Examining the roles of universities in place-based industrial strategy: which characteristics drive knowledge creation in priority technologies?

ABSTRACT Industrial strategies designed to promote innovation in a set of priority technologies through university–industry collaboration essentially institutionalize a triple helix approach to economic development. Yet, treating universities as a generic resource leaves a question mark as to which institutions are most likely to be most useful. In addition, prior evidence of uneven regional distribution of research income in these technologies suggests that place-based interventions may merely lock in pre-existing inequalities. Therefore, by controlling for spatial and temporal variations among UK universities, this paper examines whether their ability to generate knowledge in these priority technologies is dependent upon their entrepreneurial or engaged nature, and strategic orientation. Using data from the UK Higher Education Business & Community Interaction (HE-BCI) survey, the analysis finds that entrepreneurial activities such as higher levels of licensing income, start-ups and patents are associated with higher levels of research income in these priority technologies. Furthermore, higher levels of income from engagement with businesses through collaboration and contract research are also associated with higher research income in these priority technologies, while strategic orientation has no effect.


INTRODUCTION
The current approach to industrial strategy in the UK is based on three clear criteria: (1) the identification of a set of priority technologies as a focus for innovation (healthcare, medicine, clean energy, battery technologies, driverless cars, space technology and artificial intelligence); (2) the promotion of formal collaborations between firms and UK universities in order to utilize their knowledge and expertise to develop new products and processes based on these technologies; and (3) the need to 'rebalance' and 'level-up' the economy across all regions. This approach to industrial strategy is clearly influenced by proponents of the Fourth Industrial Revolution, who view enabling technologies such as those identified above as underpinning Smart Specialisation Strategies which promote innovation through their utilization (Bailey et al., 2018;McCann & Ortega-Argilés, 2015). In addition, explicitly harnessing the knowledge and expertise of multiple actors including government, universities and business institutionalizes a triple helix approach to economic development (Fini et al., 2018;Ranga & Etzkowitz, 2013;Reischauer, 2018;Steenkamp, 2019), while the concurrent aim of rebalancing the economy clearly highlights an opportunity for placebased policies to promote its objectives (Fothergill et al., 2019;Johnston & Wells, 2020).
As there has been little systematic examination of the roles of universities in place-based industrial strategies within the extant academic literature, the UK context provides an opportunity to explore this more deeply. Despite repeated references to universities within recent policy documents, little evidence is presented to outline how universities may successfully engage in this process. Indeed, universities are diverse organizations that vary in terms of their size, resources, specialisms, research capacity and engagement capabilities (Hewitt-Dundas, 2012;Huggins et al., 2012;Laursen et al., 2011;Martin & Turner, 2010). We argue that the institutionalized triple helix approach to industrial strategy draws implicitly on two ideal types of university, the 'entrepreneurial university', focused on commercialization activities such as licensing income, patenting and developing spinouts (Perkmann et al., 2013;Philpott et al., 2011;Slaughter & Leslie, 1997), and the 'engaged university' which focuses on the co-creation of knowledge through collaborative research, the provision of training activities, and the use of facilities and equipment (Breznitz & Feldman, 2012;Sanchez-Barrioluengo & Benneworth, 2019;Trippl et al., 2015). Yet, no consideration is given to which of these types or characteristics may be important.
In addition, the location of the universities may promote behaviours, resources, networks, culture, and competences that stimulate innovative and entrepreneurial behaviours (Huggins & Thompson, 2013, 2019; underlying factors that may also provide an explanation for the uneven spatial distribution of research funding in the priority technologies, suggesting that research, knowledge, and expertise in some technologies may be not be equally available to all regions (Johnston & Wells, 2020). Therefore, an institutionalized triple helix approach may instead lock in spatial inequalities rather than reduce them. Consequently, this paper addresses an important question: given the spatial inequalities in the distribution of cutting-edge research funding, do the characteristics of universities alone signal their ability to generate cuttingedge knowledge in the priority technologies?
Utilizing data covering the 10-year period from 2006/07 to 2016/17 from the UK's Higher Education Business & Community Interaction (HE-BCI) survey, this paper employs a panel model with temporal and spatial fixed effects that specifies that the ability to generate knowledge in the priority technologies is a function of the university's entrepreneurial and engagement activities and strategic orientation. By controlling for spatial and temporal influences, the paper presents a systematic analysis of the factors that influence a university's ability to generate cutting-edge knowledge. Controlling for these effects reveals that income from entrepreneurial activities such as licensing IP, patenting, and creating spinouts are all positively related to cutting-edge knowledge generation. Importantly, income from collaborative and contract research are also positively related to cutting-edge knowledge generation, suggesting that a combination of entrepreneurial and engaged characteristics are important. However, the strategic orientation of the university does not have a significant effect.
Given these findings, we argue that examining the generation of cutting-edge knowledge at the regional level may mask the importance of the universities' organizational characteristics. Therefore, place-based policymaking should seek to promote university-industry collaborations where the university partner is engaged in both entrepreneurial and engagement activities, regardless of their location, in order to ensure that university-industry collaboration matches businesses with the most appropriate university partner. Consequently, an institutionalized triple helix approach to place-based industrial strategy needs to ensure that global knowledge networks are utilized in the course of promoting local innovation.
The paper is structured as follows: the second section presents the conceptual and theoretical background through a discussion of the role of universities in the industrial strategy. The third section outlines the data used in the empirical analysis and the analytical techniques employed. The fourth section presents our findings, while the fifth section concludes and discusses their implications.

Industrial strategy and the Fourth Industrial Revolution
Over the last 20 years there has been growing interest in the enabling features of industrial strategy at international (Barca et al., 2012), national (Fagerberg, 2018), and regional levels (Bailey et al., 2018). Modern approaches to industrial strategy are based around the idea that a Fourth Industrial Revolution facilitated by technological advances is driving the reorganization of economic activity, enabled by an institutional configuration that supports these changes (Andreoni & Chang, 2016;Bailey et al., 2015;Block & Keller, 2011;Reischauer, 2018;Schwab, 2017). Consequently, proponents of the Fourth Industrial Revolution highlight the influence of 'enabling technologies' such as advances in information technology, materials engineering, and medicine that promote significant change across the whole economy (Ciffolilli & Muscio, 2018;Lepore & Spigarelli, 2020). From a regional development perspective, the Fourth Industrial Revolution is viewed as permitting the implementation of Smart Specialisation Strategies through the utilization of these new technologies in regionally embedded industries (Bailey et al., 2018;McCann & Ortega-Argilés, 2015). The Smart Specialisation process involves the identification of key priorities for regions and then building on their strengths through promoting innovation in a broad sense, i.e., the introduction of new products, processes, and technologies (European Commission, 2012). Indeed, it is argued that Smart Specialisation is most effective where a region can build on pre-existing knowledge with respect to particular technologies (Montresor & Quatraro, 2019) in conjunction with the flexibility to enable regular adaptations based on changing conditions in order to ensure regional economies are constantly responding to change (Foray et al., 2012). Therefore, the intellectual link to placebased policymaking is clear (Morgan, 2017).
Universities and place-based industrial strategy: the UK context The cornerstone of the UK's Industrial Strategy is the identification priority technologies including driverless cars, batteries, clean energy, medicine, healthcare, space technologies, robots, and artificial intelligence. These technologies form the basis of future innovation efforts designed to increase expenditure on R&D as a percentage of GDP from 1.7% to 2.4% by 2027 (HM Government, 2017). Through promoting partnerships between universities and business, industry organizations, innovation organizations, the strategy gives credence to Reischauer's (2018) argument that the Fourth Industrial Revolution promotes triple helix approaches to economic development, in particular singling out universities as an important resource for knowledge generation and promising to develop 'innovation clusters' around universities to bring together 'world-class research, business expertise, and entrepreneurial drive' (p. 67).
Reference is made to universities in three of the six themes; (1) the development of knowledge in the key technologies; (2) the commercial exploitation of the science base; and (3) supporting 'local innovation ecosystems'. The implication is that high quality university research is widespread throughout the UK, suggesting an aim that is inclusive of all universities and regions, with 98 references made to 'universities' compared with only two references to 'leading universities'. Therefore, the strategy is clearly based on the assumption that all universities can contribute to the generation of knowledge and research in the priority technologies, while the emphasis on the local innovation ecosystem belies a place-based approach to policymaking.
Given the policy focus on university-industry collaboration, we argue that policymakers place a clear emphasis on both the 'entrepreneurial' and 'engaged' characteristics of universities, which stress their focus on industrial collaboration activities (Centobelli et al., 2019;Clark, 1998;Kirby et al., 2011;Sanchez-Barrioluengo & Benneworth, 2019). While an increasing orientation towards, and engagement in, industrial collaboration, or third mission activities, is a key characteristic of both entrepreneurial and engaged universities, they cannot be considered to be an isomorphic construct (Fuller et al., 2019;Kitagawa et al., 2016;Philpott et al., 2011). Indeed, as Philpott et al. (2011) highlight, a university's entrepreneurial/ engaged characteristics are more of a continuum than a dichotomy.
First examining the entrepreneurial university, these institutions are seen as embedded in the economic and social fabric of society, taking an active role in economic development through reacting to the new demands made on them by the evolving economy, expanding and rethinking their strategies in order to set agendas accordingly (Clark, 1998, pp. 4-5). The entrepreneurial university is typically viewed as both an incubator and catalyst for development through acting as a conduit for the exploration and exploitation of knowledge (Etzkowitz, 2003;Kirby et al., 2011;Metcalfe, 2010). Consequently, the entrepreneurial university has an economic focus which places greater emphasis on commercialization activities, or so called 'hard' factors such as generating income from Intellectual Property (IP), patenting and the creation of spin-out firms, (Perkmann et al., 2013).
Therefore, the implicit assumption is that the entrepreneurial characteristics of universities, focusing on commercialization, may promote greater embeddedness within the industrial base enabling them to generate higher levels of knowledge in the priority technologies. This assertion is tested through Hypothesis 1: Hypothesis 1: Commercialization activities such as licensing, patenting and the creation of spinout firms will have a positive influence on the generation of cutting-edge knowledge in UK universities.
The 'engaged university' shares these characteristics but has a subtly different focus, combining commercialization activities with industry engagement, or so called 'soft factors', such as collaborative research with industrial partners and providing facilities and technical and training services to firms (Perkmann et al., 2013;Sanchez-Barrioluengo & Benneworth, 2019). Consequently, the engaged university also has a social focus, contributing to regional development through a broad range of activities (Breznitz & Feldman, 2012;Thomas & Pugh, 2020;Trippl et al., 2015). These engagement activities embeds the university into the ecosystem allowing an understanding of the needs and requirements of other actors (Breznitz & Feldman, 2012;Sanchez-Barrioluengo & Benneworth, 2019) and act as an anchor institution (Goddard et al., 2014). The idea that these broader engagement activities are the key to generating cutting-edge knowledge is examined in Hypothesis 2.
Hypothesis 2: Business engagement activities such as collaborative research and providing facilities, technical and training services will have a positive influence on the generation of cutting-edge knowledge in UK universities.
The strategic orientation of the university is often overlooked in terms of their knowledge generation activities (Giuri et al., 2019). Indeed, the entrepreneurial and engaged nature of universities can vary according to the priorities placed upon them by the institutions themselves (Bercovitz & Feldman, 2006). Furthermore, as there is no consensus on how third mission activities should be carried out, (Knudsen et al., 2019), how the attitudes of academics and the incentives provided to them may affect outcomes (Guerrero et al., 2016), or how technology transfer should be organized (Giuri et al., 2019). Therefore, the strategic orientation of a university cannot be ignored when it comes to understanding their propensity for creating cutting edge knowledge. This is outlined in Hypothesis three: Hypothesis 3: The strategic orientation of university, incentivizing and organizing third-mission activities has a positive effect on the generation of cutting-edge knowledge in UK universities.

Data and variables
We utilize data from both the Higher Education Business & Community Interaction (HE-BCI) survey, an annual survey of universities income from various activities and the UK Research and Innovation's (UKRI) Gateway to Research database, a searchable source of all publicly funded research projects in the UK. The HE-BCI survey is a comprehensive annual survey of industrial collaboration activities undertaken by UK universities. Given the survey is undertaken annually, is completed by technology transfer officers, and universities are legally obliged to complete it, it provides a detailed and reliable insight into the industrial collaboration activities of UK universities. Indeed, due to its comprehensive nature, and publicly available nature (see https://re.ukri.org/knowledgeexchange/the-he-bci-survey/), the HE-BCI survey is widely used as a data source in academic studies of UK universities (Fuller et al., 2019;Guerrero et al., 2015;Sanchez-Barrioluengo & Benneworth, 2019;Zhang et al., 2016). While previous analysis has been restricted to observing individual years, our analysis utilizes data across the period 2006/07-2016/17 to provide a more thorough examination through examining fluctuations in income over the period.
The Gateway to Research website (https://gtr.ukri. org/), a searchable database of publicly funded research projects in the UK, was used to identify all research projects related to the priority technologies. A two-stage systematic search strategy was then used to identify relevant projects; the first stage examined both the title and abstract to determine the focus. Following this, we then searched the project abstracts to identify whether the focus was the production of new knowledge and technology, and those where the focus was on application. For example, using this technique we were able to distinguish between projects that sought to develop new satellite technology, components, or equipment for satellites and those which sought to use satellites to examine phenomena (e.g., remote sensing). Through this process, we identified 5532 projects, which accounted for more that £2.4 billion of research funding between 2006/07 and 2016/17. These projects were broken down as follows: robots, 242; artificial intelligence, 238; driverless cars, 20; space and satellite technology, 606; clean energy, 140; healthcare, 1515; medicine, 2173; and battery technology, 598.
The HE-BCI survey was also the primary source of data for explanatory variables designed to capture the types of business engagement activities in which each university engaged. Following Sanchez-Barrioluengo and Benneworth (2019), we examine commercialization through activities as IP income, number of patents and number of spinouts created and engagement activities through total income from consultancy, contract research, collaborative research, continuing professional development (CPD) activities, the utilization of equipment and resources within the university and regeneration activity, In addition, we enhance our model through capturing the strategic intent of each university towards industrial collaboration using data from Part A of the HE-BCI survey. Therefore, several dummy variables were developed from answers to questions examining the primary focus of a university's external engagement strategy (businesses or otherwise), the existence of strong incentive for academics to engage with businesses, the existence of a majority owned subsidiary for commercial exploitation of knowledge, and whether the university possesses an incubator facility for new businesses (Table 1).
Focusing on those institutions that were granted university status, i.e., the power to award their own degrees, prior to the 2006/07 academic year 149 universities were included in the dataset as they had complete data for all variables over the time period. In addition, where mergers had occurred (for example, the Institute of Education merged with University College London in 2014, the University of Wales Trinity St David was formed from mergers of University of Wales Lampeter, Trinity University College, and Swansea Metropolitan University in 2013, and the University of South Wales was formed from the merger of the University of Glamorgan and University of Wales, Newport in 2013) all data were combined for the period.

Empirical model and estimation strategy
To examine the influence of a university's business engagement and collaboration activities on its ability to generate knowledge in the priority technologies of the industrial strategy, we estimate by employing a linear unobserved random effects panel model. To analyse the impact of all effects adjusted for temporal and spatial fixed effects, we estimate the following functional form: where y it is the logged value of research income awarded to university i in projects within the priority technologies of the Industrial Strategy during time t. Time period dummy intercepts, a t and regional NUTS-1 dummy intercepts, u j control for temporal and spatial fixed effects, respectively. x it captures a vector of unit and time variant explanatory variables, where Col represents the log of total income generated by university i at time t through collaborative research projects with external organizations; Cons is the log of the total value of consultancy research by university i at time t; Cont is Examining the roles of universities in place-based industrial strategy the total income generated through contract research income generated by university i at time t; IP represents the log of income generated from intellectual property by university i at time t; Pat represents the total number of patents possessed by university i at the end of time period t; Spinoffs represents total number of start-up businesses created by university i (excluding graduate start-ups) during time period t; CPD is the logged total value of all CPD courses run by university i during time period t; Fac is the log of the total value of income from renting of equipment and facilities by university i during time period t; Regen represents logged total income from regeneration projects, that is, those with a specific socio-economic remit, by university i at time t; z i is a vector of unit variant but time invariant variables including Inc, which represents the presence of an incubator in university i; SUB captures whether university i has a subsidiary organization for the commercialization of knowledge; BUS represents whether university i is primarily focused on business for its external engagement; and STRONG captures whether university i reports offering academics strong incentives for commercialization in terms of pecuniary rewards for business engagement activities. Composite errors (v it ) at time t in university i take the form of c i + u it , the sum of the unobserved individual heterogeneity (c i ) and idiosyncratic disturbances (u it ). The model was estimated using generalized least squares random effects (GLS-RE), with estimation based on a matrix-weighted average of a fixed effect which is generated by performing GLS on variables that have been multiplied by an idempotent matrix, transforming them into differences from their means. Further, a between-estimator is generated by performing GLS on variables that have been transformed into ones reflecting the difference between panel means and the variable means. We employed clustered robust standard errors (Whites heteroskedasticity-consistent estimator) to adjust for heteroskedasticity and first order autocorrelation.
Due to the spatial and temporal unevenness of university funding in the priority technologies (Johnston & Wells, 2020), a Wald chi squared test is also employed to test whether both year and regional coefficients are jointly equal to zero, respectively, which indicates that the inclusion of year and spatial fixed effects are required. Further applying the Lagrange multiplier test (Breusch &  , 1980) suggests that there are significant differences across universities, regions and years, thus pooled OLS estimations were not suitable. It is acknowledged that there are several other estimation approaches to our unobserved effects model, namely fixed effects (FE). Given that both estimators measure the difference between the ratio of the squared sum of residuals over time, summed over all panels, to the sum of the squared errors, any time-invariant explanatories are then removed from the equation. In addition, examining the correlation matrix showed no issues with collinearity among explanatory variables (see Table A1 in Appendix A in the supplemental data online). Examining the roles of universities in place-based industrial strategy Descriptive statistics Table 2 presents the descriptive statistics to provide an overview of the performance of the universities over the period. Between 2006/07 and 2016/17 each was awarded, on average, £16.6 million in funding for research projects focused on the priority technologies identified in the industrial strategy. The distribution of this funding was negatively skewed with 52 of the 149 universities receiving no funding for research in these areas. Income from third-mission activities is substantial. On average, each UK university generated over £394 million between 2006/07 and 2016/17 from these activities. Business research related activities such as contract and collaborative research are particularly lucrative, with average revenue for each university of £75 million and £152 million, respectively, over the period. Income from consultancy averaged nearly £62 million, while income from CPD, facilities and equipment provision, and regeneration activities averaged £34 million, £12 million and £34.8 million, respectively. Finally, IP income over the period averaged £8 million per university. In terms of commercialization activities, the average UK university had around eight start-ups registered at any one time between 2006/07 and 2016/17 and had 108 patents in its portfolio.

Pagan
In terms of strategy, Table 2 also shows that over twothirds of UK universities reported the existence of an incubator, and over half report that their strategic priority for external engagement is the business community. However, only one-fifth of UK universities report using strong incentives to encourage academics to commercialize the knowledge they create and fewer than half of UK universities (42%) have created a subsidiary organization to commercialize their research. Table 3 reports the estimates from the panel model with both temporal and spatial fixed effects included. Models 1-4 present different iterations of the model to examine robustness, with the full specification presented in model 5. The results from the panel models highlight the significant and positive effects of commercialization activities such as IP income, patents held and start-ups registered on research income in the priority technologies. Therefore, Hypothesis 1 is accepted as there is evidence that the entrepreneurial characteristics of a university have a positive influence on the generation of cutting-edge knowledge. Indeed, all entrepreneurial activities, that is, those focused on commercialization 'harder' factors are significant, providing justification for the focus on the entrepreneurial university.

EMPIRICAL RESULTS
In addition, the results highlight the significant and positive coefficients on contract research income and collaborative research income, suggesting that some characteristics of the engaged university are also important in the generation of cutting-edge knowledge. Therefore, Hypothesis 2 is partially accepted as not all engagement, or 'softer' activities are found to be significant.
However, while the significant relationships outlined above are positive, the coefficients suggest the relationships are in fact inelastic. For example, changes in both contract research and collaborative research income as well as number of start-ups bring about a less than proportionate change in research income. For example, where income from contract research is 10% higher, this equates to an additional 1.4% of research income from projects focused on the priority sectors. Based on an average level of income of £16.6 million over the period 2006/07-2016/17, this would result in extra revenue of £232,000. Furthermore, similar inelasticities are observed with respect to collaborative research income; where revenues from this source are 10% higher, an additional 2.9% of research income from projects focused on the priority sectors, or around £480,000, is received over the period. Finally, while IP income is on average a much smaller proportion of third mission income, its influence on cuttingedge knowledge creation is the largest. While still inelastic, the results show that where revenues from this activity are 10% higher, an additional 3.4% of research income from projects in the priority sectors is received, worth, on average, £564,000 over the period.
With respect to the effects of the variables designed to account for the strategic orientation of the university in relation to commercialization activities are not significant. Therefore, these policies do not appear to influence the ability of a university to generate knowledge in the priority technologies. Consequently, Hypothesis 3 is rejected.

DISCUSSION
The results presented here suggest that knowledge generation in the areas identified as a priority in the industrial strategy appears to rely on universities whose characteristics that are more akin to the model of the entrepreneurial university (Clark, 1998;Kirby et al., 2011). As such, greater levels of income from 'harder' third mission activities such as IP income, patents held, and the development of spinout firms (Perkmann et al., 2013) are all indicative of universities that produce cutting-edge knowledge.
However, the generation of cutting-edge knowledge is not driven by solely by entrepreneurial activities as 'softer' engagement activities such as collaborative research and contract research (Perkmann et al., 2013) are also found to have a positive influence. Therefore, we argue that both the entrepreneurial and engaged university models are useful for understanding the potential contribution of universities to the generation of cutting-edge knowledge in the context of the UK's Industrial Strategy. This lends further credence to arguments that the third mission activities of a university must be viewed as a spectrum rather than a dichotomy (Fuller et al., 2019); understanding universities as entrepreneurial, engaged, or otherwise does not provide an adequate understanding of the complexity of their activities. Therefore, with respect to industrial strategy formulation, a blended approach to university involvement is required, relying on those who fit both the entrepreneurial and engaged models (Sanchez- Barrioluengo & Benneworth, 2019).
Therefore, this evidence confirms that when it comes to generating cutting-edge knowledge, universities are not isomorphic (Kitagawa et al., 2016;Uyarra, 2010) and should not be treated as such by policymakers in the formulation of industrial strategies. In addition, the analysis presented here shifts the focus of policy formulation from understanding regional specializations (Johnston & Wells, 2020) to identifying universities with the entrepreneurial and engaged characteristics that signal they are generating the required knowledge.
The policy implications of this finding are highlighting that place-based industrial strategy should be more specific in its approach to identifying appropriate university partners, i.e., those that fit the entrepreneurial and engaged model. This also suggests that these policies should not  0.000*** 0.000*** 0.000*** 0.000*** 0.000*** Note: Dependent variable: log of total research funding in priority technologies in university (i) and time (t). *Significant at 10% level; **at 5% level; and ***at 1% level; cluster-robust standard errors adjusted for heteroskedasticity and AR(1) are reported in parentheses (Whites heteroscedasticity-consistent estimator); spatial (at NUTS-1 regional level) and temporal (year) fixed effects included but not reported; estimations via GLS random effects.
necessarily promote a reliance on 'local' knowledge, which may merely lock in spatial inequalities. Indeed, as Bailey et al. (2018) acknowledge, place-based Smart Specialisation policies are also driven by the 'relational embeddedness' of firms within networks, suggesting that active membership of broader complementary knowledge networks facilitated by trusted relations are the key to promoting economic development (Bruneel et al., 2017;Clarysse et al., 2014). Therefore, while effective Smart Specialisation builds on pre-existing knowledge within a region (Montresor & Quatraro, 2019), this may be best achieved through promoting collaborative partnerships with entrepreneurial and engaged universities regardless of their location. Consequently, we argue that place-based industrial strategies should consider promoting collaborations based on relational proximities (Balland, 2012;Balland et al., 2016;Bathelt & Gluckler, 2011;Ter Wal, 2014), that is, those built around the technological and organizational similarities between the firm and university partners (Chen & Xie, 2018;Gittelman, 2007;Johnston, 2020;Marrocu et al., 2013). This still allows regions to capture value and pursue Smart Specialisation Strategies that are tailored to regional strengths (Bailey et al., 2018), but crucially allows them to exploit global knowledge networks in order to avoid lock in (Bathelt & Cohendet, 2014). As such, institutionalized triple helix approaches to industrial policy can be designed to focus on the industrial specialisms of regional firms augmented with knowledge from the appropriate university partner, wherever they are located. Consequently, place-based industrial strategies can be designed to import knowledge from outside the region and ensure it is embedded into the regional ecosystem to add value to the regional economy (Bailey et al., 2015;Bailey et al., 2018).
Finally, the analysis presented here has shown that the strategic orientation of a university is not a significant factor in determining the generation of cutting-edge knowledge. Therefore, it may be that the institutional set-up of UK universities is somewhat homogenous in terms of offering similar incentives, having an explicit focus on business engagement, operating incubator facilities, and creating a subsidiary dedicated to technology transfer and licensing. The implication of this finding is that the observed differences in commercialization and engagement activities may result from the actions of individual academics rather than their institution's leadership (Abreu & Grinevich, 2013). Therefore, industrial strategy formulation should also consider the micro and meso levels of knowledge creation and transfer as well as the macro-level, contextual, level (Manniche & Testa, 2018).

CONCLUSIONS
The results presented in this paper highlight the fact that differences in the entrepreneurial and engagement activities among UK universities drive their ability to produce the cutting-edge knowledge. The results show that entrepreneurial activities such as licensing income, patenting, and creating spinouts are all positively related to cuttingedge knowledge generation. However, the only engagement activities with a positive relationship were collaborative and contract research. Therefore, as both entrepreneurial and engagement activities appear to be important, those universities that appear to be generating the required knowledge for innovation in sectors highlighted as a priority by the industrial strategy are those that combine these.
While an institutionalized triple helix approach is well motivated, particularly given the £2.4 billion of research funding in the priority areas awarded to UK universities, the pursuit of Smart Specialisation through a focus on these technologies should aim to pursue and embed knowledge from the most appropriate university sources. Importantly, we argue this may not necessarily be a local university, highlighting the need for policymakers to not only look at different spatial scales and develop placebased solutions (Organisation for Economic Co-operation and Development (OECD), 2017) but also to enable the importation and embedding of non-local knowledge. Therefore, pursuing Smart Specialisation to promote sustained regional development through capturing value from the national knowledge infrastructure requires that firms target the most appropriate universities as collaborative partners (Johnston & Huggins, 2021).
While our analysis has yielded significant insights at the institution level, one drawback is that we have not captured any of the micro-level factors which may influence these findings. For example, Abreu and Grinevich (2013) highlight the influence of individual characteristics such as age, position and gender of the academic in their propensity to engage in entrepreneurial activities. These may be particularly important given the lack of significance of strategic orientation. Furthermore, while we have identified those universities which may be sources of relevant knowledge, we cannot infer anything around the capability and willingness of academics within those institutions to work effectively with the industrial partners (Perkmann et al., 2011), nor the extent to which this means that industrial partners will judge the universities as credible partners (Johnston & Huggins, 2018). In addition, we call for further examination of potential path dependency of university knowledge generation, entrepreneurial, and engagement activities in order to examine how these may all evolve over time. Finally, we acknowledge that we have assumed that larger levels of income equate with higher levels of knowledge creation, but this does not necessarily capture its quality or impact. As such, we highlight the need for further work to examine these factors and how they may facilitate the implementation of institutionalized triple helix approaches to place-based industrial strategy.
supportive and helpful comments made during the review process. Their efforts have allowed us to produce a much better paper. As ever, all errors remain our own.

DISCLOSURE STATEMENT
No potential conflict of interest was reported by the authors.