Firm performance after incubation: public incubators and knowledge-intensive entrepreneurial firms

Entrepreneurship plays a central role in the economy, as new firms experiment with and promote novel ideas that contribute to overall growth (Wennekers and Thurik, 1999). Through a Schumpeterian approach, a long line of literature has been stressing the importance of innovative entrepreneurship to the economy (Klepper, 2016; Van Praag and Versloot, 2007). Malerba et al. (2016) and Lindholm-Dahlstrand et al. (2019) argue that particularly entrepreneurial firms based upon advanced technologies and knowledge domains can play an important entrepreneurial function in promoting entrepreneurial experimentation and causing dynamics. Despite these contributions, limited attention has been paid to how survival relates to firm performance among a larger set of firms emerging from public incubation settings. This paper addresses this gap by analyzing both survival and post-incubation performance of innovative entrepreneurial firms. Specifically, we examine how knowledge specialization during incubation influences firms' subsequent growth and development.

Our theoretical frame of reference is the Schumpeterian conceptualization of knowledge-intensive entrepreneurial (KIE) firms (Malerba and McKelvey, 2019, 2020), which face outlier conditions as compared to conventional entrepreneurs. We adopt the definition of Malerba and McKelvey (2020, p. 508), who define KIE firms as “new learning organizations that use and transform existing knowledge and generate new knowledge in order to innovate within innovation systems”.

KIE founders are exposed to fast-moving technologies and industries, while seeking to combine entrepreneurial vision with the commercialization of scientific and technical knowledge across and within sectors (McKelvey and Lassen, 2013). ^[1] These firms rely heavily on other actors in their innovation ecosystem, in terms of access to resources, contacts and realization of opportunities (Malerba and McKelvey, 2020). Given their dependency upon external knowledge and resources (as well as internal processes), KIE firms that have completed programs and training at public incubators are a relevant setting to examine survival and growth over time.

Entrepreneurs can be supported in their entrepreneurial journey and development. Public business incubators – including those associated with universities – can be understood as policy instruments for stimulating entrepreneurship. The intended mission of incubators is to assist entrepreneurs in translating their ideas into a firm by providing access to different types of resources, network contacts and talent (Klingbeil and Semrau, 2017; Kitagawa and Robertson, 2012; Markman et al., 2008). Identified as a favorable environment for founders that are new to entrepreneurship (Guerrero et al., 2021), incubators help start-ups overcome various entrepreneurial obstacles, including “liability of newness” (Stinchcombe, 1965) and “liability of smallness” (Freeman et al., 1983). From a policy perspective, incubators have been seen as an instrument to stimulate long-term regional development and economic growth via the fostering of technological entrepreneurship and high-tech venture creation (Madaleno et al., 2022; Aerts et al., 2007; Phillips, 2002; Thierstein and Willhelm, 2001; Mian, 1997).

This paper contributes to the literature on incubators by differentiating two dimensions of knowledge specialization− those that relate to (1) founder background and (2) industrial sector. Specifically, we analyze how both dimensions of incubator specialization are related to survival and longer-term performance of incubated firms. We will examine KIE firms which have completed programs at similar public incubators in one country, allowing identification of a population of firms that have all survived to the point of completing incubation. One type of knowledge specialization refers to the background of the founders. Our database contains detailed information on the background of five different types of founders of KIE firms, including but not limited to founders with backgrounds as university researchers. The different founder types represent the previous experience of the founder and are assumed to bring relevant knowledge into the venture. The general importance of founder backgrounds as key determinants of firm capabilities has long been recognized (Helfat and Lieberman, 2002). Another type of knowledge specialization within the incubator refers to the density of other KIE firms within the same industrial sector, as compared to more diverse sets of firms in the incubator setting. Apart from work by Gifford et al. (2020), these questions have not been analyzed systematically from a KIE perspective, and particularly in relation to the surviving firms, nor has specialization along founder backgrounds been the focus of prior work on public incubators.

Empirically, we used and further developed our analysis, based upon a 15-year longitudinal dataset of entrepreneurial firms (⁠$n=852$⁠) that completed incubation between 2005 and 2015 at Swedish public incubators (⁠$nk=37$⁠) ^[2]. The dataset contains detailed variables at the incubator, firm and founder level, which allows us to measure factors associated with incubator specialization; including industrial sector, founder background, as well as variables that pertain to knowledge exposure during incubation. We then augmented the database with micro-data on survival and firm performance after leaving the incubator. We first explore factors behind firm survival. We find that survival is common, and most firms persist five years post-incubation. We then relate knowledge specialization during the incubation period, to analyze later firm growth in terms of both revenue and employees.

This paper is structured as follows. In the next section we review the literature on the knowledge specialization of public incubators and hypothesize their potential relationships with KIE firm post-incubation survival and performance. Section 3 presents details of our methods and an overview of our dataset, followed by results and conclusions in Sections 4 and 5.

Incubators may also help realize the societal potential of individuals taking an entrepreneurial path, as countries appear to have varying numbers of potential “nascent” entrepreneurs (Delmar and Davidsson, 2000). Extant research has addressed the performance of incubators (Lasrado et al., 2016; Lukeš et al., 2019; Schwartz, 2009; Aernoudt, 2004; Sherman and Chappell, 1998) as well as alternative organizational design for effective incubation (Aaboen, 2009; Bergek and Norman, 2008). An extensive stream of literature addresses incubator characteristics. Prior research has shown that efficiency and networking benefits, especially those that pertain to suppliers, customers and partnerships; are associated with larger incubators (Klingbeil and Semrau, 2017; Gerlach and Brem, 2015; Schwartz and Hornych, 2010; Aerts et al., 2007; Allen and McCluskey, 1990). A more recent debate addresses the organizational design of specialized incubators, which suggests that firm performance after incubation may be related to processes during incubation, and in ways that are not yet fully understood (Sohail et al., 2023). Our conceptualization is that incubators, especially publicly funded ones, provide a way to define a set of founders and firms, which are provided institutional support to access internal and external networks of knowledge, as relevant to their business.

Our focus is on public business incubators – including university incubators – which we conceptualize as a tool for public policy for stimulating KIE firms. Founders of KIE firms require diverse resources and incubators should aim to prioritize these resources at different stages of firm development (Chan and Lau, 2005). A useful starting definition is thus “a support environment for start-up and fledging companies” (Peters et al., 2004, p. 83), even though multiple, overlapping definitions for business incubators exist (Hackett and Dilts, 2004). According to Salvador and Rolfo (2011), incubators identify resource-constrained projects that are deemed to have a high probability in succeeding and assist new firms in overcoming the “liability of newness” and “liability of smallness”. Incubators directly assist founders with the aim of transforming ideas into firms, along with the relevant resources, routines and capabilities (Breivik-Meyer et al., 2020), but may also provide indirect returns such as those that pertain to co-location alongside other KIE firms (McAdam and Marlow, 2008).

A relevant issue is how public incubators promote KIE firm creation. Following Bruneel et al. (2012), we distinguish three dimensions of incubator support. The first is shared infrastructure, including office space offered at favorable conditions (Bergek and Norman, 2008; Carayannis and Zedtwitz, 2005) which may incorporate ancillary resources such as clerical services, meeting rooms and conference venues (McAdam and McAdam, 2008), as well as specialized premises such as laboratories and specialized equipment (Grimaldi and Grandi, 2005). A second dimension is provision of business support addressing managerial challenges and skill scarcity. Given the relatively undeveloped firm-specific routines of young firms which requires a gradual process of experimental learning (Dosi et al., 2000), incubators may help with the problem of skill scarcity, offer training and business coaching (Clarysse and Bruneel, 2007) and provide access to specialized services–- especially technical, funding and legal assistance (Bruneel et al., 2012; Grimaldi and Grandi, 2005). The third dimension is access to networks, both external and internal networks to the incubator and firms located there. External professional networks, which have been identified as a critical factor for incubatee prospects (McAdam and McAdam, 2008; Larson, 1992), serve to connect firms to a range of actors such as investors, suppliers, providers of technology, patent attorneys, consultants, academic institutions and customers (Schwartz and Hornych, 2010; Hansen et al., 2000), as well as multinational corporations and other large enterprises (Vanderstraeten et al., 2016; Bergek and Norman, 2008). Moreover, internal networks within incubators, according to McAdam and Marlow (2008), may facilitate the exchange of knowledge between incubating firms (Kitagawa and Robertson, 2012; Witt, 2004), lead to the formation of alliances and the gaining credibility among peers and broker relationships with entities that expedite entrepreneurial goals (Jack, 2005; Witt, 2004).

We conceptualize public business incubators – including university incubators – as providing relevant support for entrepreneurship and the specialized knowledge search processes of KIE firms.

One type of knowledge specialization refers internally to the firm, namely founder background. The composition of tenants in terms of founder experience may provide a source of knowledge specialization within incubators. An incubator provides its firms with various possibilities to interact across different types of knowledge bases. We thus consider both direct and indirect roles that researcher-led firms play in public incubators in relation to other types of founders: students, (non-researcher) university staff, corporate spin-offs and independent inventors. Researchers engaging in other engagement activities like collaborative research for example, is far more common than researchers engaging in firm formation (Perkmann et al., 2013). Furthermore, peer influence, role models and the individual's orientation toward commercial science seems to influence this choice (Perkmann et al., 2013; Stuart and Ding, 2006). Students in technology and science have been identified as more productive than researchers in starting firms (Åstebro et al., 2012). They have also been identified as having dual roles by both developing their own ideas but also working in incubation projects based on researcher IP (Brunnström et al., 2020; Culkin, 2013). Due to the practice of interpreting inventor ownership legislation in Sweden, university staff other than researcher/teachers, are usually awarded the same rights as researchers. Åstebro et al. (2019) finds that the likelihood of university employees becoming entrepreneurs was lower than that of non-university employees, but higher in Sweden than in the US. There are good reasons to think that founders starting a corporate spin-off would have advantages, not only in firm-formation, but also post-incubation in the marketplace (Dahl and Reichstein, 2007). Their advantages are related to resources such as capital, human capital, knowledge and know-how “inherited” from the mother firm. However, these comparative advantages may diminish with time (Ortín-Ángel and Vendrell-Herrero, 2014). Although not previously analyzed much in relation to public incubators, independent inventors are present in our data. Independent inventors refer to individuals who may lack the inherited endowments and resources associated with a corporate spin-off.

The second type of knowledge specialization has to do with the density of more firms in the same industry. Incubators that have higher sector-specific specialization appear to provide better tailored facilities and equipment, as well as specialized services (Schwartz and Hornych, 2008). Within an industry, specialized networks may also enable the transfer of industry-specific knowledge which provide solutions to product and routine development (Giuliani, 2010). However, when examining diversified and specialized business incubators, Schwartz and Hornych (2010) find no association between incubator specialization and an increase in inter-firm networks; nor do they find that specialization leads to more emphasis on academic linkages. Hansen et al. (2000) argue that specialized incubators represent a new organizational form that promotes scale and scope, entrepreneurial drive and network access, thereby offering benefits previously associated with large established companies and venture capital firms. However, prior work on social network dynamics among entrepreneurs in incubators finds no systematic relationship between networking and the number of venture partners (Ebbers, 2014).

Based on our interpretation of the above theories, we propose three hypotheses. Firstly, the survival after incubation needs to be examined and described in relation to the diversity of founder background, which is an indicator of internal knowledge specialization. Secondly, the firm performance after incubation is related to the density of other founders with similar backgrounds, for example the knowledge specialization as represented by the constituent type of KIE firm. Thirdly, firm performance after incubation is related to the density of other similar KIE firms within the same industry, for example the knowledge specialization as the composition of the set of firms in an incubator. We are assuming that more knowledge specialization promotes firm performance, as conditions in the incubator may provide graduating firms with some critical conditions for knowledge creation and learning and thereby provide a competitive advantage in the marketplace.

One type of KIE ventures are started by researchers, which we are assuming have skills, knowledge and experience which makes them closer to current science. Researchers and science affect society in many ways.

Founders with a researcher background are expected to have the strongest linkages to a university. Evidence suggests that university linkages are important to firm growth during incubation as well as post-incubation (Lasrado et al., 2016), as well as increased credibility for the firm and many benefits for the university as well (Hassan, 2024). However, Rothaermel and Thursby (2005) also found that stronger university connections (defining founders being researchers) led to projects incubated for longer compared to others. Moreover, Johansson et al. (2005), in research on non-incubated spin-offs, found that firms with stronger ties to universities enjoyed higher rates of employee growth. They argued that fewer but more intimate connections would give access to basic research support which Tether (2002) found spin-offs to be reliant on. Within the Swedish context, professors' privilege law (which guarantees the rights to research ownership unless otherwise agreed) has been found to be more effective in generating researcher-led firms compared to, for example, patents and licensing (Färnstrand-Damsgaard and Thursby, 2013). Combined with stronger university ties through employment and personal networks, this in-turn may offer a further incentive for researchers to engage and potentially thrive in entrepreneurial firms.

Researchers establishing KIE firms may also impact other types of co-incubated projects. Brunnström et al. (2020) found that their presence was correlated with a shorter incubation time of other projects. To our knowledge, prior research has not extended this analysis to the post-incubation stage, and we have few indications of whether and how researchers play a role within public incubators, including those associated with universities.

Evidence for localized knowledge spillovers from science to industry is pervasive in a variety of contexts and at different levels of aggregation (Salter and Martin, 2001; Jaffe, 1989). Fini et al. (2009) describe researchers as seekers of primarily non-market-based rewards. Jensen and Thursby (2001) identify concerns regarding academic reputation, which may compete with other time uses and hinder commercialization of projects. Researchers are highly educated and sought after in the labor market; hence their opportunity costs tend to be higher than for other types of founders. Furthermore, Czarnitzki et al. (2014) argue that researchers who migrate from non-profit to for-profit firms incur higher social costs. This, they argue, creates a need for a performance premium compared to other startups in knowledge-intensive industries, which is corroborated by their finding of a positive performance premium of researchers and other university employees.

A variety of explanations may be possible. One argument is that researcher knowledge is tacit (Agrawal, 2001), and can only be accessed by those located close to the researcher. Another argument is that researchers may function as brokers, in the sense that they can connect co-located non-researcher-led startups to university researchers outside the incubator (Lissoni, 2010). A further argument is that both direct and indirect linkages to universities may enable firms to keep up to date with recent scientific advances relevant to their firms (Link et al., 2007); including scientific methodology; broad knowledge of the research field, as well as instrumentation. Löfsten (2010), in a study of 131 incubated projects in 16 Swedish university incubators, found no evidence of a relationship between post-incubation performance and an incubator's internal and external networks. However, regarding later-stage firm development, Löfsten (2010) inferred that an incubator's internal network is a potentially important factor for firm growth. Brunnström et al. (2020) found that researcher-led projects are less likely to complete incubation compared to projects led by founders from other backgrounds. In our context, public university incubators, by providing shared infrastructure, also provide repeated interactions among different categories of founders, which may help activate knowledge networks between researchers and non-researchers.

We compile our dataset from three sources. Our starting point is a database from the Swedish government agency for innovation (VINNOVA) on all 42 Swedish incubators that received funding from Vinnova and 4,800 associated projects for the 2005–2014 period ^[3]. This dataset includes variables related to incubator and firm characteristics and founder experience, covering most of the population of public incubators in Sweden.

The initial dataset consisted of 42 incubators. Five incubators were removed as they did not meet the following two conditions: (1) the incubator existed in the dataset over at least five years in the chosen time-period to enable panel data, (2) complete data were available regarding the chosen relevant variables for both the projects, start-up firms and incubators and (3) the incubator could be classified as public. This resulted in a sample of $nk=37$ incubators that could be classified as public, leaving us with 3,370 associated projects, of these 852 firms graduated during the period. We link these firm and incubator level data with two types of firm level post-graduation data on the graduated firms, following them until 2020 in terms of survival and 2019 in terms of growth. This end time-period was chosen as we did not want the onset and consequences of the COVID-19 pandemic to potentially influence our results. The second dataset comprises manually collected records of industry classification and operating status for each graduated firm ^[4] (⁠$n=852$⁠), compiled from the Swedish Companies Registration Office (Bolagsverket). The third dataset is provided by Statistics Sweden (SCB), which lists revenues and employee numbers for each of our graduated 852 firms at graduation and, three- and five-years post-graduation.

We first analyze post-incubation survival rates among graduated firms by employing a non-parametric Kaplan–Meier estimator of the survival function. Our data contain information on exited firms on a monthly interval. Exits via acquisitions are treated as censored. We also carry out a semi-parametric, Cox proportional hazards model to further analyze factors that may be associated with differences in post-incubation firm survival.

Although firms may persist in the market, they may not grow or flourish. In addition to survival analysis, we accordingly address firm performance in terms of revenue and employee growth. Our analysis of firm growth uses log-linear OLS regressions for three- and five-year outcomes. Standard errors are clustered at the incubator level to account for incubator-specific effects.

Revenue and employee data are categorical, a summary may be found in Table 1. We transform these data to arrive at variables expressed in number of employees for the year of graduation from the incubator program (⁠$EmpYear0i$⁠) and three and five years of graduation (⁠$EmpYear3i$⁠, $EmpYear5i)$ by taking the midpoints of each range in each category. Similarly, we take the midpoints for revenue for each year of observation (⁠$RevYear0i$⁠, $RevYear3i$ and $RevYear5i$⁠), expressed in thousands of SEK.

The descriptive statistics in Table 2 reveal that, post-incubation, firms grow in both employees and revenue. Mean revenue displays linear growth, albeit with increasing variability. The mean number of employees on the other hand plateaus after $EmpYear3i$⁠, yet with a high level of dispersion.

To analyze industrial specialization as addressed in H3, our dataset incorporates two levels of industrial sector categorization, one on a broader level and another on a finer grain level, a list of which may be found in Table 3.

We construct two variables, $SectorSpecLikt$ and $SectorSpecSikt$⁠, which measure the share of other firms in the same sector classification, at the broader and finer levels respectively, and at the time of entry (⁠$t$⁠) to the incubator $k$ for firm $i$ ^[5]. Note that both measures exclude firm $i$ as we are interested in exposure to other projects during incubation ^[6].

Table 2 shows that $SectorSpecLikt$ is, on average, higher than $SectorSpecSikt$⁠, which, due to the former encompassing broader categories, is to be expected. A mean of 0.397 and 0.202 for $SectorSpecLikt$ and $SectorSpecSikt$ respectively indicate that incubators are more inclined to foster projects that are industrially similar, as these figures are higher than 0.25 (for $SectorSpecLikt$⁠) and 0.05 (for $SectorSpecSikt$⁠), which would represent a perfect distribution of sectors (given the number of industrial categories in each level).

H2 address the relationship between diversity of founder experience and post-incubation firm performance. The dataset records five founder types: (1) researchers, that conduct research or teach at a Swedish university or research institute (⁠$Researcheri$⁠, 17%); (2) students, that are admitted to a course or program at a Swedish university (⁠$Studenti$⁠, 13.2%); (3) university staff, both technical and administrative (⁠$UnivEmployeei$⁠, 10.5%); (4) independent inventors, that is individuals unaffiliated with a firm or university $(Inventori,22%)$ and (5) corporate spin-offs, that is projects started by an incumbent firm (⁠$Spinoffi$⁠, 37.4%).

We define founder specialization (⁠$FounderSpecikt$⁠) as the share of projects with the same background as project $i$⁠, at the time of entry (⁠$t$⁠) to incubator $k$⁠, again excluding focal firm $i$ ^[7]. We also include an additional variable, $Backgroundsikt$⁠, representing the count of the number of other founder background types hosted by the incubator (i.e. not including the background of the founder of firm $i$⁠) at the time of incubator entry. A mean of 0.323 in Table 2 for $FounderSpecikt$ indicates that incubators tend to foster projects with founders of similar backgrounds. On average, a firm will accompany projects from 2.868 other backgrounds upon entry to the incubator.

$Researcheri$⁠, the variable of interest for H4, is a dummy taking on the value of 1 if the founder's background is that of a researcher and 0 otherwise ^[8]. 17% of founders come from a researcher background. To address the level of exposure of firms to researcher-led projects during incubation as proposed in H5, we construct the researcher exposure index $ResExposureik$ as described in Equation (1):

where $τi,k$ denotes the period of time project $i$ spent at incubator $k$⁠, $τj,k$ denotes the period of time any other project (apart from $i$⁠) spent at incubator $k$ and $rj,k$ takes on the value of 1 if $τj,k$ is a researcher-led project and 0 otherwise. Hence, $ResExposureikt$ represents the percentage of exposure to researcher-led projects as a proportion to the total exposure to all projects during the incubation period. The descriptive statistics in Table 1 indicate that on average, projects are exposed to researcher-led projects 19.3% of their total exposure time during their incubation period, with a considerable level of variation ranging from 0% to 87.5%.

We expect that incubator managers attempt to select the most promising projects from those that apply. The larger the pool of applicants, the more selective they may be. Assuming incubator managers have some ability to predict success, selectivity should relate to performance. We thus define $Selectivityikt$ as the ratio of the number of projects applying (and evaluated) to the number of projects accepted at incubator $k$⁠. In other words, $Selectivityikt$ measures the number of applicants for each accepted incubator project. This ratio is compiled for year $t$ when firm $i$ enters incubation. Thus, higher values of $Selectivityikt$ indicate a more selective incubator. Table 2 indicates that there is considerable range with this variable, with an average of 11.89 applications to a place.

$IncTimei$ describes the length of time of incubation for firm $i$ measured in months, which, on average, was 24. $IncAgeik$ records the age of incubator $k$⁠, in months, at the time of graduation for firm $i$⁠. The development stage of an incubator can have a direct relationship with the resources it makes available for its tenants (Lalkaka and Abetti, 1999), and younger incubators have limited resources due to the “liability of newness”. In our regression specifications, we take the log of both $IncTimei$ and $IncAgeik$ to account for expected diminishing returns. We also include $IncEmpik$⁠, which is the number of full-time equivalent employees of incubator $k$⁠, recorded at the time of project $i$’s entry to the incubator. Finally, $Universityk$ takes on the value of 1 if the incubator is university-owned and 0 if municipality-owned.

Pairwise correlations between all variables are shown in Table 4.

Kaplan–Meier survival curves, partitioned by founder type and estimated by months of post-incubation survival, are shown in Figure 1 (for clarity, both graphs have axis breaks just below 0.70 survival probability).

The apparent initial concavity followed by convexity suggests stronger hazards earlier in the firm's lifetime, followed by stabilization as the firm matures, a “liability of newness”. We may observe that, in general, there is little variance in survival probabilities across all firms, with at least 80% of firms surviving after 160 months (>13 years). There is no noticeable variation in patterns of survival rates across founder types, contrary to what was predicted in H1. We estimate a Cox regression to check whether there were any major determinants of post-graduation firm survival, the results of which may be seen in Table 5. The size of the incubator, measured by number of employees (⁠$IncEmpk$⁠) was the only significant estimate in the full model specification (6).

Turning to post-graduation growth performance, we begin by estimating separate models for each of the main explanatory variables (⁠$SectorSpecLikt$⁠, $SectorSpecSikt$⁠, $FounderSpecikt$⁠, $Backgroundsikt$⁠, $Researcheri$ and $ResExposureik$⁠), which are included alongside the control variables. We then estimate a full specification model incorporating all dependent variables. Regression results with respect to 3-year revenue and employee growth are reported in Table 6; results for 5-year growth are reported in Table 7.

Founder experience specialization addressed in H2, with the expectation that $FounderSpecikt$ and $Backgroundsikt$ are associated with post-graduation performance. We see no statistical significance for either coefficient, in both the employment and revenue growth models. In the five-year analysis, we likewise do not obtain significant coefficients. Thus, there is no empirical support for post-incubation performance and diversity of founders in co-incubated firms.

Hypothesis 3 addresses how industrial specialization, as measured by $SectorSpecLikt$ and $SectorSpecSikt$⁠, is related to post-graduation firm performance. $SectorSpecLikt$⁠, which measures industrial specialization when measured in a broader way, is not statistically significantly associated with revenue or employee growth. In contrast, we obtain significantly positive estimates for the narrower definition of specialization, $SectorSpecSikt$⁠, both for revenue (p < 0.10) and employee growth (p < 0.01). These results indicate that projects benefit from close cognitive proximity, providing a post-graduation performance premium via the sharing of tacit information between co-incubatees, in-turn facilitating learning and innovation as well as post-graduation linkages. Five-year results (Table 7) suggest that these benefits may attenuate over time.

Furthermore, following on from H4, firms emerging from researcher-led projects (as measured by $Researcheri$⁠), exhibit significantly higher employment growth (p < 0.01) in their first three years. However, no significant association is found for revenue growth. This may suggest that firms founded by researchers tend to have complex products and services that necessitate an abundance of skilled labor, and/or that, following the reasoning of Jensen and Thursby (2001), their products and services are still in the development phase and not market-ready, therefore requiring additional employees to finalize. Note also that the stronger employee growth of researcher-led firms is limited to the three-year analysis.

Finally, exposure to researcher-led projects during incubation, as measured by $ResExposureik$⁠, has no statistically significant association with either revenue or employment growth, counter to the prediction of H5. Thus, one may infer that co-incubation with researcher-led startups is insufficient to boost firms' post-incubation performance. Co-incubation may not be sufficient to access researchers' tacit knowledge, or access to this tacit knowledge may not have a strong impact on firm performance. This finding diverges from those obtained by Brunnström et al. (2020), who report that projects exposed to higher shares of researchers were more likely to complete incubation and doing so in less time than other projects. These contrasting outcomes may be due to the form of the exchange between project founders (e.g. advice and mentorship) rather than a more permanent transfer of knowledge. Thus, the returns from researcher exposure may be seen at the incubation stage only.

Regarding our control variables on revenue growth, $lnIncAgek$ is statistically insignificant for the first three years. $IncEmpk$⁠, on the other hand, is negative and significant for the first three years of revenue growth. This may be explained by a potential “coddling” effect of incubation, leading to firm difficulties once exposed to the marketplace. This adds to the arguments put forward by prior research that the liability of newness of firms takes effect after graduation and that incubation may, in some regards, exacerbate this liability by providing over support during incubation (Schwartz, 2013; Feldman and Pentland, 2003; Levitt and March, 1988). $Universityk$ remains positive and significant for growth in terms of revenues in both the three-year as well as the five-year models (up to model 5 in both), however for the employee-growth models $Universityk$ have negative and insignificant values throughout. A possible explanation of these results could be the presumably stronger linkages to universities of firms being incubated at university incubators may be linked to stronger growth patterns as suggested by Rothaermel and Thursby (2005) and Lasrado et al.’s (2016) studies. Although, these growth patterns are only visible for growth in terms of revenues.

For further analysis to ensure robustness, we manually scrutinized the top 1% of all firms by employee count and revenue five years post-incubation (Table 8). No discernible pattern could be found in terms of industrial sector, the incubators that hosted them or the origins of the respective businesses. Top performers came from different incubators, and their founders had diverse backgrounds. Even though industry spin-offs and independent inventors were the most common founder background, others, such as researchers and (non-researcher) university employees, were present among the top performers.