University of Birmingham Optimizing entrepreneurial development processes for smart specialization in the European Union

This paper demonstrates how the Regional Entrepreneurship and Development Index (REDI) can be used to optimize local entrepreneurial discovery processes, in a manner which can support smart specialization strategies (S3). While S3 industry prioritization is based on the identification of local strengths, regional improvement can be achieved by improving the weakest features of the local entrepreneurial ecosystem. REDI based suggestions are place-based and offer rationale for tailor-made regional policy interventions. We found that without optimizing the entrepreneurial ecosystem, the industry specialization alone may not be successful because of the inability of the ecosystem to nurture high growth ventures.

significant number of cases S3 is only being partially implemented, with aspects of the strategy development process themselves potentially undermining the S3 intervention logic.
These various findings all raise the important questions about the ability of S3 practices to foster enhanced local entrepreneurial discovery processes and the extent to which potential inhibitors or blockages to successful S3 implementation are already inherent in the early stages of the policy process itself. In order to better overcome any such inherent inhibitors so as to ensure that the fostering of entrepreneurial discovery processes is genuinely at the heart of the design of the S3 strategy, there needs to be a way of identifying up front the key inhibitors and blockages to the local entrepreneurial search processes.
The entrepreneurship ecosystem (EE) approach provides a way for understanding how we can enhance regional development (Mason & Brown, 2014;Stam, 2015) and the EE approach shares many common features with the S3 approach. The systemic view, the centrality of the entrepreneurial discovery process (EDP), the importance of bottom-up policy actions relying on local actors, and the important shaping and constraining roles played by regionspecific forces, are all key features of both concepts. Similar to S3, EE also aims to foster the emergence of innovative, high growth firms, but the emphasis of EE is different in that it focuses on a wider set of individual and institutional factors, not only on the innovative features of the local system. At the same time, while S3 underlines the importance of specialization and diversity, embeddedness and cluster development, EE takes a broader view by looking at all the interrelated agents, actors, organizations, and institutions (Acs, Åstebro, Audretsch, & Robinson, 2016), including issues such as entrepreneurial attitudes, abilities and aspirations.
In terms of S3 policy, the research question we address is how can we consistently identify and measure the strengths and weaknesses of the regional EDP, building on insights from the EE literature, in a manner which lends itself to helping S3 policy prioritization processes? At the moment, what has been missing in S3 for both analytical and policy-making considerations is a consistent framework for empirically assessing the scale of the strengths and weaknesses of the various EE elements (and their interconnections) and the likely role that they play in fostering or inhibiting EDPs. Ideally, such a framework needs to be able to capture all of these different elements in a manner which allows us to understand both local specificities and broader national and regional comparisons. The ability to more accurately identify the strengths and weaknesses of the local EDP should help to ensure that ex ante S3 priority-setting is more firmly-grounded on objective criteria and less determined by local institutional pressures, and this in turn should also allow for more realistic ex post evaluation processes.
The purpose of this paper is to identify the inhibitors to local entrepreneurial discovery processes, in a manner which can support S3 policy prioritization processes. We argue that the Regional Entrepreneurship and Development Index (REDI) are ideally suited to play such a role. REDI can produce consistent quantifiable measures which capture both the strengths and weaknesses of the individual features of the EE as well as the local EE as a whole. This is important for S3 because while S3 is a combination of top-down and bottom-up processes, the vertical S3 policy prioritization based on the fostering of EDPs is needed because horizontal policies alone have not been able to help move many regions into the knowledge economy (Foray, 2015), and nor have many largely top-down largely sectoral approaches. Having a better systemic understanding of the strengths and weakness of the local EE and a clear sense of the magnitude of the individual EE elements should help S3 prioritization, because many of these features cannot be identified simply by deliberation among stakeholders, nor by looking at comparison cases, nor by considering individual datasets. As such, the largely horizontal perspective afforded by the EE-REDI framework can enhance the primarily vertical parts of S3 processes.
The rest of the paper is structured as follows. The next section provides the conceptual backgrounds about the intersection of S3 and EE concepts. Section 3 explains the structure and the calculation methodology of the REDI.
Section 4 discusses on how the REDI could contribute to S3 policy implementation by providing a solution to four S3 policy caveats as (i) measuring the necessary basic conditions for smart specialization in a mix of 125 NUTS 1 and NUTS2 European Union regions; (ii) identifying the institutional and individual weaknesses at local levels; (iii) providing a comprehensive view about the harmonization of the components of EE; and (iv) presenting some simulations on how additional policy efforts could be optimized to alleviate bottlenecks of the regional ecosystem.
Finally, the paper concludes with some policy suggestions and discussions about limitations and future research domains. We found that industry prioritization is necessary but not sufficient condition to sustain high growth firms.

| ENTREPRENEURIAL ECOSYSTEM BASED POLICIES AND SMART SPECIALIZATION STRATEGIES
In this section, we provide a conceptual background to the EDP, which is the central concept of our research. EDP is mostly a spontaneous, practice-oriented procedure including opportunity recognition and exploitation, trials and errors and learning-by-doing techniques. However, it is also something that can be partially shaped and reshaped affected by the contextual and institutional factors influencing the individuals. These institutional factors interact with the participating agents and can affect their incentives, their trust relations and their interactions with other agents (Rodríguez-Pose, 2017).

| Smart specialization strategies and the entrepreneurial discovery process
Having its roots in the regional innovation system, smart specialization builds on the coupling of different theoretical findings (Foray, David, & Hall, 2011;Hassink & Gong, 2019). With the common goal of creating a theoretical framework for explaining (regional) economic growth potential, S3 has been built with contributions from different research fields inter alia regional science, economic geography, innovation theory, and entrepreneurship (McCann & Ortega-Argilés, 2013OECD, 2013). On the policy side, S3 "came as a reaction to the failures of old-style dirigismes and from the frustration with hands-off government policies" (Kyriakou, 2017, p. 5). The strength of S3 is that it works as a practical melting-pot of theories complemented with the conclusions drawn from the experiences of earlier policy concepts and implementation strategies of the EU and the OECD (del Hermosa, Elorduy, & Eguía, 2015;OECD, 2013).
In the past few years, S3 has gained widespread acceptance in academic and policy arenas the European Union and beyond (Kyriakou, 2017). Politically, it has become a fully-institutionalized strategy framework which serves as an ex ante conditionality in the current Structural Funds programming period (European Commission, 2014c). S3 2 is first and foremost a policy prioritization framework aimed at finding ways to enhance the scale and effectiveness of entrepreneurial processes trying to develop regions' indigenous potential. S3 aims to promote innovation and entrepreneurship via enhanced technological diversification, embeddedness and connectivity (Foray, 2014;McCann & Ortega-Argilés, 2015, 2016a and this is to be achieved by better policy prioritization and experimentation. The idea behind S3 is that policy resources must be prioritized on those activities, technologies or sectors where a region has the most realistic chances to develop wide-ranging and large-scale impacts and which also develop and build on many different local and interregional linkages and connections (Foray et al., 2012). A common feature in this policy context must be that the entrepreneurial actions contain a sufficient degree of experimentalism and self-discovery (Hausmann & Rodrik, 2003) as this is essential in all forms of innovation and entrepreneurship.
S3 is now seen as a new policy framework that has transformed policy thinking from either largely top-down vertical sectoral approaches or primarily horizontal innovation policy programmes (focused on improving human capital, accelerating transfer of technologies, creating incubators, cluster-policy implementation) to a holistic, inclusive, place-based bottom-up and smart policy mix approach which combines both vertical and horizontal perspectives (Kyriakou, 2017;Nauwelaers, Forte, & Midtkandal, 2014). Identifying smarter goals for a given region is only a beginning, because S3 is not intended to be a one-off process, necessary simply to respond to ex ante conditionalities, but 2 S3 Platform has a repository of RIS3 from member states and regions, http://s3platform.jrc.ec.europa.eu/home rather an ongoing process of governance and policy-making upgrading (Balland, Boschma, Crespo, & Rigby, 2019;McCann & Ortega-Argilés, 2016b;Thissen, Van Oort, Diodato, & Ruijs, 2013).
With regards to entrepreneurship, S3 distances itself from traditional innovation policy and industry policy frameworks (OECD, 2013) by emphasizing the role of EDP (Foray, David, & Hall, 2009). The argument here is that local agents are best positioned to search for the ex ante knowledge and identify the unique local characteristics, assets and competitive advantages of their region, and then discover the priorities regarding their innovation resources and capacities that can lead new market opportunities (European Commission, 2012;McCann & Ortega-Argilés, 2015). Finally, this bottom-up process should result in a "collective strategy" with the broad engagement of the key actors using "an inclusive governance structure, a capacity-building toolbox, and an evaluation system" (Sotarauta, 2018, p. 4).
The smart specialization literature considers EDP as one of the central elements of S3 (Foray, 2019;Martínez-López & Palazuelos-Martínez, 2019) and probably the most ambitious element when it comes to its practical implementation S3 (del Hermosa et al., 2015;Krammer, 2017;Ranga, 2013;Sotarauta, 2018). The EDP "that it is the heart of the S3 approach is by construction an inclusive, continuous, embedded and bottom-up process" (Kyriakou, 2017, p. 5). In order to maintain the originality of place-based approaches, the European Commission does not want to narrow the scope of the bottom-up EDPs by providing precise policy recommendations which would limit many opportunities and therefore shies away from methodological normativity. This intention is quite explicitly expressed in the recently published handbook on "Implementing Smart Specialization Strategies" (European Commission, Gianelle, Kyriakou, and Cohen 2016;Navarro et al., 2014). However, one thing is increasingly certain, that an evidence-based analytical framework is highly recommended to get a clearer picture of the institutional constellation in which local agents interact (Kotnik & Petrin, 2017;McCann & Ortega-Argilés, 2014, 2016bRodríguez-Pose, 2017).
The success of any place-based policies, such as S3, is influenced by the institutional context at the regional level (Coffano & Foray, 2014). Less developed regions often have problems with their institutional capabilities and therefore EDPs are "hard to trigger and, even more so, to keep alive between administratively rigid governments and week business sectors that lack both trust and experience in mutual collaborations" (Kroll, 2019b, p. 36). According to Benner (2019Benner ( , p. 1791, the EDP has two institution-related functions: it helps to discover specific regional institutional patterns, and offer policies "either consistent with existing institutions or aiming at institutional change." Therefore, the effective regional policy requires understanding regionally specific obstacles, mainly as institutional settings (Kroll, 2019b) "instead of copying 'best practices', translating policies to a region's institutional context can be useful" (Benner, 2018, p. 14).

| Entrepreneurial ecosystem and the entrepreneurial discovery process
Entrepreneurship research has changed considerably over the last two decades. While early entrepreneurship scholars focused on the entrepreneur itself and/or on the creation of the new venture, recent researchers consider the entrepreneur not in isolation but within a context of the environment (Welter, 2011(Welter, , 2019. When someone compares the present definitions of entrepreneurship to older ones, one can recognize the movement from the individually focused one-dimensional view to the environmental/contextual multidimensional approaches up to the most recent EE concepts (Acs, Autio, & Szerb, 2014;Autio, Nambisan, Thomas, & Wright, 2018).
The EE approach makes a clear distinction between entrepreneurial outputs or activities and its antecedent, interconnected "systemic" factors. Moreover, EE scholars also differentiate potentially high impact, high growth entrepreneurial outputs from low impact entrepreneurial activities (Stam, 2015). At the same time, the nature of the connection between EE and the whole entrepreneurship process has only been emerging most recently (Audretsch & Belitski, 2017).
The notion of EDP is developed from a Kirznerian perspective (Foray, 2017;Kirzner, 1979;Roman & Nyberg, 2017) and also reflects the view of Hausmann and Rodrik (2003) regarding the phenomenon of self-discovery (Foray & Goenaga, 2013). The concept is based on the observation that "the knowledge about what to do is not obvious. It is knowledge 'of time and place'; this is local knowledge which is dispersed, decentralized and divided. It is hidden and needs to be discovered" (Foray, 2016(Foray, , p. 1433. EDP is, by nature, spontaneous, and the discovery of a new idea leading to high impact potential venture startup is a result of trial and error experimentation (Acs et al., 2014;Fiet & Patel, 2008). For opportunity discovery, we need valuable opportunities and enterprising individuals (Shane & Venkataraman, 2000). While some entrepreneurship scholars highlight the individual aspects of opportunity recognition such as traits, personal networks or prior experiences (Ardichvili, Cardozo, & Ray, 2003), others emphasize the role of the environment (Welter, 2011). Fundamentally, EE provides the locally embedded contextual, "systemic factors that interact and influence the identification and commercialization of entrepreneurial opportunities" (Audretsch & Belitski, 2017, p. 2). Successful EE supports potential entrepreneurs to be able to discover and exploit valuable business opportunities by offering growing markets, favorable culture, formal and informal infrastructure, and finance (Spigel & Harrison, 2017). While systems can be examined at a national level, ecosystems are local and regional constructs. Hence, the focus of regional entrepreneurship policy, namely the improvement of the EE, provides a fertile field for the EDP and the potential emergence of high impact startups, and ultimately regional growth.

| Smart specialization and regional entrepreneurship policy
While both S3 and EE based regional entrepreneurship policy aims to improve EDP, there are slightly different emphases from the policy perspective. For example, from the perspective of participating actors/agents, S3 is more exploratory than REP by identifying entrepreneurial agents-firms (suppliers, manufacturers, service providers), innovators, higher education institutions, research institutions-policy-makers, leaders and all stakeholders who take part directly or indirectly in the EDP (Coffano & Foray, 2014;McCann & Ortega-Argilés, 2016a;Rodríguez-Pose,-2017). Meanwhile, the REP focuses on those formal and informal, direct and indirect, institutional factors-such as markets, infrastructure, culture, talents, finance, networks, supporting institutions and services-that could lead to the emergence of high growth ventures (Spigel & Harrison, 2017;Stam, 2015). In this sense REP is more regulatory than S3.
While both S3 and EE based REP highlight the entrepreneur as the key figure in EDP, S3 is more focused on the agency while the EE emphasis is on the institutional aspects. As we demonstrate in the next section, REDI combines both the individual and the institutional sides of EE, therefore balancing both the agency and the institutional characteristics of a regional EE. Whereas S3 highlights the bottom-up nature of voluntary participation, EE based REP puts more emphasis on the institutional development, therefore it is more a top-down policy approach. However, harmonization is also a key in policy implementation talking about either stakeholders (S3 approach) or institutional components (EE) (Acs et al., 2014;Santini, Marinelli, Boden, Cavicchi, & Haegeman, 2016). Another common feature of EE based REP and S3 is that both underline the reliance on local strengths and space-based, tailor-made policy initiation (Autio & Levie, 2017;Mason & Brown, 2014;Spigel, 2017).
Summing up, S3 and ecosystem based REP view the EDP process and the emergency of high growth, innovative ventures partially differently, although these two approaches share many commonalities and are largely complements rather than substitutes, therefore opening up the possibility of informing each other in the S3 policy prioritization process. EDP can be viewed at the intersection of the RIS and the EE. From the S3 perspective, it is the interaction of the industry specialization/diversification and the improvement of the EDP that leads to the emergence of new industries and new, potentially high growth firms. The initial experience of S3 experimentation was to focus on the industry specialization (Capello & Kroll, 2016) and neglected the EDP and the new firm formation aspects. Probably this is one of the main reasons while the expected outcomes of S3 policy in innovation, development, growth and job creation failed from the initial expectations. This view was reinforced by two conceptual frameworks that emerged in the 1990s to explain the evolution of the information technological revolution: the first was national systems of innovation; the second was Porter's diamond that defined a system of regional clusters that propelled a country to prominence. Clusters and systems of innovation had two assumptions in common. First, they both argued that institutional embeddedness was important and second, they both relied on existing firms to implement and deploy the new technologies! Both of these approaches had a large theoretical literature, empirical research and policy recommendations. Because they both left out of their analysis the role of new firms that was Boyan Jovanovic's great insight-that new firms were needed to implement new technologies-they were limited in their usefulness for implementing the new information technologies (Greenwood & Jovanovic, 1999). Why new firms were left out of these approaches is an interesting subject . The systems of innovation approach was in part a Swedish discovery and helps explain both the Scandinavian disdain for startups and the European Union's unwillingness to view innovation and entrepreneurship in the same unified approach (Sandström, Wennberg, & Karlson, 2019).

| REGIONAL ENTREPRENEURSHIP AND DEVELOPMENT INDEX (REDI): STRUCTURE, METHODOLOGY, AND THE DATASET
We now ask the question of how do we measure EE in order to understand the EDP working as a bottom-up process that is viewed as the weakness of both the S3 and EE concepts? We suggest in this paper that the REDI policy tool, by gauging the EE, can help us to reveal the effectiveness of the entrepreneurial discovery process that leads to the emergence of new industries and high growth firms. It provides measurable and verifiable data at the regional level for helping with the optimization of effective S3 policy strategies.
F I G U R E 1 The conceptual model of smart specialization strategy based policies and regional entrepreneurship policy influences According to Acs et al. (2014, p. 119) entrepreneurship can be seen as "a dynamic, institutionally embedded interaction between entrepreneurial attitudes, ability, and aspirations, by individuals, which drives the allocation of resources through the creation and operation of new ventures." Originally the Global Entrepreneurship Index (GEI) was created on the national level (Acs et al., 2014;Acs, Szerb, Ortega-Argilés, Aidis, & Coduras, 2015). While the entrepreneurship literature has acknowledged the importance of national differences in entrepreneurship, regional differences have also been confirmed to be equally or even more essential underlying the importance of the regional context in the individual entrepreneurial decision-making (Bosma, 2009;Mason & Brown, 2014;Stam, 2015).

| The REDI structure and calculation
REDI has been formed to measure the level of the entrepreneurship ecosystem in a regional context. This index uses measures of individual-level entrepreneurial attitudes, abilities, and aspirations as weights to adjust the magnitude of institutional and contextual factors in regulating the quality of the entrepreneurial dynamics. Moreover, the index helps to provide guidance on resource allocation decisions in the economy towards high-productivity uses (Acs et al., 2014). In the REDI theoretical concept, entrepreneurs are seen as operating a trial-and-error resource allocation dynamic by mobilizing resources to pursue perceived opportunities (Qian, Acs, & Stough, 2013). This feature is also along the lines of the smart specialization's entrepreneurial discovery (Foray et al., 2012). However, contextual conditions moderate the potential impact of such individual-based resource allocations-for example, the availability of high-quality human capital would regulate the growth potential of a new venture. Scarcity in high-quality human capital would constrain the ability of new ventures to meet their recruitment needs to support rapid growth. In the REDI, therefore, framework conditions are not seen as direct drivers of entrepreneurial attitudes, abilities, and aspirations, but rather, as institutional and contextual regulators.
REDI is a systemic tool, in the sense that the system components are thought to "co-produce" system-level outcomes. In practice, this property is operationalized through the penalty for bottleneck (PFB) algorithm, which "penalizes" strong pillars for gaps-or bottlenecks-in pillar-level performance. This means that the REDI methodology is potentially useful for profiling entrepreneurial ecosystems, where a similar co-production dynamic is thought to be in operation (Autio et al., 2018;Autio & Levie, 2017). As the REDI methodology is able to highlight gaps in the entrepreneurial dynamic, it also provides a potentially useful template to guide policy action. The first version of REDI has been developed under the professional supervision of DG REGIO, 3 and its importance is demonstrated by the fact that both the index itself and the results of the regional investigation have been included in the EU's 6th Cohesion Report (European Commission, 2014b). Table 1  building blocks are the 14 pillars which contain, simultaneously, 36 regional individual, regional and country-level institutional variables. As compared to the Global Entrepreneurship Index (GEI) the institutional variable components of REDI are much richer. Regional level variables aim to reflect the local spillover effects of agglomeration (size of the region, market potential), connectivity, networking/clustering, social capital, education systems, human capital, the effects of knowledge spillover and innovation, the role of regulation, the quality of governance and also of finance. Each pillar was created as a product of individual-and institutional-level variables. Careful scrutiny of the relative differences between individual pillars, both within a given region and across benchmark regions, can provide good initial guidance for the search for prospective strengths and weaknesses across regions from a benchmarking perspective.
These pillars comprise three sub-indices: entrepreneurial attitudes (5 pillars); abilities (4 pillars); and aspirations (5 pillars). 4 The entrepreneurial attitude (ATT) sub-index aims to identify the attitudes of a region's population as they relate to entrepreneurship. Opportunity recognition, start-up skills, risk acceptance, networking building potential, and cultural perceptions formulate these attitudes. The entrepreneurial abilities (ABT) sub-index is principally concerned with measuring certain important characteristics of both the entrepreneur and the start-ups with high growth perspectives such as start-up motivation, human capital, technology-absorption potential, and market nicheidentification capabilities. The entrepreneurial aspiration (ASP) sub-index refers to the distinctive, qualitative, strategy-related nature of entrepreneurial start-up activity. Product and process innovation, growth strategy, cluster formation, internationalization, and finance frame these aspirations.

| REDI as a policy tool
A key feature of the REDI is its unique methodology based upon the Penalty for Bottleneck (PFB) method that makes it possible to incorporate the system-perspective to the calculation method. Practically it means that the value of each pillar in a region is penalized by linking it to the score of the pillar with the weakest performance in that region. This simulates the notion of a bottleneck, and if the weakest pillar were improved, the particular sub-index and ultimately the whole REDI would show a significant improvement. On the contrary, improving a relatively high pillar value will primarily enhance only the value of the pillar itself, and in this case, a much smaller increase of the whole REDI index can be anticipated. Moreover, the penalty is higher if differences between the bottleneck and the actual pillars are higher (Acs et al., 2014). This idea reflects the classical notion of public policy T A B L E 1 The structure of the regional entrepreneurship and development index

Structure of the REDI 3 sub-indexes/14 pillars
National and regional institution variables where the aim is to correct for market failures (Bator, 1958;Stiglitz, 1989) and also the growth policy notion of the "second-best" targeting to improve the most binding constraints of development (Hausmann, Rodrik, & Velasco, 2008). Moreover, region-specific bottlenecks can also be interpreted as restraints in the entrepreneurial discovery process. Here, the main aim of public/entrepreneurship policy is to correct for entrepreneurship ecosystem failures.

Regional level individual variables
The advantage of REDI, as compared to other indices, is its capability to incorporate both the individual and institutional contexts in one model. Emphasizing both angles the REDI acknowledges the multidimensional nature of entrepreneurship (Wennekers & Thurik, 1999), against those measures which are still one-dimensional and frequently using start-up, self-employment or business density rates (Iversen, Jorgensen, & Malchow-Moller, 2008).
These output measures are problematic because they mix together many different types and qualities of start-ups or business organizations from "muppets" to "gazelles" (Nightingale & Coad, 2013). Recent conceptual developments of the "entrepreneurial ecosystem" represent the latest evolution about the measurement of entrepreneurship (Autio et al., 2018;Pitelis, 2012;Spigel, 2017) and the REDI provides the most structured and comprehensive approach to entrepreneurial ecosystem measurement, up to now.
In terms of the possible limitations of REDI, one of the fundamental restrictions is that any regional regulating context behind entrepreneurship ecosystems might be more complex than an index such as the REDI could fully capture. Understanding the REDI index outcomes is less straightforward than in the case of one-dimensional measures (using only one variable), but a potential criticism of our method-as with any other index-might be the apparently arbitrary selection of indicators and the neglect of other important ones. In order to limit this risk, in all cases, we aimed to collect and test alternative individual/institutional factors before making our selection. Of course, our choice is constrained by the limited availability of data.
The REDI methodology also makes several simplifying assumptions. First, in assigning equal weights to each pillar, it thereby assumes that all pillars always contribute equally to the outcomes of the entrepreneurial dynamic.
In so doing, the method also assumes that one best configuration for the entrepreneurial system exists-one in which all elements are maximized and in balance. Second, is the arbitrary selection of the magnitude of the penalty that is based on a "rule of thumb" 10% penalty on the average. The other problem is that we cannot exclude fully the potential that a particularly good feature can have a positive effect on the weaker performing features.
While this could also happen, many of the entrepreneurship policy experts hold that policy should be based on the correction of market failures (Acs, Audretsch, Lehmann, & Licht, 2016;Audretsch, Grilo, & Thurik, 2007;Lundström & Stevenson, 2006). The identification and the correction of the weak links is also the main focus of the efficient operation of the entrepreneurship ecosystem (Mason & Brown, 2014;Stam, 2015). The REDI focus on system bottlenecks tends to focus attention to fixing gaps, which potentially may come at the cost of maximizing system strengths. An important novelty of our index-building is the way the pillars are combined (aggregated) into subindices. Most indices simply use the (weighted) average of the pillars; others apply a dimension reduction methodology, such as factor analysis. We provide a different approach. The basic assumptions of our methodology are that the performance of the system is determined by its weakest performing part and the pillars can only be partially substitutable with one another. The PFB relates the pillar values to the lowest pillar value.
The penalty depends on the magnitude of the differences; for greater deviation, the penalty is greater. The PFB provides valuable policy suggestions for enhancing entrepreneurship by improving the weakest pillar in the system. Altogether, we claim that the PFB methodology is theoretically better than the arithmetic average calculation. The average method has no any theoretical/conceptual basis by assuming equal weights to all the components. PFB based weighting addresses directly the system failure and provides a theoretical basis to policy suggestions. However, the PFB adjusted REDI is not necessarily an optimal solution since the magnitude of the penalty is unknown.
The most important message for economic development policy is that improvement can only be achieved by abolishing the weakest link of the system which has a constraining effect on other pillars, and consequently on the EE as a whole.   Table 2. Individual variables are described in detail in Appendix Table A1.

| The REDI dataset
Since the GEM dataset lacks the necessary institutional variables, we complemented it for the index with other widely-used and relevant data derived from a variety of available sources. For a detailed description and sources of institutional data see Appendix Table A2. An important note is that the benchmarks were calculated by taking into

| REDI AND ITS CONTRIBUTION TO SMART SPECIALIZATION STRATEGIES: A PRACTICAL APPLICATION
In this section, we present how REDI could contribute to S3 policy implementation by providing a solution to the following four S3 policy caveats emphasized by the literature. The sharpest critique regarding the S3 concept that it does not emphasize enough (at least in the beginning) that the success of the S3 strategies is largely determined by the institutional setup in which nations or regions are embedded. Capello andKroll (2016, p. 1396) call attention the fact that nations and regions characterized by diverse institutional capacity have to face and handle various challenging place-based situations, and therefore they argue that "smart specialization could provide a common political rationale." The notion of institutional capacity refers to the ability of nations/regions supporting or hampering the absorption of those new aspects, ideas which continuously attack their institutional arrangement.
The limitations of the S3 concept are particularly apparent for less developed regions (LDRs), which struggle with the phenomenon of regional innovation paradox (Oughton, Landabaso, & Morgan, 2002). It means that LDRs are deficit regarding their institutional capacity (Rodríguez-Pose, 2013), and therefore they have to face many "atypical obstacles", such as lack of creativity, limited marked opportunities, top-down style of governance lieu of regional autonomy, heavy reliance on external resources etc. (see in detail Krammer, 2017). While particularly the LDRs have a greater need for strategic reshape of their policies and innovation-related institutional setup to avoid the danger of stalling in regional "lock-in situations" (Capello & Kroll, 2016;McCann & Ortega-Argilés, 2016a). The S3 requires less developed regions to get rid of their old unfavourable institutional setup and introduce novel institutional arrangement. Therefore, as Morgan (2017, p. 17) indicates "the detrimental power of policy path dependence" should be seriously considered as well. (p. 17).
According to Capello and Kroll (2016)  Grillitsch (2016, p. 29-30) emphasizes the importance of "institutional harmony" by stressing that conflicting institutions can discourage trust among different stakeholders. Regions lacking institutional harmony can "fall into an institutional conflict trap that might not only dissipate collective efforts but also impede trust-based social relations or prevent them from being built" (Sotarauta, 2018, p. 6.).
S3 needs to ensure the continuous character of the entrepreneurial discovery process. The process should avoid reducing EDP to a mere consultation on top-down choices, based on ex cathedra analysis. Instead, stakeholders should be present at the creation (Kyriakou, 2017, p. 5).Next, we present how REDI could contribute to these S3 policies related to caveats. We follow the four points listed at the beginning of this section. All sub-sections finish with a proposition and we also note some potential caveats.

| Smart specialization strategies in less developed regions
The first point refers to a typical practical problem of the application of S3 policy in less developed regions (LDRs). It seems that some development of the individual and institutional capabilities are required for the successful specialization policy. If basic requirements are missing then there is no place for industry specialization. REDI is able to measure the level of the entrepreneurship ecosystem on a 0-100 point scale; hence providing an overall picture about the level of existing preconditions for specialization. Table 3 contains the REDI scores for each of the 125 regions.
As can be seen later the ex post evaluation of the implementation will obviously reveal the effectiveness of this strategy. If it was unsuccessful, it is a clear sign for the region that its EE is still not ready for the proper for the execution of S3 strategies, while basic regional conditions are still missing.

| Institutional and individual development
The second critical points referred to the problem of identifying the deficits in the local institutional capacity. While but also the widely interpreted institutional features. Although, REDI pillars are created as the interaction of the various individual initiations and institutional contexts it is possible to separate them. For doing it, we have followed the REDI score calculation methodology and used the non-penalized institutional and individual components scores (in details see Szerb et al., 2017). By calculating the share of institutional scores to individual (SII) ones for every region we can identify if a specific region is relatively weak in the institutional component (share is lower than 1) or in the individual features (share is higher than 1). Figure 1 presents the connection between the SII and REDI scores.
In Figure 2, we highlighted the 1.00 SII score with a dotted line, as an optimal share. SII scores range from 0.38 (Macroregion doi, Romania) to 1.33 (Pohjois-ja Ita-Suomi, Finland). The third-degree polynomial line in Figure 2 implies that REDI scores are determined mostly by the institutional quality that is along the line of the institutional economics claim.
If SII is lower than 1 then a certain region has a weakness in its institutional setup, so its policy should aim to alleviate these institutional deficiencies. This is the case in many regions of less developed countries. If SII is higher

Proposition 2b Regions with relatively higher individual development should improve their institutional features.
Caveat As REDI is determined more by institutional factors as compared to individual ones, the selected optimal share of one may not be appropriate. Further research is necessary to search for an optimal share of institutional and individual variables.
The connection between the share of the individual and institutional variables and REDI scores

| Harmonization of the components
Regarding the institutional harmony, the REDI methodology is based on the harmonization of its 14 pillars, therefore it is particularly suitable for identifying the potential conflict trap domain. REDI measures not only the level of regional institutional capacity but also the interconnected effect of the individual and the institutional factors. Differences in EE are clear when we classify the countries according to the 14 pillars of entrepreneurship (Table 4).  more successful regions in that particular feature. The EU's best practices could be helpful to look for practical solutions. 6 Differences across regions are even higher if we examine individual regions and not clusters. Region-specific obstacles as bottlenecks, that is, lower score pillars can be identified both in higher developed regions and LDRs.
Proposition 3 EE is optimal when the 14 pillars of REDI are harmonized. The region-specific policy should focus on increased balance of the 14 pillars of EE. Without the harmonization of the components, the EE cannot fully exploit its potential opportunities, resources are wasted, and the whole EDP is inhibited.
Caveat 3 While classical public policy intervention is based on the alleviation of market, in this case system failures, we cannot rule out the possibility that higher-performing components could counterbalance weakly performing elements of the system.

| Policy optimization
While early S3 policy suggestions centered on bottom-up policies, over time, a more realistic top-down and bottom-up mix approaches have gained space (Foray, 2017(Foray, , 2019. As a systemic index for entrepreneurial ecosystems underpinnings of S3, the REDI provides an opportunity for enhancing the EE by alleviating the bottlenecks and optimizing the additional resources. The PFB algorithm penalizes system pillars according to gaps exhibited by the most poorly performing pillar or pillars, that is, the bottleneck pillar(s). As explained above, the idea is that systems with strong weaknesses cannot fully leverage their strengths, or to put it another way, weakly performing bottleneck pillars hold back entrepreneurial ecosystems performance in situations where system pillars coproduce system performance. A corollary implication of this assumption is that entrepreneurship policy efforts supporting EE can work most effectively when it seeks to alleviate systemic bottlenecks. A notable advantage of REDI is its capability to show the relative size and magnitude of the bottleneck(s). Instead of further enhancing systemic strengths, it may be more effective to alleviate the bottlenecks that prevent the system from fully leveraging its strengths.
Using the logic above, we performed a set of simulations exploring the effect of regional entrepreneurship policies designed to alleviate systemic bottlenecks.
The PFB method calculation implies that the greatest improvement in system performance can be achieved by alleviating the weakest performing pillar-the bottleneck pillar. In the simulation, each bottleneck pillar is alleviated to a point where it ceases to be a bottleneck. At this point, any further effort is allocated to the second-most binding constraint within the system, again up to a point where this constraint is no longer the most binding constraint within the system. By successively alleviating most binding constraints, our simulation therefore provides an idea of how policy effort should be allocated to achieve an "optimal" outcome, defined as the largest possible increase in the REDI index score.
Our simulations seek to identify the benchmark allocation of policy effort that targets to increase the REDI index score each of the 125 EU regions by 5 REDI points. Table 5 shows the result of our optimization exercise for the selected country regions of Denmark, Estonia, France, and Hungary. In this case, the additional units are distributed across constraining pillars until a 5-point increase in the REDI index score has been achieved in each region. The percentages indicate the distribution of additional policy effort across the constraining pillars, reflecting the relative severity of the pillars in the respective region. In Table 5 Total effort represents all the amount of the inputs that the region is spending for entrepreneurship in natural units. 7 It is the sum of the average normalized values of the 14 pillars. The percentage numbers under the pillar names are the unit (amount) of inputs necessary to add to the particular pillar value in order to reach the required alleviation of the pillar constraint. Zero value indicates that no additional input is needed, as the pillar is currently not a binding constraint. The total additional effort column provides the overall sum of the required additional units. Larger numbers indicate that more inputs are necessary for overall performance improvement in a given region, as compared to regions with lower scores. More uneven profiles are ones where significant relative differences exist across different pillars-in particular, where some pillars exhibit significantly lower values than other pillars. Thus, a more uneven profile signals the existence of more pressing constraints. In addition, an uneven profile also means that greater benefit can be achieved by focusing most of the additional policy effort into a small number of bottleneck pillars because bottleneck alleviation enables the system to more fully utilize its existing strengths. The most efficient outcome can be achieved in regions where there is one single pressing bottleneck, which is able to absorb all of the additional policy effort required to produce a five-point increase in the REDI index value.
According to Table 5, there are huge differences in the allocation of the inputs. For example, in the case of Midtjyland (DK04), the five-point increase can be produced by alleviating the Globalization bottleneck alone. This is reflected in the relatively small additional input allocation required (0.009 units). Other Danish regions also have such an 'uneven' profile requiring additional inputs in Globalization, Finance, Process Innovation, and High growth. In contrast, Estonia has a relatively "even" profile, and the simulation suggests that additional policy effort needs to be distributed relatively evenly across Cultural support, Globalization, and Finance pillars. This also means that there are few pressing bottlenecks in the Estonian regionthe implication is that greater additional inputs are required to achieve a five-point increase in the Estonian entrepreneurship system performance (0.292 units). French regions show larger differences in EE as compared to Danish ones: Île de France (FR1) is one of the leading regions with well-balanced pillars. Sud-Ouest (FR6)-similar to DK4 has only one bottleneck that is Globalization. Centre-Est (FR7) also needs to improve only one pillar, in particular, Startup skills. Hungarian regions are at the bottom of the ranking, but the entrepreneurship system profiles of the country's regions show a relatively well-balanced performance. As a consequence, a high amount of additional inputs is necessary to reach a five-point increase in the REDI scores (0.214-0.520). In addition, multiple pillars need to develop in the Hungarian regions, mostly Cultural support, Risk perception, and

Financing.
While REDI cannot mobilize local players the analysis of the individual and institutional components of the REDI pillars makes it possible to identify the lack of vital local individual and/or institutional features of the EE that could negatively affect EDP.
Proposition 4 REDI provides a methodology on the optimal allocation of additional resources to improve regional EE. This region-specific policy mix could contribute to increased EDP and the emergence of high growth, innovative startups, and ultimately regional growth.
Caveat 4a REDI, by itself, is not appropriate to offer assistance on sectoral/industry specialization.
Caveat 4b REDI cannot provide guidance on how to promote the learning process or a solution on how to make local players involved in S3. REDI only detects the lack or the presence of particular local factors influencing EDP. If many factors are missing in the particular region then it is worth thinking about a more appropriate development strategy for S3.

| SUMMARY AND CONCLUSION
Moving away from traditional, top-down and largely uniform innovation policies, the EU has turned to a more bottom-up, region-specific, place-based policy, spearheaded by the Regional Research and Innovation Strategies for Smart Specialisation (RIS3) or simply the S3 agenda. By the mid-2010s, all EU regions have developed their own S3 priorities, although the practical implementation of S3 is still only partially successful. This has encouraged researchers and politicians to further develop S3 both from the theoretical and from the practical, policy-action sides.
S3 includes industry prioritization and reliance on the entrepreneurial discovery process (EDP). While early S3 experiences focused on the industry prioritization and neglect EDP, here we aim the spotlight on EDP. While EDP is a self-governed process itself based on trials and errors, its functioning is determined by the entrepreneurship ecosystem development of the particular region. Hence, the improvement of EE could contribute to more successful EDP and therefore of the overall S3 process.
In this paper we present the Regional Entrepreneurship and Development Index (REDI), as a holistic measure of EE. Our methodology is based on a systemic and multidimensional approach, and we demonstrate how the REDI could contribute to S3 policy implementation by providing some improved solutions to four S3 policy caveats. First, we provide a measure of the entrepreneurship ecosystem for a mix of 125 NUTS 1 and NUTS2 European Union regions. If basic conditions are missing then policies should focus more on improving the entrepreneurial discovery process rather than trying to specialize. Second, REDI is able to identify institutional and individual weaknesses at local levels. Third, REDI methodology provides a comprehensive view of the harmonization of the 14 pillars of the entrepreneurship ecosystem for each region, hence able to identify potential policy domains. Fourth, with the help of the penalty for bottleneck methodology REDI presents some simulations on how additional policy efforts could be optimized over the 14 pillars to improve the REDI scores and enhance EDP. REDI based suggestions are place-based and they are parallel to the tailor-made policy nature of S3. While S3 industry prioritization is based on the identification of local strengths, REDI improvement can achieved by improving the weak features of the EE. We found that without optimizing the entrepreneurial ecosystem, the industry prioritization alone may not be successful because of the inability of the ecosystem to be able to nurture high growth potential ventures.
Even if the assumptions are restrictive it should be kept in mind that the policy portfolio simulation offers many benefits that go above and beyond what traditional indices can offer. The most important benefit is in drawing attention and highlighting system dynamics in regional EEs. This reinforces a systemic perspective to policy analysis and design over a traditional, siloed standpoint, exactly as is argued for in the smart specialization agenda. A policy scenario simulation, which highlights interconnections within the system, forces policy analysts and policy-makers to think outside individual policy silos and consider the system performance as a whole. This, then, should help smart specialization policy-makers also to think about trade-offs between different allocations of policy effort and to judge their effectiveness against a system-level performance benchmark. If correctly used, therefore, a policy portfolio simulation should facilitate agreement on system-level policy priorities for driving smart specialization by promoting awareness of different policy scenarios.

Opportunity Recognition
The percentage of the 18-64 aged population recognizing good conditions to start business next 6 months in area he/she lives,   were forced to pay a bribe in the last 12 months to obtain any health care in the region/area.

Business Environment
Quality of Governance

Business Strategy
The Nature of competitive advantage was multiplied with the unweighted average of the three indicators of the Business Sophistication variable. marketing, distribution, advanced production processes, and the production of unique and sophisticated products). spill over into the economy and lead to sophisticated and modern business processes across the country's business sectors. The variable of Nature of competitive advantage is a part of the Technological readiness pillar. The data captures answers to the question: "What is the nature of competitive advantage of your country's companies in international markets based upon?"

Nature of Competitve Advantage
(1 = low-cost or natural resources; 7 = unique products and processes). . Concentration of Financial Sector regional Regional employment in financial services sector as percentage of total regional employment.

Employment, K-
Cluster Observatory