e.g.OECD, 2016).Dierks argues that system innovation thinking has potentially radical implications for innovation policy mixes and therefore analyses how it was taken up within the science, technology and innovation directorate of the OECD.He also provides an account of why this way of thinking has not been institutionalised within the OECD's core activities.The analysis shows how legacy effects in the form of institutional, cognitive and political 'sticking points' prevented such a translation and inscription into proposals for policy mixes for sustainability transitions.While there was some momentum behind the uptake of the concept of system innovation within the OECD in 2014 and 2015, driven by its technology and innovation policy working group consisting of OECD staff and policy makers from member countries, Dierks finds that several purposeful efforts to inscribe the concept into the organisation's core activities have largely failed.The project ended in early 2017 and without further initiatives it is doubtful whether the system innovation concept will be able to shape innovation policy advice by the OECD in terms of policy mixes for sustainability transitions.The paper by Dierks presents a powerful reminder of how difficult it is to overcome path dependencies and develop novel ideas about suitable policy mixes to promote sustainability transitions, even in organisations which have acknowledged the importance of the challenge and also cognitively started to think in terms of policy mixes.
Building on the idea that implementation processes are key for policy outcomes and therefore for sustainability transitions to occur, Céline Mavrot, Susanne Hadorn and Fritz Sager (this issue) make a case for looking at the implementation context of policy mixes.In their conceptualisation this context includes the specific setting within which a policy instrument is implemented (such as firms, schools or families as different kinds of social settings) as we all the specific target group of each instrument (who are nested within particular social settings).
The authors argue that focussing on the specific settings where interventions are implemented instead of whole policy domains allows for a more precise understanding of policy making and policy implementation processes.Including target groups into the analysis of policy mixes is argued to be crucial since the recipient side of policy may be as important as the sender side of policy (strategies, goals, instruments) in explaining outcomes (Rogge and Reichardt, 2013).
Building on Rogge and Reichardt (2016) the aim of the paper is to propose a conceptual framework for analysing the effectiveness of policy mixes and specifically their ability to induce behavioural change.Empirically, Mavrot et al analyse an unusual case from an STI perspective: the implementation of tobacco control policies in Switzerland.They convincingly argue it to be an exemplary case for analysing transitions which involve significant behavioural change driven by public policy, as well as an emblematic case of an attempted managed decline of a specific product and its associated industry.The analysis triangulates different sets of quantitative and qualitative indicators in order to assess the implementation of eleven subnational policy mixes.Mavrot et al find that taking into account implementation contexts of the policy mixes improves our understanding of policy compliance at the individual level, which is crucial in policy-driven transition processes.For example, they highlight that target groups can only be addressed successfully if the many actors involved in policy implementation cooperate effectively.The paper thus demonstrates how insights generated by the policy implementation and evaluation literature as well as health policy insights can be used to finetune our conceptualisation of policy mixes for sustainability transitions by paying more attention to the recipient side of policy mixes.

Actors & agency
One of the criticisms of policy mix research is that analysis often remains on a fairly abstract level, for example regarding policy layering or the co-evolution of technology and policy.
However, to really understand the nature and dynamics of such processes it is important to look at the agency of actors driving them.Within the literature on sustainability transitions there is some research specifically focussing on actors and agency (Farla et al., 2012; for a review, see Fischer and Newig, 2016;Wittmayer et al., 2017).Recently, Duygan et al (2019) developed a heuristic to analyse the determinants of agency in transition processes.There are also studies focussing on the role of specific actor types such as incumbents (Smink et al., 2013), intermediaries (Kivimaa, 2014), community-based initiatives (Seyfang and Haxeltine, 2012) or users (Schot et al., 2016) in transitions.Especially the literature on the politics of transitions (Avelino et al., 2016; for a recent summary, see Köhler et al., 2019) or the policy studies literature (Howlett et al., 2009;Sabatier and Weible, 2014) provide many useful starting points for thinking about different kinds of actors and their involvement in policy processes which can be utilised in the analysis of policy mixes for sustainability transitions (Kern and Rogge, 2018).
Two of the papers in this special issue focus specifically on the actors involved in developing policy mixes.
The paper by Marie Byskov Lindberg, Jochen Markard and Allan Dahl Andersen (this issue) focusses on the EU's energy policy mix, for which it provides an assessment of key industry actors' policy preferences.This is argued to be important since transition pathways unfold as a result of continuous struggles of actors over policy goals and instruments.Given the interest in the direction of transitions, and specifically how actors and policies influence the direction, the paper asks which policies and which actors favour which kind of pathway.
Lindberg et al also propose a methodological approach to analyse major conflict lines in an ongoing transition by focussing on policy preferences.One of the contributions of the paper is therefore to demonstrate how actors and policy preferences can be explicitly included in the study of policy mixes.In terms of potential transition pathways, the analysis distinguishes between two dimensions: the degree of sustainability and the degree of disruption.Lindberg et al first map policies according to these two dimensions (which pathway do they support?) and then similarly map key industry actor preferences according to the same dimensions.They develop a systematic approach for analysing and comparing policies and actor preferences through establishing a set of criteria for both dimensions and developing a coding system.The numerical coding along the two dimensions enables a graphical representation of the analysis.
The analysis shows that there generally is a large overlap between current policies and actor preferences in the direction of a centralised renewable energy system, but also finds that there are many actors with strong preferences in favour of a more decentralised and more ambitious renewables pathway, which is also supported by some policies.Overall, the study demonstrates an interesting way to integrate the analysis of actor preferences into the analysis of policy mixes.
Future work could build on this approach and try to combine such mapping with a more processual account looking at the politics of the underlying policy processes and how actor preferences are shaping the content of policies.

The contribution by Karin Ingold, Isabelle Stadelmann-Steffen and Lorenz Kammermann
(this issue) focuses on citizens, which they argue to be an understudied target group of energy transition policies.Their research question is: "What are citizens' preferences towards new instruments promoting renewables in an instrument mix situation, and what factors influence these instrument mix preferences?".Conceptually, Ingold et al draw on the social acceptance literature as well as on sustainability transitions research.Their framework differentiates between three different potential drivers for citizens' instrument preferences in an instrument mix setting (individual values, potential to be energy prosumers, existing policy mix) which are used to develop a number of hypotheses.The current instrument mix is assessed by looking for all policy instruments supporting three main renewable energy technologies, classifying them into four categories (tax incentives, subsidies, ban and information), and characterising the mix in terms of its density (number of instruments present) and intensity (e.g. the amount of resources used, strictness of standards, etc.).Empirically, the focus is on the Swiss energy transition.The authors draw on a survey among Swiss residents, with data being analysed by estimating logistic multi-response models, which disentangle the different potential drivers for citizen preferences.The results show that it is mainly individual factors that contribute to the acceptance of additional policy instruments compared to context-related factors, and that respondents acknowledge that a mix of instruments will be required to meet the goals.
Importantly, the analysis shows that there is no systematic relationship between the existing cantonal instrument mix and individual's instrument mix preferences.The hypothesis that instruments introduced in the past affect current instrument preferences was therefore not supported, though the authors argue that this might be a consequence of the way the policy mix was operationalised in the survey.Regardless, the methodology developed in this paper could potentially be applied to analysing policy maker or industry actor preferences', as the literature on policy path dependence and policy feedbacks suggests that such associations are plausible.

Evaluating policy mixes
Policy evaluation is complex and challenging for a number of reasons, even when focusing on single instruments rather than policy mixes (Vedung, 2017).One main challenge is the problem of attribution, i.e. how one can link a specific policy intervention with a particular outcome which is claimed to have followed from the existence of the policy.In the domain of science, technology and innovation policy, theory-based evaluation has been proposed to draw on theories of the systemic nature of innovation to assess innovation policy (Arnold, 2004;Molas-Gallart and Davies, 2006).However, so far only a handful of European countries (Austria, Denmark, Finland, Ireland and the Netherlands) comprehensively consider policy mix questions in their evaluation of system-oriented innovation policy (Borrás and Laatsit, 2019).
In the context of mission-oriented innovation policy and transformative policy mixes for sociotechnical system change the challenge of policy (mix) evaluation becomes even more pronounced (Arnold et al., 2018;Fisher et al., 2018;Kivimaa et al., 2017).On the one hand, sustainability transitions and experiments are complex, uncertain and co-evolving processes calling for tailored evaluation approaches (Luederitz et al., 2017;Nieminen and Hyytinen, 2015); on the other hand policy mixes can be characterised by very different rationales, thereby complicating the question about suitable evaluation criteria, and by policies originating from different policy levels which continually interact with one another in complex ways (Magro and Wilson, 2013).As a consequence, paying attention to policy mixes for sustainability transitions challenges traditional policy evaluation approaches and has implications for the governance of evaluation practices (Boni et al., 2019), as one paper in this special issue shows.
The contribution by Edurne Magro and James Wilson (this issue) analyses the interaction between governance processes and policy mix evaluation in the context of regional smart specialisation strategies.The challenge of evaluating such policy mixes, many of which aim to contribute to sustainability transitions, is to better understand the impacts of such policy mixes and how they contribute to the processes and direction of transformation.Smart specialisation strategies are meant to set the direction of travel through an experimental 'entrepreneurial discovery process' involving a wide range of regional stakeholders from government, business, research and civil society.The aim of the paper is to analyse how the strategic intelligence and learning gained through appropriately governed evaluation processes can be used to boost regional capacity building.The paper's assumption is that the usefulness of the evaluation process for learning very much depends on whether the governance of the evaluation process is seen as legitimate by the actors involved in setting the direction of the smart specialisation strategy.Conceptually, the paper draws on the policy evaluation, governance and smart specialisation literatures.The authors focus on three dimensions of smart specialisation policy mixes: their multi-level nature (verticality), their directionality, and experimenting with new governance arrangements to involve stakeholders involved in the prioritisation process.
Empirically, the paper explores these issues through a case study of the Basque Country in Spain.Magro and Wilson find that in the context of advanced manufacturing (one of the priority areas under the Basque smart specialisation strategy) there are strong elements of verticality since there are five different administrative levels involved.In terms of the directionality of the policy mix they find the co-existence of neutral instruments with no directionality, instruments that support specific directions, and instruments which have been made directional to serve the strategy through a policy patching process (e.g.amending eligibility criteria for existing subsidy schemes).The Basque government is shown to have utilised a two-pronged governance approach based on experimentation: a 'hard governance' arrangement in terms of the regional government's oversight of the implementation and evolution of the strategy (and policy mix); and a 'soft governance' arrangement in terms of the wider engagement with business, research and civil society in an 'entrepreneurial discovery process'.Overall, this paper provides insights into the complexity of real-world policy mix evaluation challenges and what kinds of governance arrangements might be conducive to produce the policy learning required to develop successful smart specialisation policy mixes.
3.5 The co-evolution of policy mixes and socio-technical systems The policy sciences literature on policy mixes acknowledges the path-dependent nature of policy making and is interested in the processes through which policy mixes evolve over time.
Concepts like policy patching versus packaging (Howlett and Rayner, 2013) or policy layering (Howlett and Rayner, 2007;Kern and Howlett, 2009) have been used to study such policy mix dynamics.In the field of sustainability transitions, there have been several contributions which have argued for a need to better understand these processes of policy development in the wider context of dynamics within the socio-technical systems which the policies are meant to address (for example Hoppmann et al., 2014;Reichardt et al., 2016).The argument is that in the context of transitions, the policy mix and the socio-technical system co-evolve over time and that transition scholars need to better understand these dynamics in order to provide better policy advice.This is reflected in recent calls for a greater emphasis of transitions research on policy feedbacks (Roberts et al., 2018;Schmidt and Sewerin, 2017).
The paper by Duncan Edmondson, Florian Kern and Karoline Rogge (this issue) develops a novel conceptual framework which aims to shed light on the co-evolution of policy mixes and socio-technical systems over time by focusing on policy feedbacks.The framework builds on both the sustainability transitions literature as well as the policy feedback literature within policy sciences to better conceptualize the specific mechanisms through which this co-evolution happens.The policy feedback literature is interested in the consequences of policy making on subsequent rounds of policymaking and its politics, with the core idea being that policy is path dependent.The framework draws on this research strand and proposes three types of effects through which policy can shape developments of the socio-technical system (resource, institutional and interpretive effects) and three feedback mechanisms through which developments within the socio-technical system influence further policy developments (sociopolitical, administrative and fiscal feedbacks).The framework draws on recent developments in the feedback literature in which both possibilities of positive (self-reinforcing) feedbacks as well as negative (self-undermining) feedbacks are considered.The utility of this framework is illustrated through an empirical application to the case of the zero carbon homes policy mix in the UK.This ambitious policy mix initially led to some momentum towards lower carbon housing but because of a number of developments, the original policy ambitions were undermined, target groups lost their faith in the seriousness of the commitment and lobbied against it, so that ultimately the zero carbon homes ambitions were abandoned by the government.The case study therefore illustrates the case of a failed transition and how despite initial strong policy commitment and some positive feedbacks occurring, it can be difficult for actors to maintain commitment to an ambitious policy mix in the face of difficult external circumstances (financial crisis, housing crisis).The paper contributes to understanding the maintenance of policy mixes as a challenging political process and proposes that policy makers seeking to support sustainability transitions need to attempt to design policy mixes capable of generating positive feedback, thereby strengthening political support over time.
The paper by Ping Huang (this issue) explores the co-evolution of a vertical policy mix across several levels of policy making (urban, regional and national) with industrial path creation.By focusing on the interactions across policy levels Huang addresses a gap in the empirical policy mix literature in which most analyses focus on the policy mix interactions and dynamics on one level of policy making (typically the national one).He also stresses that such an analysis must take a co-evolutionary perspective since policy mixes co-evolve with socio-technical systems.
The aim of the paper is therefore to better conceptualise interactions between multi-level policy mixes and their co-evolution with industry development.Conceptually, the paper draws on insights from policy mix research and the concept of anchoring from economic geography which is understood as the 'mobilization and recontextualization of both regional and extraregional resources in a particular technological field within the region'.Empirically, the analysis traces the industrial path creation of solar water heating (SWH) technology in Shandong Province in China and its interactions with policy mix developments at several levels of policy making.It finds that there are both bottom-up and top-down patterns of policy mix development across levels of policy making and that interactions between the levels of policy making evolved over time from unidirectional (top-down) to bidirectional (top-down and bottom-up) patterns in line with industry maturation.This is the case because of vertical policy learning dynamics which are argued to have led to an increase in the comprehensiveness of the policy mix.The paper serves as a reminder for how important local policy developments can be in the context of multi-level policy mixes, but also that the temporal dimension is important as policy mixes need to change over time in line with the phase of the transition (which some in the literature have called 'policy sequencing', see Meckling et al., 2017).

Synopsis
In this section, we have introduced five strands within policy mix research and positioned the contributions of the papers in the special issue within them.Table 4 offers a comprehensive summary of these nine contributions.As can be seen, six of these papers build on the extended policy mix concept developed by Rogge and Reichardt (2016), while the other three utilize the policy mix definition offered by Kern and Howlett (2009).Six papers make conceptual contributions to the emerging policy mix literature, while four papers also offer empirical contributions and three of the papers make methodological contributions.Empirically, and despite an open call for papers, the large majority of contributions focuses on energy transitions.
In addition, the geographic scope of the papers in this special issue is largely limited to Europe, Northern America and China.Methodologically, the special issue includes not only qualitative but also quantitative as well as mixed methods contributions, with many offering unique and novel approaches for the emerging literature on policy mixes, and others proposing new standards in the field.

Conclusion
Research on policy mixes has significantly increased over the last few years and the topic has also received increasing attention from policy makers interested in directing and accelerating socio-technical systems towards sustainability.This article has taken stock of the literature on policy mixes for innovation and sustainability transitions and analysed emerging collaborations between scholars from innovation and policy studies.It also summarised the contributions to the special issue on policy mixes for sustainability transitions and positioned them within five major research strands.
All papers included in this special issue draw on some policy sciences theories, concepts or methodological approaches and in our view this has led to a major enrichment of policy mix research within the field of science, technology and innovation (STI), both in terms of conceptual frameworks as well as in terms of methodologies.In addition, all contributions in the special issue demonstrate the utility of their approaches in delivering insightful empirical analyses despite the inherent complexity of policy mixes.We hope this will stimulate further work within the STI community drawing on insights from the policy studies field.
Furthermore, the contributions in the special issue use a variety of qualitative and quantitative approaches to shed light on different dimensions of and research questions about policy mixes for sustainability transitions.This variety is encouraging and we hope that many scholars will build on these approaches in future analyses of policy mixes.We feel that the methodological standards set by the contributions are high and that over time more analyses along similar lines will lead to a productive and systematic research programme on policy mixes for sustainability transitions that can provide valuable insights for policymaking.
However, we also critically note that several of the analyses in the special issue do not provide very specific policy advice.Potentially, this is partly a feature of being a relatively new field of research, but more importantly this may also be related to the theories drawn upon and the methodologies utilised.In fact, most approaches in the field of policy sciences are about understanding and explaining policy processes and their outputs, not about evaluating their outcomes.In contrast, econometric studies used to evaluate policy outcomes have so far largely focused on single policy instruments, with only a few attempts on evaluating the complex policy mixes studied in this special issue.We argue that much could be gained from combining both strands of research for an integrated understanding of policy and innovation processes, their outputs and outcomes, and the co-evolution of policy mixes and socio-technical systems.
We are convinced that such thinking in terms of broader policy mixes is relevant for policy makers, and in fact reflects much more their everyday experience compared to textbook assertions about "best" policy options.As such, policy mix research enables policy makers to advance their thinking about policy complexity, and provides a terminology and analytical tools to make sense of this complexity.For example, the contribution by Ossenbrink et al (this issue) enables policy makers to develop a bottom-up perspective from the angle of 'policy target groups' when their thinking may often be dominated by a top-down approach.The research of Ping and Edmondson et al enables policy makers to adopt a more dynamic view of policy and its coevolution with important dynamics within the socio-technical systems.Yet, it also reminds them that this is not straightforward given path dependencies and power asymmetries in both, including within policy making organisations (also see Dierks, this issue).It also reminds policy makers of the challenge of the (non)acceptance of policies by target groups and how to address this (see Ingold et al. this issue).Finally, it allows policy makers to benchmark their policy approach with other countries and encourages them to (re-)assess the potential of 'layering' through policy patching as a realistic design strategy for more effective policy mixes (see Schmidt and Sewerin, this issue).
In summary, we argue that the policy mix literature is mainly of value to policy makers in terms of providing a heuristic or meta perspective on policy making as an ongoing, complex learning process.Clearly, there is still much to be learned from actual policy experiences about the way horizontal and vertical policy mix dynamics unfold and shape outcomes, how the co-evolution of the socio-technical system and policy influence policy outcomes, and how therefore policy mixes need to be adjusted over time to be able to foster socio-technical transitions.Making progress along these lines in terms of better understanding but more importantly through gaining practical experience of managing such complex processes needs to be a joint endeavour of scholars and policy makers interested in promoting sustainability transitions.We hope this special issue is providing a useful starting point for this journey.

Figure 1 :
Figure 1: Policy mix publication trend

( 5 )
. Finally, there are three top publication outlets in the 'innovation' group, which together account for 38 of the 141 journal articles (i.e.27%): Energy Research and Social Science, Research Policy and Energy Policy.

(
with two non-University research organisations in the top group, together contributing 8 of the 12 German publications), followed by Canada (with 11 publications, with all but one from the top 2 publishing universities) and the Netherlands (with 10 publications).The top authors in the 'policy studies' group are Michael Howlett, with five publications on policy mixes, Pablo del Rio (3) and Jan Börner (3).The top publication outlet in the 'policy studies' group is the journal Energy Policy (having published 6 of the 72 articles in the field), followed by Energy Research and Social Science (with 5 articles).It is noteworthy that there is only one dedicated policy studies journal in the top journal group, namely Policy Sciences (with 4 articles).Overall, the field is rather dispersed with publications spread out amongst 56 different journals, with 50 of these only having published one policy mix article.

Figure 5 :
Figure5: Citation tree of the 'policy studies' group within the policy mix field(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017) Figure 6 displays colour-coded clusters based on how frequently key words are used in the same publication.Again, three different clusters emerge, with a central position of policy design and instruments.While the green cluster (top left) mainly captures studies addressing the topic of policy instruments and innovation in the fields of climate, environmental and renewable energy policy, the blue cluster (bottom left)

Figure 6 :
Figure6: Co-word cluster analysis of 'policy studies' group using author key words(1997- only 38 were found to have collaborated with another author in the combined group.The coauthorship between these connected authors is shown in Figure7.The visualisation shows five clusters of closely related publications (vanEck and Waltman, 2013, p. 5).The size of each node represents the number of documents the author is associated with and the thickness of the line connecting authors represents the frequency of collaboration.

Figure 7 :
Figure 7: Collaboration between policy mix authors in innovation and policy studies(1994- 2017)

Table 1 :
Search terms for bibliometric policy mix analysis and resulting number of studies1

Table 2 :
Ranking in 'innovation' group by number of policy mix publications (#) per organisation

Table 3 :
Ranking in 'policy studies' group by number of policy mix publications (#) per organisation The central role of researchers in the UK and Germany in policy mix research in the combined 'innovation and policy studies' groups is confirmed by the country rankings, with the UK representing 24.2% of all articles in this group (46) followed by Germany with its share of 18.4% (35 publications).The top 4 authors in the combined 'innovation and policy studies' group are Karoline Rogge(8 publications), Elvira Uyarra (6), Michael Howlett (5) and FlorianKern (5).Finally, most research has been published in the journals Energy Research & Social Science (16 articles), Energy Policy(13)and Research Policy(12).

Table 4 :
frequency of co-authorship.From the 190 publications in the 'innovation and policy studies' group a total of 464 authors were identified (411 of them with at least one citation), of which Ranking for combined 'innovation' and 'policy studies' groups by number of  publications per organisation, country, author and journal (1994-2017)Note: The following cut off points apply regarding the minimum number of publications: organisation (at least 5 publications), country (at least 10), author (at least 4), and journal (at least 4). 4 All 14 publications of the University of Sussex originate from the Science Policy Research Unit -SPRU, while all 11 publications of the Fraunhofer Society originate from the Fraunhofer Institute of Systems and Innovation Research ISI.

Table 5 :
Overview of contributions to the special issue