University Graduates and New Green-Tech-Based Entrepreneurship: Evidence from Italian Regions

Abstract

Universities serve as catalysts for knowledge creation across territories, promoting innovation and economic development through different channels. This paper investigates the role of university graduates as a location determinant of new green-tech-based firms (NGTBFs) across Italian NUTS-3 regions over the period 2011–2017. We examine whether universities, as providers of high-skilled human capital, affect the spatial distribution of new green ventures. Adopting a patent-based definition of NGTBFs and an econometric framework accounting for regional heterogeneity, we analyse the impact of university graduates on green firm creation. The results show that higher education fosters green entrepreneurship primarily through the channel of producing doctoral and STEM-oriented graduates, who serve as key drivers of NGTBF formation. Interestingly, the analysis reveals marked spatial heterogeneity across Italy’s North–South divide, with stronger associations of PhD and STEM graduates in Southern regions, where specialised human capital appears to compensate for weaker innovation systems. These findings deliver clear policy implications, suggesting that strategies aimed at promoting green entrepreneurship should prioritise advanced, STEM-oriented human capital and explicitly account for regional contexts, rather than relying on uniform higher education expansion approaches.

1. Introduction

The green transition is no longer a matter of environmental preference but a defining global economic and societal imperative to preserve the environmental integrity of territories. At its core, it aims to decouple economic growth from ecological degradation by transforming energy systems, production processes, and consumption patterns towards sustainability (OECD, 2011; United Nations, 2015). Within the European context, this transformation has been institutionalised through ambitious policy frameworks such as the European Green Deal, which sets out the European Union’s roadmap for achieving climate neutrality by 2050 and accelerating decarbonisation across all sectors of the economy (European Commission, 2019). Complementary industrial policy initiatives, such as the Net-Zero Industry Act, further aim to strengthen Europe’s capacity to manufacture strategic clean technologies and support the deployment of net-zero solutions across the EU (European Union, 2024).

Beyond their economic and technological implications, these policy initiatives highlight how the green transition is increasingly embedded in a broader societal transformation that places education and skills development at its core. The shift towards sustainability requires not only technological innovation but also the formation of new competencies, knowledge bases, and learning processes capable of supporting systemic change across sectors and territories (Wiek et al., 2011).

Such a transformation is inherently systemic, demanding coordinated action across multiple layers of society and governance. Its success depends on the collective engagement of diverse institutional stakeholders, including governments, international regulatory bodies, large corporations, SMEs, financial actors, research institutions, and civil society. Within this multi-level framework, education systems—and higher education in particular—play a pivotal role in shaping the capabilities that enable actors to participate effectively in the transition (Brundiers et al., 2010; Coll-Martínez et al., 2022).

In this context, new green entrepreneurship emerges as a key mechanism through which sustainability-oriented knowledge and skills are translated into tangible technological and organisational change (Dean & McMullen, 2007). Green entrepreneurial activity, centred on discovering and exploiting opportunities to introduce innovations with positive environmental impact, is inherently knowledge-intensive and relies on the availability of advanced competencies and specialised expertise (Gast et al., 2017). Understanding the factors that foster or hinder its geographical diffusion is therefore a strategic priority, as the spatial patterns of green venture creation determine where new green technologies take root, scale, and ultimately reshape regional economic trajectories.

Among the various determinants shaping this process, knowledge—and the human capital through which it is embodied—stands out as a critical driver (Coll-Martínez et al., 2022). Access to advanced education, technological capabilities, and highly skilled individuals fundamentally conditions the ability of regions to generate novel green solutions and support the emergence of eco-innovative firms (Orsatti et al., 2024; Quatraro & Scandura, 2024). In this sense, the green transition can be understood not only as a technological or economic transformation but also as an educational challenge, requiring the development of new forms of expertise, interdisciplinary knowledge, and innovation-oriented mindsets.

Universities occupy a central position within this framework. Traditionally recognised as engines of knowledge creation, they are increasingly evolving into key actors in the formation of green entrepreneurial ecosystems (Acosta et al., 2011; Bonaccorsi et al., 2014). The literature shows that universities contribute to entrepreneurship mainly through three interrelated mechanisms (Bertoletti & Lelli, 2026), which can equally be applied to the context of green entrepreneurship. The first is human capital formation: by training graduates equipped with scientific, technical, and innovation-oriented competencies, universities directly expand the pool of individuals capable of identifying and exploiting green entrepreneurial opportunities—particularly in technology-intensive domains such as renewable energy, clean manufacturing, or circular economy solutions, where translating scientific knowledge into viable ventures requires deep disciplinary expertise (Gast et al., 2017; Wiek et al., 2011). The second mechanism operates through knowledge spillovers. In particular, universities generate codified and tacit knowledge through research activities that, when absorbed by local actors, can seed new firm formation (Acs et al., 2009). The third mechanism concerns the universities’ third mission (Martin, 2012). Through technology transfer offices, spin-off programmes, incubators, and collaborative research partnerships with industry, universities foster the connectivity between academic knowledge and market actors that is essential for green ventures to scale (Agasisti & Bertoletti, 2019; Bonaccorsi et al., 2014), While these three mechanisms are complementary and likely to interact, this paper focuses primarily on the human capital formation channel. On the one hand, it represents the most direct and proximate link between university activity and entrepreneurial outcomes, as it is through graduates that universities most systematically transfer the competencies required to found and develop new ventures. On the other hand, it remains the least explored channel in the context of green entrepreneurship. Despite this relevance, the existing literature has primarily focused on other determinants of sustainable entrepreneurship, including regional technological specialisation (Corradini, 2019), the presence of supportive stakeholders such as policymakers and investors (Cojoianu et al., 2020; Giudici et al., 2019), and the role of green demand (Colombelli et al., 2025; Lelli et al., 2025). By contrast, the role of universities as producers of human capital shaping the competencies underlying green entrepreneurship remains underexplored—a gap that is particularly relevant given the growing recognition that the success of the green transition depends not only on innovation systems, but also on the capacity of education systems to generate the skills and knowledge required to sustain it. Against this background, this paper investigates whether and how university graduates contribute to the creation of new environmentally oriented ventures across Italian NUTS-3 regions, formally asking:

• RQ: to what extent and through which mechanisms do universities’ graduates influence the emergence and geographical diffusion of new green entrepreneurship across regions?

The paper makes three contributions. First, it links the emergence of new green-tech-based firms (NGTBFs) to regional models of human capital and knowledge creation, integrating the green entrepreneurship literature with frameworks developed around knowledge spillovers and university-driven innovation (Acs et al., 2009; Bonaccorsi et al., 2013). Second, it introduces a granular perspective on university human capital, distinguishing between total and STEM graduates and incorporating a measure of PhD intensity, while also considering university-level characteristics such as institutional prestige and governance structure (public versus private). Third, it explicitly accounts for territorial disparities within Italy, particularly the North–South divide, providing a more nuanced understanding of how university-generated human capital interacts with local contexts to shape the geography of green-tech-based start-ups (Crescenzi et al., 2007). Focusing on Italy provides a particularly relevant empirical setting. The country has witnessed a steady expansion of environmentally oriented entrepreneurship, supported by a national policy framework introduced in 2012 to promote innovative start-ups through fiscal incentives, simplified regulatory procedures, and targeted support for R&D. At the same time, Italy is characterised by pronounced regional disparities in economic development, innovation capacity, and the spatial distribution of research institutions—making it a valuable laboratory for examining how local knowledge resources and human capital contribute to the geographical emergence of green technology-based entrepreneurship. The empirical analysis covers Italian NUTS-3 regions over the period 2012–2017, employing Poisson regression models to account for the count nature of new firm formation, with additional specifications exploring territorial heterogeneity. The results indicate that the availability of university graduates—particularly those with scientific and research-oriented backgrounds—significantly contributes to the regional formation of green technology-based start-ups, highlighting the role of higher education as a key lever for promoting sustainable entrepreneurship and supporting the green transition. The remainder of the paper is structured as follows. Section 2 outlines the empirical approach and describes the data and methodology. Section 3 presents the empirical results. Section 4 concludes and discusses the policy implications.

2. Background Literature

2.1. The Role of Universities as Location Determinants of Entrepreneurship

The study situates itself between the literature on entrepreneurship and regional innovation, drawing on the Knowledge Spillover Theory of Entrepreneurship (KSTE), which conceptualises entrepreneurial activity as a mechanism through which new knowledge generated within institutions—particularly universities—is transformed into commercial opportunities. This framework provides a useful lens for understanding how academic knowledge and human capital can stimulate the emergence and spatial diffusion of green entrepreneurial ventures. A substantial body of research highlights the central role of universities in fostering entrepreneurship by generating and disseminating knowledge within regional economies. Building on the Knowledge Spillover Theory of Entrepreneurship (KSTE), entrepreneurial opportunities are understood to arise from new knowledge created and circulated among economic actors (D. B. Audretsch et al., 2006). This perspective emphasises that knowledge exchanges within territorial systems stimulate the development of new technologies and ventures (Acs et al., 2009; D. B. Audretsch & Lehmann, 2005). The framework has since evolved into the Knowledge Spillover Theory of Innovation (KSTE&I), which argues that opportunities emerge when knowledge generated, but not fully commercialised, by incumbent institutions becomes accessible for entrepreneurs to transform into marketable innovations (D. Audretsch et al., 2025). Yet the capacity to exploit such knowledge is uneven across regions. While some areas sustain long-term economic dynamism through continuous technological advancement, others with limited innovative capacity struggle to initiate new technological trajectories (Sbardella et al., 2018). Among the institutions capable of shaping these regional knowledge environments, universities hold a prominent position. Entrepreneurial universities, in particular, function as catalysts for knowledge creation and diffusion, contributing to economic and social development through their teaching, research, and entrepreneurial missions (D. Audretsch & Belitski, 2022; Guerrero et al., 2015). By producing skilled human capital, advancing scientific knowledge, and promoting the formation of spin-offs and new ventures, universities significantly enhance regional innovation capacity and entrepreneurial dynamics (Cunningham et al., 2017; Cunningham & Menter, 2021; Guerrero et al., 2015). Empirical evidence consistently shows that universities shape the locational patterns of new technology-based firms. Early research on German firms demonstrated that location choices respond not only to the quantity of university output but also to the specialisation of the scientific knowledge produced locally (D. B. Audretsch & Keilbach, 2004). Regions hosting universities with higher knowledge capacity and stronger research performance tend to generate disproportionately more technology startups, reflecting the role of universities as sources of localised knowledge spillover (D. B. Audretsch & Lehmann, 2005). Technology-based firms often cluster in proximity to universities in order to access tacit knowledge, specialised expertise, and collaborative networks that reduce the costs of innovation and technological development. Beyond knowledge availability, the characteristics and quality of universities further shape their impact on entrepreneurial activity (Bertoletti & Lelli, 2026), a point further supported by the growing literature on the multidimensional measurement of higher education systems’ performance (Nepomuceno et al., 2024). Research-intensive universities, especially those with strong applied research profiles, tend to be more efficient in transferring knowledge to external commercial actors (Bertoletti & Johnes, 2021). Evidence from the United Kingdom shows that elite research-intensive universities, such as the Russell Group, primarily influence regional economies through the creation of spin-offs, whereas other institutions contribute more through knowledge transfer activities (Guerrero et al., 2015). At the same time, cross-country evidence highlights that the channels through which universities affect entrepreneurial dynamics vary across institutional and regional contexts. In Spain, for instance, the supply of graduates rather than research output emerges as the most important driver of high-tech firm formation (Acosta et al., 2011). In Italy, universities specialising in applied sciences and engineering display a strong positive association with new firm creation, particularly in service-based sectors (Bonaccorsi et al., 2013). Additional evidence from the United Kingdom indicates that high-quality research universities contribute to the spatial clustering of research activity and commercial enterprise in their vicinity (Abramovsky et al., 2007), reinforcing the importance of geographic proximity in facilitating knowledge spillovers. While the literature highlights the critical role of universities in fostering technology-based entrepreneurship, the emerging field of green entrepreneurship has paid comparatively limited attention to these institutional dynamics. Existing studies on the regional determinants of green entrepreneurial activity emphasise factors such as technological specialisation in environmental technologies (Corradini, 2019), knowledge bases and innovation capabilities (Coll-Martínez et al., 2022; Colombelli & Quatraro, 2019), or demand-side drivers and policy incentives (Lelli et al., 2025). Only a few studies highlight the role of academic inventors in the development of green technologies (Orsatti et al., 2024; Quatraro & Scandura, 2019, 2024).

Although these contributions recognise the importance of university-generated knowledge in fostering green start-ups, they leave a significant gap in understanding the role of universities as producers of highly qualified human capital.

2.2. University Human Capital and New Green Entrepreneurship

Human capital represents a fundamental channel through which higher education systems shape economic development, innovation processes, and entrepreneurial dynamics (Becker, 1964; Schultz, 1961). Within this framework, higher education contributes to the accumulation of knowledge, skills, and competencies that underpin technological advancement and long-term economic growth (Castelló-Climent & Hidalgo-Cabrillana, 2012; Sianesi & Van Reenen, 2003). From an educational perspective, human capital should not be interpreted merely as a quantitative output of higher education systems but rather as a multidimensional construct encompassing skills, competencies, and cognitive capabilities acquired through formal learning processes (Marginson, 2019). While empirical research has often relied on simplified proxies—such as the number of graduates—this approach neglects important differences in the composition and quality of human capital. As a consequence, it may lead to incomplete or biassed assessments of the contribution of universities to economic and entrepreneurial outcomes (Hanushek & Woessmann, 2010).

These effects are not evenly distributed across space but are strongly mediated by the geographical organisation of higher education systems. Regions with greater access to universities not only exhibit higher participation rates in tertiary education among local populations but also tend to retain a significant share of graduates within their local labour markets (Card, 2001; Valero & Van Reenen, 2019). Through these mechanisms, universities contribute to the formation of localised stocks of human capital that are embedded within regional economies and support local innovation dynamics.

Moreover, the impact of human capital extends beyond local boundaries. Within regional development frameworks, knowledge and skills generated by higher education institutions may diffuse across neighbouring areas through spillover mechanisms (Faggian et al., 2019). This implies that university-generated human capital operates both as a local asset and as a source of interregional externalities, thereby reinforcing its relevance as a driver of regional development processes (Agasisti & Bertoletti, 2022).

Building on this understanding, human capital also constitutes a key mechanism linking education to entrepreneurial activity. However, entrepreneurship is shaped by the interaction of individual, organisational, and contextual factors (Bergmann et al., 2016). At the individual level, human capital plays a central role in enabling both the recognition of entrepreneurial opportunities and their transformation into new entries. Empirical evidence confirms that, whilst individual characteristics remain the primary determinant of entrepreneurial propensity, university environments and regional conditions significantly influence how human capital is formed and mobilised for entrepreneurial purposes (Bergmann et al., 2016).

This link becomes particularly salient in the context of the green transition. Green entrepreneurial activity is inherently knowledge-intensive and relies heavily on specialised, science-based, and sustainability-oriented competencies (Colombelli & Quatraro, 2019; Gast et al., 2017). In this sense, the composition and quality of human capital—rather than its mere quantity—become critical in shaping the emergence of green ventures.

However, despite a growing body of literature recognising human capital as a potential antecedent of green entrepreneurship (Coll-Martínez et al., 2022; Colombelli & Quatraro, 2019; Lelli et al., 2025), existing contributions have not explicitly examined the role of university-generated human capital in this process. Notably, Coll-Martínez et al. (2022) highlight the importance of knowledge embedded in university graduates for the development of green technology-based entrepreneurship, thereby pointing to a clear gap in the literature and motivating a more systematic and detailed investigation.

2.3. Conceptual Framework and Research Hypothesis

Building on the arguments presented in the previous section, this paper develops a conceptual framework that identifies universities and, in particular, their capacity to generate differentiated forms of human capital as a key location determinant for the emergence of new green technology-based firms. The underlying mechanism is that universities, by training and supplying graduates to the regional labour market, expand the local stock of human capital available to support or directly engage in entrepreneurial activity. When sufficiently developed, this stock may reduce barriers to firm creation by enhancing the cognitive capacity, technical competence, and problem-solving abilities required to identify and exploit entrepreneurial opportunities.

However, not all forms of human capital contribute equally to green entrepreneurial dynamics. The overall availability of higher education graduates represents a baseline condition reflecting the general level of skills and knowledge within a region, potentially facilitating entrepreneurial activity by increasing the supply of individuals with general analytical and organisational capabilities (Marvel et al., 2016).

More importantly, the disciplinary composition of human capital is expected to play a decisive role (Hanushek & Woessmann, 2010). A regional graduate base with a strong STEM orientation is likely to exert a stronger impact on the emergence of new green technology-based firms, as green entrepreneurship is inherently science- and technology-intensive and relies on specialised knowledge and engineering capabilities (Colombelli & Quatraro, 2019; Fritsch & Aamoucke, 2013).

In addition, the framework explicitly considers the role of advanced, research-oriented human capital. Doctoral graduates (ISCED8) represent a distinct component characterised by high levels of scientific expertise and research capability. Their involvement in knowledge creation and frontier research makes them particularly relevant for green entrepreneurship, which often depends on emerging and complex technological domains (Quatraro & Scandura, 2019).

Finally, the framework acknowledges the geographical dimension of these mechanisms. The impact of university-generated human capital is unlikely to be uniform across regions, as it interacts with local conditions such as industrial structure, institutional quality, and innovation capacity. Although the direction of such geographical variation remains theoretically indeterminate at this stage—and is therefore treated as an open empirical question—disentangling this spatial dimension is an integral part of the framework, as it reflects the inherently localised nature of both knowledge production and entrepreneurial processes. Within this perspective, it is plausible that in regions characterised by weaker innovation systems, specialised forms of human capital may play a compensatory role, generating stronger marginal effects on the creation of new green firms (Crescenzi et al., 2007).

Based on these considerations, the following hypotheses are proposed:

H1. 

The overall stock of university graduates is positively associated with the regional emergence of new green technology-based firms.

H2. 

The share of STEM-oriented graduates has a positive and stronger effect on the emergence of new green technology-based firms compared to general graduates.

H3. 

The presence of doctoral (PhD) graduates has a positive and significant effect on the emergence of new green technology-based firms, reflecting the role of research-oriented human capital.

H4. 

The effects of university-generated human capital on green entrepreneurship vary across regions. While the direction of this variation remains theoretically indeterminate, university-generated human capital may exert stronger effects in regions characterised by weaker innovation systems, where it can compensate for the absence of complementary institutional and economic resources.

3. Data and Methodology

3.1. Data Sources

This study relies on a novel dataset of New Green Technological-Based firms, encompassing Italian NUTS-3 regions between 2011 and 2017, integrating information from several complementary databases. The EPO PATSTAT database serves as the primary source for detecting NGTBFs and mapping their presence throughout EU regions. Our methodology is consistent with contemporary research contributions (see Lelli et al. (2025)) while providing fresh perspectives on identification techniques that diverge from the conventional approach of matching firm-level records with green patent portfolios, as demonstrated in previous studies (Coll-Martínez et al., 2022; Colombelli & Quatraro, 2019; Corradini, 2019; Giudici et al., 2019). To classify green patents, the study adopts standard procedures found in the scholarly literature (e.g., Cicerone et al. (2023)), specifically employing the Cooperative Patent Classification (CPC) framework and concentrating on technologies addressing climate change mitigation, identified through Y02 codes. Beyond patent classification, the database encompasses granular details about patent applicants, including their status (corporate or individual), spatial coordinates at various NUTS levels, and prior inventive output. The European Tertiary Education Register (ETER) constitutes the second key data source, offering insights into university attributes including institutional classification, disciplinary distribution of graduates, student numbers, and organisational form. These characteristics facilitate our examination of universities’ contribution to the emergence of New Green Technology Enterprises. The analysis restricts attention to institutions designated as “university” in ETER1, while excluding art institutes, research institutes, military schools, and conservatories, given their institutional diversity. Similarly, virtual universities are not included owing to the absence of physical infrastructure. When institutions operate multiple campuses, the study focuses exclusively on the principal location as specified within ETER, as the dataset provides institutional-level data without distinguishing between campuses. This approach is consistent with the strategy commonly adopted in the literature (Agasisti & Bertoletti, 2022; Bertoletti et al., 2022). Moreover, only a limited number of universities operate campuses in other regions, and these universities are generally smaller and less populated, implying a marginal impact on overall institutional performance. Finally, a set of control variables is constructed using Eurostat data in order to account for regional socio-economic conditions that may influence the analysed phenomena. These include indicators such as per capita GDP, population concentration, and labour force participation.

All variables are aggregated to the NUTS-3 regional level, which constitutes the unit of analysis throughout the study. Specifically, NGTBFs are counted by region based on the spatial coordinates of patent applicants as recorded in PATSTAT. University-level characteristics and graduate flows from ETER are similarly aggregated to the NUTS-3 level by summing or averaging across institutions located within each region. In particular, for variables reflecting institutional characteristics—namely foundation year and the share of public universities—regional values are computed as weighted averages across universities within each region, using the number of enrolled students as weights, following the approach adopted by Bertoletti et al. (2022). Regional socio-economic controls from Eurostat are directly available at the NUTS-2 or NUTS-3 level. The resulting dataset is therefore a region-year panel, in which each observation corresponds to a NUTS-3 region in a given year, and all variables reflect regional aggregates rather than firm- or institution-level records.

3.2. Dependent Variable: NGTBFs

The study’s outcome variable measures the count of New Green Technology-Based Firms located in EU NUTS-3 regions spanning the period 2011 to 2017. We employ a more restrictive conceptualisation of NGTBFs, consistent with Lelli et al. (2025), that differs from alternative definitions used elsewhere in the literature (Coll-Martínez et al., 2022; Colombelli & Quatraro, 2019; Corradini, 2019). Under this stricter formulation, NGTBFs are characterised as enterprises whose inaugural patent filing constitutes a green patent and whose total patenting activity comprises at least 50% green patents. The operationalisation of this definition involved three sequential phases. Initially, green patents were identified by reference to Climate Change Mitigation Technology designations (Y02 codes) within the Cooperative Patent Classification (CPC) system, with retention limited to applications containing at least one such code. Subsequently, green enterprises were designated as those having at least half their patent filings classified within these green technology categories. The final stage involved identifying new green technology-based firms by drawing on the framework established by Lelli et al. (2025), specifically targeting firms where their first green patent application aligns temporally with their initial patent submission overall. The resulting dataset captures characteristics of Italian new green-tech-based firms disaggregated at the NUTS3 regional scale. It warrants emphasis that because firm establishment dates are unavailable in our data, we cannot definitively confirm whether all identified firms qualify as new according to demographic criteria. Rather, by focusing on enterprises whose inaugural green patent matches their overall first patent filing, we identify actors whose inventive trajectory begins with green technological orientation, namely, new green technological-based firms. In line with contemporary research practices, we presume that firm founding approximates the timing of initial patent registration, a presumption substantiated by (Ewens & Marx, 2024), whose analysis indicates that approximately 90% of U.S. firms established post-1975 were no more than ten years in operation at the point of their first patent application. As already recognised in the relevant literature (Coll-Martínez et al., 2022; Colombelli & Quatraro, 2019; Corradini, 2019; Lelli et al., 2025), the distribution of new green-tech-based firms is geographically uneven, and it might vary across different time periods. The distribution of NGTBFs over the year and macro-area (North/South)2 is presented in Figure 1. For the purpose of the analysis, we adopt a simplified territorial classification distinguishing between North and South. In this framework, the North includes both Northern and Central Italian regions, while the South refers to the Mezzogiorno. This choice reflects the well-documented structural divide in the Italian economy, where Southern regions typically display lower levels of higher education and weaker entrepreneurial ecosystems compared with the rest of the country (Abramo et al., 2016). Consequently, throughout the paper, references to the North should be interpreted as including both Northern and Central regions.

Figure 1 illustrates the temporal evolution of average NGTBFs across Italian macro-regions from 2011 to 2017, revealing a persistent and substantial North–South gap in green technology entrepreneurship. Throughout the entire observation period, Northern regions consistently maintain higher average levels of NGTBF formation. In contrast, Southern regions exhibit markedly lower levels, demonstrating that the North generates roughly three to four times more green technology-based firms than the South on average. This stark disparity underscores the deeply entrenched territorial inequalities that characterise Italy’s innovation landscape and confirms that green entrepreneurship follows broader patterns of regional economic divergence. Notably, the gap between the North and South remains remarkably stable over time despite the European debt crisis and subsequent economic stagnation, which particularly affected investment in innovative and capital-intensive ventures during the mid-2010s, suggesting that the relative disadvantage of Southern regions in generating green technology entrepreneurship is deeply rooted in structural differences in human capital, innovation infrastructure, and economic development rather than being merely a temporary or cyclical phenomenon.

3.3. Independent Variables and Controls

This analysis examines how universities’ production of human capital influences the emergence of new green technology-based firms. The research employs three primary independent variables to capture different dimensions of university human capital output: (i) the total number of graduates at ISCED levels 5 and 7 (bachelor’s and master’s degrees); (ii) the total number of graduates at ISCED level 8 (doctoral degrees); and (iii) the STEM orientation of graduates from local universities.

The first variable captures the overall availability of highly educated human capital within a region and represents a standard measure in the literature examining the relationship between universities and regional innovation. A large body of research highlights that the production of university graduates constitutes one of the main channels through which universities contribute to local knowledge creation and entrepreneurial dynamics (see, for instance, Acosta et al. (2011); Bonaccorsi et al. (2013)). In this sense, the stock of graduates represents the baseline indicator of university-generated human capital that may support the emergence of new knowledge-based and technology-oriented firms.

The second variable focuses on doctoral graduates, who represent a more research-intensive component of university human capital. PhD holders typically possess advanced research capabilities and are more directly involved in the creation and transfer of scientific knowledge (Pablo-Hernando, 2015). Their presence may therefore strengthen the link between universities and the formation of knowledge-intensive firms, particularly in technologically advanced sectors such as green technologies.

STEM skills are widely recognised as key inputs for innovation and technological development (Fritsch & Aamoucke, 2013). A higher concentration of STEM graduates may therefore increase the likelihood of new firm formation in technology-intensive sectors, including environmentally oriented industries that rely on scientific and engineering competencies for the development of green technologies.

Together, these variables allow us to assess how different dimensions of university-generated human capital—general higher education, research-oriented training, and technological specialisation—contribute to the regional emergence of green technology-based entrepreneurship.

Figure 2 illustrates a pronounced regional disparity in graduate numbers across the period examined. The North consistently demonstrates the highest graduate output, maintaining levels between approximately 3400 and 3900, albeit with a notable decline around 2016 followed by a substantial recovery. Conversely, the North exhibits the lowest figures, remaining relatively stable at approximately 2600–2800 throughout most of the timeframe, with a marginal increase evident towards 2017. Notably, despite some temporal fluctuations, the hierarchical relationship between them remains consistent throughout the period, suggesting the persistence of underlying structural factors that influence differential graduate production between the North and South.

Figure 3 shows the trends of the graduates at the ISCED8 level over time, confirming the previous one. The figure shows a persistent and widening territorial divide in the average number of ISCED 8 graduates between Northern and Southern Italy over the period considered. Altogether, the patterns suggest an increase in spatial polarisation in the formation of advanced human capital in different areas. Regarding the STEM orientation, measured as the share of students and graduates in science and technology, no particular differences occur between North and South of Italy, which obtain results of 5.92% and 5.97%, respectively.

The analysis also incorporates a set of controls at the university level, including the university’s founding year and its financial status as a public institution, and whether it is an applied science university. These variables enable us to assess whether older, established universities and public institutions, which typically have greater resources, broader networks, and stronger ties to regional innovation ecosystems, demonstrate different patterns of green entrepreneurship generation compared to newer or private institutions. Finally, the research includes a second set of controls for regional socio-economic conditions, specifically GDP per capita and population density. These variables account for spatial heterogeneity in economic development and urbanisation that may independently influence both university characteristics and green entrepreneurial activity. A brief summary of the variables is shown in the following table (see

Table 1. Summary of the variables.

Variable	Source	Brief Description
New green-tech-based firms	EPO-Patstat	Number of new green-tech-based firms
Graduates ISCED57	ETER	Number of graduates at ISCED5 and ISCED7
Graduates ISCED8	ETER	Number of graduates at ISCED8
Graduates STEM	ETER	Share of graduates at ISCED levels 6–7 in STEM fields (ISCED-F 05–07) over total graduates at ISCED levels 6–7
Foundation year	ETER	Average foundation year of universities in the region
Public universities	ETER	Share of public universities in a region
GDP per capita	Eurostat	Gross domestic product per capita of the region
Population density	Eurostat	Population density of the region

Source: authors’ elaboration.

).

3.4. Empirical Model

The empirical analysis models the regional formation of new green technology-based firms using count data methods. Given that the dependent variable represents the number of newly established NGTBFs, it takes the form of a non-negative count variable. For this reason, we employ a Poisson regression model, which is widely used in the empirical literature to analyse count data and the determinants of firm formation (Cameron & Trivedi, 2013). The Poisson specification allows us to account for the discrete and non-negative nature of the dependent variable and to estimate the expected number of new firms conditional on the explanatory variables. The model is estimated including two-way fixed effects for time and region, which allow us to control for unobservable heterogeneity across territories and common shocks over time that may affect entrepreneurial activity. The baseline model is specified as follows:

$$NGTBF{s}_{r,t}={\beta }_0+{\beta }_{1}Gra{d}_{r,t}+{\beta }_{2}Un{i}_{r,t}+{\sigma }_{r,t}+{\eta }_r+{\mu }_t+{ϵ}_{r,t}$$

(1)

where NGTBFs indicates the number of new green-tech-based firms established in a given NUTS-3 region (r) in each calendar year (t). Grad represents a vector of the variables capturing different dimensions of university human capital—graduates at ISCED levels 5–7, doctoral graduates (ISCED 8), and STEM-oriented graduates—for each region and year. Uni represents a vector of controls at the university level in a region r at time t. Finally, $\sigma$r,t represents a vector of regional socio-economic controls at the NUTS-2 level and $ϵ$r,t the error term. We add multiple fixed effects, including regional and time fixed effects, allowing us to control for time-invariant regional characteristics and year-specific shocks.

Although the Poisson estimator is well-suited for modelling count outcomes; it assumes equality between the conditional mean and variance of the dependent variable. To account for potential overdispersion (Greene, 1994), which is common in firm entry data, we perform robustness checks by estimating the model using a negative binomial regression specification. Finally, to ensure robustness of our estimation, we tested our estimates with different model specifications (e.g., Negative Binomial and Linear Probability Model) in Appendix A.

4. Results and Discussion

4.1. Baseline Specification

This section presents the baseline empirical results together with a set of robustness checks aimed at validating the underlying empirical strategy. We begin by reporting alternative model specifications that introduce different measures of regional human capital separately (

Table 2. Empirical estimation results.

Variables	(1) NGTBF	(2) NGTBF	(3) NGTBF
Graduates_ISCED57 (thousands)	0.0341
	(0.0241)
Graduates_ISCED8 (thousands)		1.2200 *
		(0.6570)
Graduates_STEM			2.0550 ***
			(0.6410)
Foundation year	0.0002	0.0002	0.0002
	(0.0001)	(0.0001)	(0.0001)
Public Universities	0.3270	0.2740	0.0978
	(0.3890)	(0.3890)	(0.3120)
GDP per capita	0.0583	0.0641 **	0.1160 ***
	(0.0439)	(0.0299)	(0.0119)
Population density (thousands)	0.3440 *	0.3030	0.2400
	(0.1940)	(0.2090)	(0.2180)
Constant	−2.8780 **	−3.0070 ***	−4.3110 ***
	(1.2180)	(0.8850)	(0.4270)
Observations	707	707	650
Pseudo $R^2$	0.3400	0.3400	0.3400
Region FE	YES	YES	YES
Year FE	YES	YES	YES

Notes: Clustered standard errors at NUTS2 in parentheses. Graduate variables and population density are expressed in thousands. *** $p<0.01$, ** $p<0.05$, * $p<0.1$.

), allowing us to isolate the contribution of each dimension of tertiary education. We then turn to the full specification including all covariates simultaneously (

Table 3. Results of the empirical estimation—Baseline Model.

Dependent Variable	(1) NGTBF	(2) NGTBF
Graduates_ISCED57 (thousands)	$-0.017$	$0.983$
	$\left(0.021\right)$	$\left(0.021\right)$
Graduates_ISCED8 (thousands)	$1.430$ ***	$4.174$ ***
	$\left(0.320\right)$	$\left(1.335\right)$
Graduates_STEM	$1.779$ ***	$5.923$ ***
	$\left(0.564\right)$	$\left(3.338\right)$
Foundation year	$0.001$ **	$1.000$ **
	$\left(0.000\right)$	$\left(0.000\right)$
Public universities	$0.001$	$1.001$
	$\left(0.316\right)$	$\left(0.316\right)$
GDP per capita	$0.075$ **	$1.077$ **
	$\left(0.037\right)$	$\left(0.040\right)$
Population density (thousands)	$0.169$	$1.000$
	$\left(0.212\right)$	$\left(0.000\right)$
Constant	$-3.180$ ***	$0.042$ ***
	$\left(1.055\right)$	$\left(0.044\right)$
Observations	650	650
Pseudo $R^2$	$0.350$	$0.350$
Region FE	YES	YES
Year FE	YES	YES

Notes: Clustered standard errors at NUTS2 level in parentheses. Graduate variables and population density are expressed in thousands. Column (1) reports raw coefficients; Column (2) reports incidence rate ratios (IRR = $e^{\beta }$). *** p < 0.01, ** p < 0.05, * p < 0.1.

), which provides a comprehensive assessment of their joint contributions.

Table 2 reports three alternative specifications examining the relationship between regional human capital and the creation of NGTBFs. Column (1) includes the total number of tertiary graduates (ISCED 5–7), column (2) focuses on doctoral graduates (ISCED 8), and column (3) considers STEM-oriented graduates. Across these specifications, the overall number of tertiary graduates does not exhibit a statistically significant association with NGTBF formation. The estimated coefficient is close to zero, indicating that even sizeable increases in generally higher education attainment translate into negligible changes in the expected number of NGTBFs.

By contrast, doctoral graduates display a positive and weakly significant effect (column 2). Given the Poisson specification, the coefficient can be interpreted as a semi-elasticity. In particular, an increase of 0.1 units (corresponding to 100 additional doctoral graduates) corresponds to roughly a 13% increase in expected NGTBF creation, pointing to the relevance of advanced research-oriented human capital.

The strongest effect emerges for the share of STEM graduates (column 3), whose coefficient is large and highly significant. Since this variable is measured as a share, the estimate implies that, for instance, an increase of 10 percentage points corresponds to an increase of nearly 23%, highlighting the critical role of technical and scientific skills in driving green entrepreneurship.

Table 3 reports the full specification including all human capital measures simultaneously. The results confirm the patterns observed in Table 2 while providing a clearer picture of their relative importance. The coefficient for ISCED 5–7 graduates remains statistically insignificant, reinforcing the conclusion that the sheer volume of general higher education does not drive NGTBFs creation. In contrast, doctoral graduates exhibit a positive and highly significant coefficient (Table 3), confirming both the robustness and the economic relevance of this effect. Similarly, the share of STEM graduates retains a large and statistically significant coefficient, indicating that field-specific competencies remain a key driver of green entrepreneurship even when controlling for other forms of human capital. Column (2) translates these estimates into incidence rate ratios, offering a more immediate sense of economic magnitude. Each additional thousand doctoral graduates is associated with a 4.174-fold increase in the expected number of NGTBFs, while a one-unit increase in the share of STEM graduates multiplies the expected rate of green firm formation by a factor of 5.923, identifying field-specific human capital as the strongest predictor in the baseline specification. The IRR for ISCED 5–7 graduates of 0.983 remains virtually indistinguishable from unity, confirming the absence of any meaningful impact of general graduate supply on green firm formation.

Turning to the control variables, GDP per capita consistently shows a positive and statistically significant association across both Table 2 and Table 3. The magnitude of the coefficient suggests that more economically developed regions experience disproportionately higher levels of NGTBF creation, reflecting stronger innovation ecosystems and better access to financial and institutional resources. The foundation year of universities becomes positive and statistically significant in the full model (Table 3), indicating that more recently established institutions are associated with higher levels of green entrepreneurship. This result may reflect the greater adaptability of younger universities, which could be more oriented towards emerging technological fields, interdisciplinary research, and closer engagement with innovation-driven sectors. By contrast, the share of public universities does not display a statistically significant effect in any specification, suggesting that governance structure alone is not a decisive factor once other characteristics are taken into account. Population density appears positive but generally insignificant across both tables, indicating that agglomeration effects are not a primary driver of NGTBF formation in this context.

Taken together, the evidence from Table 2 and Table 3 provides clear support for the theoretical framework and allows for a direct assessment of the proposed hypotheses. First, the lack of a statistically significant coefficient for the stock of university graduates leads to a rejection of Hypothesis 1. Contrary to expectations, the overall stock of general higher education graduates does not appear to be a sufficient condition for fostering the emergence of green technology-based firms. These findings are consistent with those reported by Bergmann et al. (2016), who show that although individual characteristics are important drivers, organisational and regional contexts also play a significant role, exerting heterogeneous effects depending on both the idea and the stage of development of the new venture. This result also challenges conventional human capital perspectives that emphasise the quantity of education, suggesting instead that the relevance of higher education for green entrepreneurship depends critically on its composition and level.

However, strong support is found for Hypothesis 2, confirming that STEM-specific competencies play a crucial role in driving green entrepreneurship, as suggested by Coll-Martínez et al. (2022). This result is consistent with the complex nature of green technological activities (Barbieri et al., 2023), which rely on scientific expertise, engineering capabilities, and the ability to develop and commercialise environmentally oriented innovations. The results also provide robust support for Hypothesis 3, with doctoral graduates’ estimates highlighting the importance of research-oriented human capital. This finding reinforces the idea that advanced scientific training and involvement in knowledge creation processes are key enablers of green technological innovation and entrepreneurial activity (Orsatti et al., 2024; Quatraro & Scandura, 2019).

To further strengthen the robustness of our findings, we re-estimate the specifications reported in Table 2 using one-year lagged explanatory variables. Introducing a temporal lag between human capital variables and NGTBF creation serves two main purposes. First, it helps mitigate concerns of reverse causality, as current levels of green entrepreneurship are unlikely to influence past university output. Second, it allows us to move beyond contemporaneous correlations and better capture the underlying causal mechanism, which plausibly involves a time gap between the production of human capital and its translation into entrepreneurial activity. In particular, the process through which graduates, especially those with advanced or STEM training, enter the labour market, accumulate experience, and eventually engage in firm creation is unlikely to occur instantaneously. The results remain qualitatively unchanged (see

Table A1. Empirical estimation results—lagged specification (1 year lag).

Variables	(1) NGTBF	(2) NGTBF	(3) NGTBF
L.Graduates_ISCED57 (thousands)	0.0342
	(0.0227)
L.Graduates_ISCED8 (thousands)		1.1800 *
		(0.6250)
L.Graduates_STEM			1.1820 **
			(0.5020)
Foundation year	0.0002	0.0002	0.0002
	(0.0001)	(0.0001)	(0.0001)
Public Universities	0.0593	0.0669 **	0.1250 ***
	(0.0426)	(0.0290)	(0.0142)
GDP per capita	0.0005 *	0.0004	0.0005 *
	(0.0002)	(0.0003)	(0.0003)
Population density (thousands)	0.3130	0.2540	0.4230
	(0.3590)	(0.3660)	(0.3790)
Constant	−3.0450 **	−3.2130 ***	−4.7660 ***
	(1.2210)	(0.8860)	(0.4860)
Observations	606	606	554
Pseudo $R^2$	0.3400	0.3400	0.3500
Region FE	YES	YES	YES
Year FE	YES	YES	YES

Notes: Clustered standard errors at NUTS2 in parentheses. Graduate variables and population density are expressed in thousands. *** $p<0.01$, ** $p<0.05$, * $p<0.1$.

in Appendix A.1), reinforcing the robustness of the baseline estimates and supporting the interpretation of a stable relationship between specialised human capital and green entrepreneurship. Longer lag structures are not explored due to the limited time dimension of the dataset, which would substantially reduce the available sample size and compromise estimation reliability. As such, the one-year lag specification represents a reasonable trade-off between addressing endogeneity concerns and preserving sufficient statistical power.

Finally, as discussed in Section 3.4, we perform additional robustness checks using alternative model specifications, namely a Negative Binomial model and a Linear Probability Model. The results are reported in

Table A2. Results of the empirical estimation—negative binomial and LPM specification.

Dependent Variables	(2) NGTBFs	(3) Log NGTBFs
Graduates_ISCED57 (thousands)	−0.0265	0.0237 **
	(0.0261)	(0.0667)
Graduates_ISCED8 (thousands)	1.7300 **	−0.0430
	(0.7900)	(0.3217)
Graduates_STEM	1.8120 ***	0.4680 *
	(0.4620)	(0.2630)
Foundation year	0.0003 **	0.0001 **
	(0.0001)	(0.0000)
Public Universities	−0.0510	−0.1130
	(0.2590)	(0.0750)
GDP per capita	0.0818 ***	0.0337 ***
	(0.0227)	(0.0067)
Population density (thousands)	0.0827	−0.0442
	(0.1940)	(0.4630)
Constant	−3.1750 ***	−0.6000 ***
	(0.6680)	(0.1600)
Observations	650	650
Region FE	YES	YES
Year FE	YES	YES
Pseudo $R^2$	0.2350	0.4970

Notes: Clustered standard errors at NUTS2 in parentheses. Graduate variables and population density are expressed in thousands. *** $p<0.01$, ** $p<0.05$, * $p<0.1$.

. These alternative specifications broadly confirm the baseline findings. The Negative Binomial model, which is more appropriate for over-dispersed count data, yields estimates that are highly consistent with the Poisson results, particularly with respect to the positive and significant associations of doctoral (ISCED 8) and STEM graduates. The Linear Probability Model, while less suitable for modelling count outcomes and, thus, included primarily as an exploratory exercise, does not fundamentally alter the main conclusions. In particular, the positive role of STEM specialisation remains robust, while the effects of other human capital variables show weaker and less stable patterns, as expected given the limitations of this specification.

4.2. Regional Heterogeneity: The North–South Divide

The second part of the analysis examines whether the relationship between university-generated human capital and green technology-based firm formation varies across regions, focusing on Italy’s North–South divide, a defining feature of the country’s socio-economic structure. Given the substantial differences in economic development, institutional quality, and innovation capacity between Northern and Southern regions, it is plausible that the contribution of human capital to green entrepreneurship is not uniform across space. To this end, we augment the baseline specification by interacting a dummy variable equal to one for Southern regions with each of the key human capital variables. This approach allows the marginal effects of different types of graduates to vary systematically between Northern and Southern Italy. The inclusion of these interaction terms provides a direct empirical test of Hypothesis 4. The resulting estimates are reported in

Table 4. Interaction between graduates and Southern regions.

Variables	(1) NGTBF	(2) NGTBF	(3) NGTBF
Graduates_ISCED57 (thousands)	0.0189	0.0144	0.0215
	(0.0196)	(0.0196)	(0.0205)
Graduates_ISCED8 (thousands)	0.2550	0.4030	0.2240
	(0.5520)	(0.5500)	(0.5710)
Graduates_STEM	1.6520 **	1.6590 **	1.5670 **
	(0.7090)	(0.7160)	(0.7590)
Graduates_ISCED8 (thousands) × South	2.3800 ***
	(0.8080)
Graduates_ISCED57 (thousands) × South		0.0683 ***
		(0.0180)
Graduates_STEM × South			1.6980
			(1.5210)
South	−1.1150 ***	−1.1040 ***	−0.9560 ***
	(0.3480)	(0.3340)	(0.3340)
Constant	−2.7170 ***	−2.7630 ***	−2.6530 ***
	(0.5970)	(0.6030)	(0.5860)
Observations	692	692	692
Pseudo $R^2$	0.3400	0.3400	0.3400
Controls	YES	YES	YES
Region FE	NO	NO	NO
Year FE	YES	YES	YES

Notes: Clustered standard errors at NUTS2 in parentheses. Graduate variables are expressed in thousands. All models include the full set of control variables used in the baseline specification, namely foundation year, share of public universities, GDP per capita, and population density. *** $p<0.01$, ** $p<0.05$, * $p<0.1$.

.

The results reveal significant heterogeneity in the role of human capital across regions. The coefficient on the South dummy is negative and statistically significant across all specifications, confirming that Southern regions exhibit lower levels of green technology-based firm creation, conditional on observable characteristics. Focusing on the interaction terms, the evidence suggests that the contribution of advanced human capital differs markedly across regions. In particular, the interaction between doctoral graduates and the South dummy is positive and statistically significant, indicating that the marginal effect of PhD-level human capital is substantially stronger in Southern regions. This result provides strong empirical support for the compensatory mechanism outlined in Hypothesis 4, whereby research-oriented human capital substitutes for missing institutional and innovation-related resources. A similar pattern emerges for graduates at ISCED levels 5–7, whose interaction with the South dummy is positive and significant, although smaller in magnitude. This result may indicate that, while general higher education is not a key driver of NGTBF formation at the national level, it becomes more relevant in less developed regions where baseline levels of human capital are lower. By contrast, the interaction term for STEM graduates is not statistically significant, suggesting that the effect of field-specific technical skills does not differ substantially between Northern and Southern regions. This may reflect the fact that STEM competencies are a fundamental input for green entrepreneurship regardless of the regional context. In relation to Hypothesis 4, this finding indicates that not all forms of human capital exhibit compensatory dynamics, and that certain competencies may operate as universally binding constraints for green firm creation.

To provide better interpretability to these interactions,

Table 5. Interaction between graduates and Southern regions—incidence rate ratios.

Variables	(1) NGTBF	(2) NGTBF	(3) NGTBF
Graduates_ISCED8 (thousands) × South	10.831 ***
	(8.750)
Graduates_ISCED57 (thousands) × South		1.068 ***
		(0.018)
Graduates_STEM × South			5.464
			(8.312)
Observations	692	692	692
Pseudo $R^2$	0.3400	0.3400	0.3400
Controls	YES	YES	YES
Region FE	NO	NO	NO
Year FE	YES	YES	YES

Notes: Clustered standard errors at NUTS2 in parentheses. Graduate variables are expressed in thousands. All models include the full set of control variables used in the baseline specification, namely foundation year, share of public universities, GDP per capita, and population density. *** $p<0.01$, ** $p<0.05$, * $p<0.1$.

presents the IRRs for the interaction terms between graduates and southern regions. The IRR for doctoral graduates (ISCED 8) interacted with the South is 10.831, implying that the rate of new green-tech-based firm formation is over ten times higher in southern regions for each additional thousand PhD graduates. Similarly, the interaction between bachelor and master graduates (ISCED 5–7) and South yields an IRR of 1.068, meaning that the positive effect of undergraduate and postgraduate supply on green tech entrepreneurship is approximately 6.8% larger in southern regions. Finally, the interaction between STEM graduates and South, while yielding an IRR of 5.464, is not statistically significant. Overall, the results indicate that while Southern regions lag behind in terms of NGTBF creation, certain types of human capital, particularly doctoral training, play a relatively stronger role in these contexts, partially compensating for structural disadvantages. Taken together, these findings provide consistent support for Hypothesis 4, highlighting that the impact of university-generated human capital is context-dependent and may be amplified in regions where complementary innovation system components are weaker.

As an additional robustness check, we estimate the baseline model separately for Northern and Southern regions, with all models including the full set of control variables and year fixed effects, and ensuring that the estimated differences capture meaningful heterogeneity rather than compositional effects. The results, reported in

Table A3. Results of the empirical estimation—baseline model: the North/South divide.

Dependent Variable	(South) NGTBF	(North) NGTBF
Graduates_ISCED57 (thousands)	0.0177	0.0274
	(0.0404)	(0.0176)
Graduates_ISCED8 (thousands)	3.9800 ***	1.4300 ***
	(1.2800)	(0.2990)
Graduates_STEM	2.4480 ***	1.7600 ***
	(0.6380)	(0.6480)
Foundation year	0.0010 **	0.0001
	(0.0005)	(0.0001)
Public Universities	−0.7530 *	0.2350
	(0.4480)	(0.3490)
GDP per capita	−0.1270	0.1010 ***
	(0.0848)	(0.0349)
Population density (thousands)	−0.7090	0.2940
	(0.6610)	(0.2040)
Constant	−0.4800	−3.9320 ***
	(1.7530)	(1.0530)
Observations	214	436
Pseudo $R^2$	0.1838	0.3286
Region FE	YES	YES
Year FE	YES	YES

Notes: Clustered standard errors at NUTS2 in parentheses. Graduate variables and population density are expressed in thousands. *** $p<0.01$, ** $p<0.05$, * $p<0.1$.

, are consistent with the interaction analysis and confirm the presence of substantial regional heterogeneity in the relationship between human capital and NGTBF formation.

5. Conclusions

This paper investigates the role of university graduates as a location determinant for Italian new green-tech-based firms across NUTS-3 regions over the period 2011–2017. Addressing an evidence gap in the existing literature, we examine whether local high-skilled human capital fosters the formation of NGTBFs. Following established approaches in the literature on new green-tech-based entrepreneurship (Coll-Martínez et al., 2022; Colombelli & Quatraro, 2019; Lelli et al., 2025), we adopt a patent-based definition of new green tech firms and develop an econometric strategy that accounts for regional heterogeneity in the spatial distribution of these firms with respect to local high-skilled human capital endowments.

The results presented in this study have relevant implications for regional entrepreneurship dynamics. In particular, the findings suggest that the expansion of higher education in quantitative terms may not, by itself, be sufficient to foster green technology-based entrepreneurship (Bergmann et al., 2016). Rather, the evidence points to the importance of more specialised forms of human capital, indicating that green entrepreneurial activity may depend more strongly on the composition and quality of knowledge production than on its aggregate scale (Coll-Martínez et al., 2022; Colombelli & Quatraro, 2019). This interpretation is consistent with previous research emphasising the role of advanced and field-specific skills in driving innovation and technology-based entrepreneurship (e.g., Vona et al. (2015)). Second, the results consistently highlight the central role of advanced and specialised human capital, with doctoral graduates presenting a positive correlation with new green tech firm creation. This underscores the importance of research-intensive capabilities in fostering green innovation and sustainable development (Orsatti et al., 2024; Quatraro & Scandura, 2019). Third, the STEM orientation of university output emerges as a key driver of regional green entrepreneurship, with coefficients that are large and highly significant in all specifications. This finding reinforces the idea that green technological innovation is inherently science- and engineering-intensive, and this result puts universities as a key player in the regional innovation ecosystems, emphasising their role in producing STEM-oriented human capital (Coll-Martínez et al., 2022). Fourth, the analysis of the North–South divide that characterises the Italian case reveals substantial spatial heterogeneity in the role of university human capital. In Southern regions, doctoral and STEM graduates exhibit markedly stronger associations with NGTBF formation, suggesting that highly specialised human capital plays a compensatory role in less developed innovation systems. In contexts where entrepreneurial ecosystems and institutional supports are weaker, scarce high-level skills appear to generate larger spillovers, and this represents a relevant element for regional policymakers in developing sustainable trajectories for their region.

The outcomes of this paper contribute to the current literature on new green entrepreneurship, providing the first empirical evidence on the role of universities as a provider of high-skilled human capital, a fundamental driver in the enhancement of these companies. In particular, this research disentangles the contribution of different levels and orientations of university graduates, highlighting the distinctive importance of doctoral-level and STEM-specific local human capital in fostering new green technology-based firms. By linking university output to regional entrepreneurial outcomes, the paper complements and extends the knowledge spillover theory of entrepreneurship (Acs et al., 2009; D. Audretsch et al., 2004), explaining how the offer of high-skilled local graduates might represent an important driver for green venture creation. Moreover, this research highlights regional differences in this relationship, exploring the North–South divide, revealing a strong spatial heterogeneity in these mechanisms, and demonstrating that advanced and specialised university human capital plays a compensatory role in less developed regional innovation systems. In doing so, the paper contributes novel insights into how universities shape the geography of new green-tech-based entrepreneurship and provides a more nuanced understanding of the conditions under which higher education institutions can effectively support the green transition.

Several limitations of this study should be acknowledged, as they also highlight important avenues for future research. First, the analysis adopts a technology-based definition of green entrepreneurship that is well established in the literature (Coll-Martínez et al., 2022; Colombelli & Quatraro, 2019; Corradini, 2019; Lelli et al., 2025). However, this approach captures only a subset of green entrepreneurial activity, excluding non-technological and service-oriented firms. Future research could extend the analysis by employing broader definitions of new green entrepreneurship to provide a more comprehensive analysis of the phenomenon.

Second, while the paper takes an important step towards unpacking the role of university-generated human capital, it remains a first approximation. Data limitations prevent a detailed analysis of specific curricula or fields of study and their direct connection to green-related competencies. Future research could overcome this limitation by incorporating more granular information on educational content, for example, through curriculum analysis using advanced data mining techniques and large language models. In addition, the lack of data on graduate mobility—particularly relevant in the European context, where skilled individuals frequently move across regions—limits our ability to fully capture the spatial allocation of human capital and may bias estimates of the relationship between locally produced graduates and regional green entrepreneurship.

Third, the temporal and geographical scope of the analysis represents an additional constraint. The selected time period reflects a trade-off between data availability and the need to avoid major exogenous shocks—such as the 2009 global financial crisis and the COVID-19 pandemic—that could confound the observed relationships. While this choice enhances internal validity, it limits the ability to assess longer-term dynamics. Similarly, the focus on Italy is motivated by its pronounced regional disparities and heterogeneous performance in eco-innovation, which provides a suitable setting for examining spatial heterogeneity. Nonetheless, extending the analysis to a broader European context would allow for more systematic cross-country comparisons and contribute to a deeper understanding of the relationship between universities and green entrepreneurship.

Finally, the estimates may be affected by reverse causality and omitted variable bias. Regarding the former, regions with higher levels of green entrepreneurial activity may attract universities or influence their educational offer, generating a feedback loop between NGTBF formation and the production of STEM and doctoral graduates. Regarding the latter, unobserved regional characteristics such as the presence of green industrial clusters or the quality of regional governance may simultaneously drive both university output and green firm creation. Although the use of lagged specifications and the inclusion of region and year fixed effects help mitigate these concerns, they cannot fully address time-varying endogeneity. A more rigorous treatment would require instrumental variables such as historical university foundation dates or shifts in national higher education funding policies, which fall outside the scope of the current dataset. For these reasons, the results should be interpreted as robust correlational evidence rather than strictly causal estimates, and future research should prioritise the development of credible identification strategies.

The research carries clear policy implications for the design of strategies supporting the green transition. The findings highlight that high-skilled and specialised human capital, particularly doctoral and STEM-trained graduates, represents a critical driver in fostering new green-technology-based entrepreneurship, suggesting that policies focused solely on expanding tertiary education in quantitative terms are insufficient. Instead, three distinct policy levers can be identified. At the national level, policymakers may prioritise structural investments in research-intensive education, specifically by expanding doctoral programmes in STEM fields and strengthening the research capacity of universities. This may include dedicated funding schemes for green-oriented PhD training, such as industrial doctorates in partnership with clean technology firms, or the creation of green research centres within universities. At the regional level, the results call for place-based strategies that explicitly account for territorial heterogeneity. In less developed regions, where specialised human capital generates disproportionately large spillovers, targeted interventions such as incentives to attract and retain PhD and STEM graduates or the creation of regional university-industry platforms focused on green technologies could help compensate for weaker innovation ecosystems. At the university level, institutions could be encouraged to align their educational offer with the competencies demanded by the green transition, for instance, through the integration of sustainability-oriented curricula within STEM programmes and the promotion of entrepreneurship education for doctoral students. More broadly, one-size-fits-all approaches are not suited to fostering green entrepreneurship; differentiated, place-based strategies that account for regional innovation capacity, institutional maturity, and human capital endowments are needed to effectively support a sustainable and inclusive green transition.