Types of studies.

We did not screen studies by study design. We did, however, exclude commentaries and editorials, literature reviews that were not systematic reviews, conceptual and theoretical, and opinion pieces.

Type of interventions.

We included studies that described physical STEM assets (incubators, science parks, accelerators, research parks, laboratories, testing centres, and innovation centres) or services provided by STEM assets (business support, technology transfer, training programmes, testing, information, diagnostics, and networking events such as conferences). We also considered studies that described programmes delivered by universities that support: innovation (product/process); knowledge exchange; cluster development, supply chain development; technical qualifications/degree level apprenticeships; and graduate programmes and internships where there was a possibility that they might draw on a STEM asset. We included studies that reported on businesses, entrepreneurs, innovators, intermediaries, or professional and business services working with STEM assets. We excluded studies where the participants were social enterprises and further education providers–unless working with a university or STEM asset–given their lack of focus on research.

Types of outcome measures.

We included studies that provided information on at least one of the following outcome measures: innovation capability (process innovation, product innovation, or knowledge exchange) regional economic growth (start-ups, cluster development, or supply chains), business performance (turnover, productivity, profits), sustainability, inclusive growth, upskilling and employment, and environmental impact.

Geography.

We included studies from OECD member economies only (comprising at the time of writing, Australia, Austria, Belgium, Canada, Chile, Colombia, Costa Rica, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Israel, Italy, Japan, Korea, Latvia, Lithuania, Luxembourg, Mexico, Netherlands, New Zealand, Norway, Poland, Portugal, Slovak Republic, Slovenia, Spain, Sweden, Switzerland, Turkey, United Kingdom, United States). This decision was made so that studies included are from country contexts with comparable per capita incomes, as well as broadly democratic, capitalist governance structures and institutions.

Selection of studies.

The searches were undertaken by GB, CB and PY and exported into the Endnote database. Papers were double-screened on abstract and title and inclusion was based on consensus (AP, CB, DS, FP, KH, KK, PY, RL) with a third reviewer (CB, GB) adjudicating where necessary. At this stage, articles in languages other than English, French, German, Greek, Spanish, and Turkish were screened out, reflecting the language skills of the team. Articles that appeared to fulfil the inclusion criteria were retrieved. Full texts were screened and included texts were summarised in a standardised data extraction template. Data extraction was performed by a single reviewer and checked by a second reviewer (AP, CB, CI, DS, GB, FP, KH, KK, PY, RL) The data extraction template was designed by GB and based on the research questions, which were outlined by SC.

The searches identified 1,556 articles, and from these 162 studies were independently selected by two reviewers based on titles and abstracts (see Fig 1). 128 studies were excluded from the 162 identified. An overview of excluded studies is provided in S1 Table. The most common reasons for excluding studies were that they describe education or training programmes without reference to a STEM asset (n = 20), there was no physical STEM asset (n = 17), or the case study was based in a non-OECD country (n = 14). We were unable to obtain 25 articles through online, local or national resources available to us.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. PRISMA chart summarising selection of papers.

https://doi.org/10.1371/journal.pone.0287005.g001

Data extraction and analysis.

For each text, the reviewers independently undertook data extraction using a piloted data extraction form. A blank form can be found in S2 File. Information was collated on the type of STEM asset (university, accelerator, incubator, science park, research park, fixed Infrastructure, labs, testing centre, innovation centre, collaboration, research centre, or other), university involvement, level of STEM asset autonomy and relationship with universities, geographical focus, resources, activities and outputs, performance metrics, available datasets on the asset, timeframe covered by the article, policy lessons derived in the article, and gaps in evidence. Due to the anticipated heterogeneity of study designs and reported outcomes, the extracted data is presented as a narrative synthesis and it was not possible to undertake a systematic assessment of risk of bias.

Narrative synthesis.

The following analysis is organised by type of STEM asset (where assets have been grouped according to the similarity of their functions), in each case identifying key findings relating to their university relationship, their inputs (resources and activities undertaken), the adopted methodological approach of the paper, and any measured impact.

Incubators and accelerators.

Regarding the inputs for incubators and accelerators, it is possible to identify three broad groups of resources and activities which are apparent across the papers: facilities, funding, and support services [15–27]. Facilities refer to access to fixed infrastructure, such as office and manufacturing space, for businesses, graduates, or other entrepreneurs. Sources of funding vary, from local authorities to supranational authorities to private investments. Services typically include mentoring, business development workshops, skills training, academic expertise and knowledge acquisition, and networking, with the addition of curriculum content and extracurricular activities and experiences specific to graduates. Two of the articles focus specifically on pre-incubation [25,28], which has a much greater emphasis than incubators on activities that develop initial ideas into viable business ventures. Crisan et al identified three different levels of support provided by accelerators, ranging from limited to extended. They find that university accelerators usually offer limited support, with a focus on idea generation and validation [14].

Seven of the 14 incubator-focused studies [15,18,20,23,25–27], as well as one of the accelerator studies [29], were considered to have a high level of involvement with universities. Given that university business incubators are initiated by, and frequently located in or very close to, universities, this should not be surprising. However, in several instances, university engagement with the entrepreneurial team formation [23] and the operation of [25] their incubators were identified as having a significant role in shaping both incubator performance and the wider business ecosystem. Universities contribute to entrepreneurial teams in various ways, for example in board formation, or through softer interventions such as coaching and targeted educational offerings. Another study (although not focused on incubators) found that university-affiliated incubators worked with the most innovative firms which were thought to have the biggest growth potential [30].

The methodological approaches to the 16 studies were diverse, ranging from quantitative (n = 3), to qualitative (n = 6), to mixed method (n = 6) research strategies. The three purely quantitative articles differed in their specific approaches and research designs, which included statistical analysis informed by Structural Equation Modelling [28] and ANOVA analysis [22]. The quantitative elements of the mixed methods approaches included surveys [16], network analysis [19] and hypothesis testing using Pearson Chi-Square tests [23]. Among the qualitative and mixed method approaches, a case study design was frequently adopted [16–18,20,21,23,26,27,31], with interviews (typically semi-structured) being the most commonly used method [15,17,18,25,27,29,31]. However, these similar approaches were selected from differing theoretical frameworks, such as the absorptive capacity framework [27], a communications perspective [19], a more contextual Theory-based Impact Evaluation [17,18], and the Identity-Legitimacy-Life Cycle (ILLC) model [26].

Different methodological approaches adopted different measures for incubator and accelerator success. These ranged from the simple–such as employment and introduction of new products [29] or a relative count of new firms created across time [24]–to a wider range of 21 benchmark indicators [15]. As one study highlighted [25], the meaning of success to each individual incubator is likely to differ, ranging from the number or sustainability of firms established, to the quality of training and experience provided. Moreover, the IILC, which measures an incubator’s success by its ability to transition between its stages in development, necessitates differentiating between specific proxy measures used across time, transitioning from initial ad hoc monitoring to external standardised monitoring [26]. Follow-on funding secured by accelerated businesses may also be used as a performance indicator [14]. Thus, what is clear from the incubator and accelerator-related literature is that there is no one-size-fits-all measurement for evaluating participant success. Indeed, several different measures might potentially be required within the bounds of a single study, whether cross-sectional or longitudinal [25,26].

It is noteworthy that none of the studies focused their attention on measuring the broader impact of incubators and accelerators on the economy or society; they all focused on measuring the specific impact on the population of participating firms and incubatees. Positive commercial impacts were apparent across several studies, including increased firm survival rate, productivity, turnover growth, private investment, job creation, and new venture creation [17–19,21,22]. At the pre-incubation level, positive individual entrepreneurs’ perceptions of an incubator’s performance increased the desirability and confidence in starting and incubating a business [28], and another study evidenced incubatees’ continued improvement post-incubation [22]. One study–categorised as ‘university’ but which included analysis on incubators–found that incubation formed a condition for the most innovative young companies to achieve the highest growth rates [30]. A case study on the UBI Global-ranked top university-based tech incubator highlights the scale of the impact which such STEM assets can have, with the Digital Media Zone having helped over 400 companies, raised over $600 million, and created over 3,700 jobs since its inception [26]. The incubation literature also draws attention to non-commercial impacts such as providing professional development and learning opportunities for entrepreneurs, students, and university staff alike, raising the profile of the associated university and increasing connections with the wider community [20,26], as well as developing an entrepreneurial culture [14].

Surrogate entrepreneurship, whereby an experienced entrepreneur is added to a start-up’s senior leadership team (and the high level of university involvement it embodies) was identified as contributing to commercial benefits for incubatees [23]. But overall, the literature collectively explained the positive impacts in light of the inputs highlighted above: facilities, funding, and services involving human and/or social capital. Indeed, a lack of such inputs was identified by incubatees in one study as a constraint [25]. The social constructivist perspective of the Fisher ILLC model adopted in one paper explains incubator success as a function of its ability to gain differentially defined legitimacy from each new set of stakeholders it engaged with (and thus access to their respective resources needed to develop) [26].

University focused literature.

Much like the papers discussing university business incubators, the seven articles categorised as ‘university’ highlighted the role of facilities, funding, and services involving human and/or social capital as key input resources and activities [30,32–34], which when not present were identified as constraining factors to innovation [35]. However, in this context, the university’s role as a mediator [36] can be understood as adding a fourth category of inputs which refers to its ‘third mission’ activities centred on business innovation. This category gives a prominent role to knowledge as a resource itself [30] and as the focus of entrepreneurial activities involving its transfer and diffusion, such as patenting and commercialisation, encouraging academic spinout companies, and coordinating collaboration between various stakeholders [35]. Such activities are typically conducted by technology transfer offices/centres [13,35], often as part of larger technoparks [33].

This subsection of the literature, then, acknowledges universities as prominently located spaces to generate and disseminate STEM knowledge, leading to local, regional, and national impacts such as organisational learning, business innovation, job creation, and economic growth [32,33,35,36]. However, it is equally as cautious in assigning them too much importance for generating such impacts. Indeed, not all universities have developed an entrepreneurial culture [30], and other factors, such as the availability of skills, have been identified as being more significant for regional growth [34]. Moreover, only two of the seven papers noted a high level of involvement between universities and the other relevant stakeholders discussed [33,35].

Methodologically, the university-focused literature was split fairly evenly: quantitative (n = 3), qualitative (n = 2), and mixed methods [13]. The two more simplistic quantitative designs involved administering surveys (one cross-sectional, one longitudinal) to businesses to understand their relationship with, and perceptions of, university activities [34,36]. The more advanced quantitative analytical techniques employed regression analysis, to identify contributing factors to spinoff generation as a way of measuring spillover [32]; fuzzy cognitive map modelling, to identify causal determinants of the success of technology transfer offices (TTOs) [13]; and crisp-set qualitative comparative analysis, to look for relationships between various support infrastructures and the growth of young innovative companies [30]. The latter used the change in employment level as a proxy for company growth, whilst the fuzzy cognitive map modelling measured TTO performance using the following proxies: number of spin-offs established, number of patents awarded, license income, income from industry-sponsored research contracts, and consulting income. One of the qualitative studies (both of which were case studies) also sought to understand the success of TTOs in promoting technology transfer but did so using semi-structured interviews with managers [33].

Innovation centres, labs, research centres and research parks.

The remaining STEM assets are referred to as a group in this final section given i) the smaller number of papers which concern them (n = 10) and ii) that they share similar functions, as well as strong relationships with universities. It is notable that seven of these studies were categorised as having a high level of university involvement with the respective STEM asset, for which the two main reasons were specific location and their functional positionality as mediators between industrial firms and academia. Regarding the former, assets were typically based at, or within proximity to, the main university campus [37–40], whether they operated within a specific department [39,40] or almost independently of university management [41].

Regarding the specific inputs and activities undertaken in their mediating role, the STEM assets provide both concrete facilities and physical equipment, and a favourable environment for innovative collaboration between entrepreneurial firms as well as academic knowledge transfer to such firms [37,38,40–42]. Key to these innovation-related activities is hybrid researchers who work in the space directly between academia and industry [41], helping to close the gap which has typically been referred to as the ‘valley of death’. Not only is their specific academic knowledge particularly valued, but so too are the other components of researchers’ human capital, such as professional competencies, awareness of institutional culture, and accessible networks [37,41]. In one particular case, collaborative and interdisciplinary R&D, whilst supported by staff, is directed by university students, who both gain valuable skills and experience whilst contributing to highly sought-after innovative solutions to real industrial problems [39]. The environment and facilities provided by the STEM assets in another case also led to the development of linkages between community college students and local businesses through activities such as training, mentoring, and networking [37].

As well as knowledge transfer and innovation, assets frequently hold specific training programmes [41,43] and provide other services such as production process improvement and product testing [41]. One innovation centre provided assessments of business practices as well as technology clinics to assess needs and identify solutions to common issues [40]. In the case of research parks, supporting services also included food and drink, real estate and maintenance, which are provided by secondary organisations based on the site, making it an effective ‘micro cluster’ [38].

Funding for such assets is provided by governments at the regional, national, and supranational scale [41] as well as governmental agencies, membership fees from firms, donations, and spending by participants at in-person events and workshops [43]. However, the funding structure was also identified as an area of strain and even conflict with universities, as overhead resources for centre management, were difficult to source, and time for industry engagement was in conflict with academics’ other responsibilities, in particular developing academic publications [40,44].

The research approaches adopted by these ten studies were fairly evenly split. The two papers relating to labs [37,39] both opted for qualitative case study designs. Three studies on innovation centres adopted qualitative approaches, conducting semi-structured interviews with individuals working at the centres [40,41,44]. One of these researched an interdisciplinary approach between STEM and social sciences. In this case, the goal of the innovation centre was to simultaneously improve productivity as well as employment quality at supported businesses, and the paper documented the implementation from management and employee perspectives using repeated semi-structured interviews [44]. In contrast, another paper concerning innovation centres used a quantitative approach involving survey data and multiple linear regression techniques to better understand the factors affecting company turnover growth [45]. The remaining four papers, relating to research centres and research parks, all used quantitative methods. One of these also looked at identifying the factors, such as the number of patents and finance, which best predict the growth of companies based at research parks, but used the Mann-Whitney statistic method [42]. The final three articles sought to measure the wider impact of their respective STEM assets, but the impact was measured differently in each case. The research park study estimated its indirect and induced employment, differentiating between ‘core’ and ‘supporting’ firms based on site [38], whilst the research centre studies tried to estimate its regional and national economic impacts [43,46]. In one case, a proxy measure of regional economic impact was estimated by combining direct and indirect centre expenditures (generated using a regional input-output model) with many other additional impacts including income relating to licensing and intellectual property, taxable income from start-up firms which originated at the centre, and the cost savings for firms who hired graduates of the centre. To estimate the national economic impact, the regional proxy was complemented with a consumer surplus model to accommodate for spillover effects from which the public benefits [43]. Finally, one paper used administrative data and a double-matching approach to identify two control groups and causally estimate regional spillover effects from the research centre. While they found a positive short-term effect on employment, particularly employment of managers, there were no significant long-term effects of the STEM asset discernible on the regional economy [46].

Turning directly to the evaluation of the impacts generated by innovation centres, labs, research centres and research parks, several insightful points arise from these studies. Firstly, whilst the activities of these STEM assets and their positionality vis-à-vis academia and industry outlined above seemed suggestive of a direct relationship between existence and commercial impact, the effect is somewhat more nuanced, especially regarding innovation centres. Indeed, both studies conclude that interactions between the centres and firms alone are insufficient to explain economic impacts such as growth and increased employment. However, when combined with other factors–such as public financial support and other support infrastructures–such interactions become statistically significant [41,45].

Secondly, quantitative studies, whilst beneficial, are also problematic in several ways. One research centre study highlighted the problem of time lag, with the uptake and application of R&D and tangible indicators of other economic impacts taking a number of years to evidence before they could be measured and assessed [43]. Moreover, and of particular pertinence to this systematic review, the same paper also concluded that only using quantifiable measures considerably underestimates the scale of impact that STEM assets generate. Indeed, benefits can be unintended and unanticipated, such as providing local employers and community college students the space for training, as one of the lab studies documents [37]. Furthermore, quantitative studies are less able to effectively measure certain benefits such as improved reputation [37,42].

This links with the final point that the literature calls to attention: the non-commercial impacts generated by STEM assets. Not only can STEM assets positively affect regional and national employment [38,43], but society can benefit in a multitude of ways from the practical application of innovations in a wide breadth of sectors, from health care to education, to the environment [39]. Furthermore, interdisciplinary approaches that bring together social scientists, natural scientists and engineers may produce social impacts, such as improved job quality [44]. Similarly, multiple stakeholders can benefit from the activities which assets undertake, from the experience, skills and training developed by students and young career researchers in the short term to the businesses and firms who come to benefit from such human capital in the long term [37,43].