1. Introduction
Knowledge created at universities plays an important role in driving innovation and contributes to the economy and society, while its importance has especially grown in recent decades [
1,
2,
3]. Hence, it is also of major interest to decision makers in the public sector, who try to determine the most suitable ways of transferring knowledge from universities and achieving maximum return on public resources [
4,
5].
The methods of transferring knowledge created at universities into society are diverse and can be both official and unofficial [
6,
7]. This study focuses on company creation, as the most explicit and essential way of monetizing academic knowledge [
5,
7,
8,
9,
10]. However, the extant literature indicates that although firm creation as a form of academic entrepreneurship is becoming a mature topic [
11], it is still understudied [
12,
13], and less exploited than the potential alternatives [
14].
As academic workers themselves are usually involved in technology transfer and business creation processes, their individual assets become topical [
15,
16]. In order to explain why some academic workers are more likely to create spin-off companies than others, Landry et al. [
12] created a theoretical concept relying on the resource-based theory of the firm. According to this concept, academic workers as entrepreneurs have access to different resources, which in turn vary from one individual to another and affect the likelihood of firm creation.
The effects of academic assets on company creation might vary during a career. One explanation for the latter could be that scientific productivity is not constant, nor linear, during the careers of academic staff, and hence, life-cycle effects in research productivity exist [
17,
18,
19]. Perkmann et al. [
2] concluded that individual life-cycle effects in academic engagement need further investigation, as the evidence is still inconclusive. Moreover, a longitudinal analysis should be conducted, as the profiles of academic workers evolve over time [
10,
11]. Relying on the latter, the commercialization engagement of academics could be enhanced by accounting for the specific phase of their academic career and other individual characteristics. Therefore, the design of commercialization incentives and measures should also follow the life-cycle approach, for instance, early-stage academics exposed to publishing pressures should be rewarded for committing resources to commercialization [
2].
Academic entrepreneurship can be considered a special case in innovation studies, as in the era of open science the provider (academic worker) and recipient (academic spinoff founder) of open innovation (OI) in the outside-in model (see further [
20]) are the same. The latter is especially important as, despite the increasing diffusion of OI created at universities [
21,
22], its absorption by SMEs other than academic spinoffs has remained modest [
23]. According to the theoretical concept of relational involvement, by Perkmann and Walsh [
24], academic workers creating a firm and retaining their job in academia could be among the best examples for studying OI, mainly because the creation and commercialization of innovations can be seen as a lengthy interactive process [
25], rather than two distant discrete events. Strong internal R&D competences, organically inherent to academic spinoffs, have been found to make firms benefit more from OI [
26].
Relying on the aforementioned motivation, and by linking the resource-based and life-cycle theories, this study aims to find out which academic assets are linked with firm creation by academic staff at different academic career stages. By doing so, the study aims to be the first to outline the dynamic linkage between academic achievements and a specific type of commercialization. The reliability of the results is enhanced by the usage of a factual longitudinal whole population dataset of Estonian academic workers, while the existing research has in turn mostly relied on sample-based questionnaires. The results provide a population level track, which specific academic assets matter at varying career stages, therefore enabling extending the so far statically applied theoretical concept of academic assets as determinants of firm creation into the academic life-cycle context.
The rest of the paper encompasses the following sections. In the next part, the theoretical foundations of academic entrepreneurship, academic assets, and academic life-cycle are discussed and interlinked. The third part is devoted to explaining the data, variables, and method applied. Thereafter, the results are provided, based on which theoretical propositions are summarized. This is followed by a section discussing the results in the light of the extant empirical research and innovation literature. After providing practical implications, the last part concludes the paper, while also providing limitations and future research directions.
3. Study Design
The data for this study include the whole population, i.e., all academic workers from all four non-specialized public universities of Estonia (i.e., Tartu University, Tallinn University, Estonian University of Life Sciences, and Tallinn Technical University) from the year 2017. Two very small, specialized public universities (i.e., Estonian Academy of Music and Theatre, and Estonian Academy of Arts) focused on arts, music, and theatre studies were excluded from the analysis. The positions of the included universities, according to different QS rankings (e.g., EECA or subject rankings) [
50], are relatively high, thus the study focuses on advanced institutions.
The academic workers included had to be employed at the university with at least a 0.5 position. The latter guarantees that the individuals included are not occasional, part-time workers, as lecturing by practitioners at universities is common in Estonia [
51]. All academic positions, starting from the most junior and ending with professorships were included. The database of academic workers was merged with data from the Estonian Business Register (EBR), which enabled understanding when exactly each of the individuals founded a firm. In total, the database includes 1929 unique academic workers, out of whom 306 have become entrepreneurs at some stage of their career. The latter is defined as an individual becoming a company founder irrespective of ownership share, and is coded as FOUNDER. The latter results in a proportion of 15.9%, which is around twice as low when compared with the 33.2% value of the ratio of all firms and people in Estonia in the same age group as in this study in the year 2020 (calculated based on [
52,
53]). Thus, the entrepreneurial orientation in the academia portrayed through this specific commercialization type is much lower than in the general population. Still, the latter is not caused by difficulties in starting and running a business, as Estonia ranks among the top countries with respect to the ease of doing business (see the country rankings in [
54]), while the country also has an advanced innovation system [
55] and public universities are favoring spin-off entrepreneurship by creating the respective infrastructure (see e.g., [
56]). Other forms of commercialization tend to dominate spin-off entrepreneurship among Estonian academic workers (see e.g., [
57]), probably because running a firm is more time consuming. Nevertheless, since 2011 business creation has been made very easy in Estonia, as firms can be started without paid in capital and fully digitally [
58]. Partly due to this, 41% of firms originated between 2011 and 2015 in the dataset. All the respective firms are private SMEs, while the dominating size group is micro-firms. In addition, in the year 2017, the majority of the owners were still board members as well, thus being actively engaged in the daily management of their respective firms. The firm size and corporate governance setup in this study corresponded well to the overall Estonian population of firms (see e.g., [
59]).
Information about the academic performance of individuals was collected from the Estonian Research Information System (ERIS), where the CVs of academic workers are available based on certain universal fields. For each of the academic workers, five independent variables were coded (see
Table 1). Supervision experience (SUPEXP) is accounted as the number of supervisions of master’s theses and doctoral dissertations. As these require different amounts of effort, for summing up, doctoral dissertation supervision was weighted with four, i.e., their supervision was considered four times more laborious. In addition, as in the Estonian higher education system the Bologna approach (3 years bachelor and 2 years master’s studies, instead of the former 4 + 2 years) was introduced in 2002, the former master’s theses were weighted with two. Bachelor theses were not counted in the supervision experience, as information about them was not available in ERIS; moreover, in various curriculums there is no bachelor thesis.
Previous involvement in scientific and applied scientific projects (PROJECT) was accounted as the number of projects the academic worker has been involved in as a project head or one of the main executors. We summed these two project types because ERIS does not provide a full distinction between them and organically many projects include both domains as well. In addition, it was not possible to account for project budgets, as it would demand knowledge of how many people were factually involved in them, but in case of many projects the latter is not documented in ERIS.
Authorship of intellectual property (IP) was accounted as whether the academic worker has authorship of intellectual property objects (almost exclusively patents in the dataset). We did not account for the number of intellectual property objects, as only on rare occasions was the number more than one.
The number of high-quality publications (PUBLIC) was accounted as the number of articles in international peer-reviewed journals, chapters in monographs by international publishers, and Web of Science indexed conference proceedings (categories 1.1, 1.2, and 3.1 in ERIS). Solely written monographs were not counted, as this information has not been listed homogenously throughout the history of ERIS, and thus was not applicable. The last variable indicates the presence of a doctoral degree (PHD) for a specific academic worker.
As the aim of this paper is to study how academic assets are linked to firm creation at different career stages, four career stage groups were created. As these individuals have been constantly involved in academia, we applied biological age as a proxy of career length. The first considers academic workers up to 30 years old (STAGE_1), the second 31–40 years old (STAGE_2), the third 41–50 years old (STAGE_3), and the last all academic workers whose age is at least 51 years (STAGE_4). Roberts [
9] applied five-year steps in creating age categories, but in this study, it would lead to a too low frequency of entrepreneurs in each age group. Moreover, it is reasonable to apply wider categories, as in the academic scene a few years is a too short period for fundamental changes in academic assets to occur. However, we acknowledge that different grouping logics could be applied, and thus, a robustness test was provided with only two groups (≤40 years and >40 years, respectively). We controlled for career stage effects by creating separate groups, not by including relevant control variables in the equation, for the following reasons. First, the association of independent and dependent variables can be too sophisticated to be handled by control variables in the models. Second, when for each career stage a separate model is composed, the coefficients and significances of independent variables can be directly compared.
Table 2 documents the number of observations by career stage group and also by academic field. The latter categorization was added as, since science workers have a higher likelihood of being an academic entrepreneur, the created codified knowledge can be more easily commercialized, and scholars in that area have more links with firms [
13]. Similarly to Mõttus et al. [
13] the non-science field includes social sciences and arts and humanities. The number of observations (2314 individuals) was larger than the unique number of individuals (1929). This is because some individuals in the dataset passed through multiple age groups throughout their career. The earliest firm creations originated in the year 1995, and thus, we started to consider the academic characteristics from 1994. Therefore, the early career stages of older academic workers, falling into the Soviet era, are not the focus of this study.
For founders, the values of academic assets were collected from the year previous to founding the firm. In the case of academic workers having not founded a firm, the latter should be handled differently, as in their case there is no specific time available in order to position academic performance. As four career stages are used in the analysis, the most logical way is to account for each non-founder’s median value for each of the independent variables at the specific career stages. As the age of the academic worker at firm creation is random at each of the career stages, the median time of firm creation is always exactly in the middle of each of the career stages. This also lends support to the suitability of using the median value in case of non-founders.
For each of the four career stages, a separate logistic regression model was composed, with dependent and independent variables from
Table 1. After this the population of academic workers was divided into those working in the science field or not (i.e., social sciences and arts and humanities), and separate four logit models were composed for both of those populations. In addition, statistical tests were conducted to supplement the descriptive analysis available in the results section.
4. Empirical Analysis
4.1. Descriptive Statistics and Logistic Regression Models
Based on
Table 2 it can be concluded that the highest proportion of founders came from STAGE_2. At that stage, probably their position in academia has been secured, PhD defended, and first substantial results from research obtained. This finding is consistent with the results of Roberts [
9] that two thirds of academic entrepreneurs founded a new company at the age of 28–39, and in addition with Karlsson and Wigren [
39], who concluded that age is negatively correlated to starting firms. In the non-science field, firm creations occur to a certain extent earlier than in science. Namely, the share of companies created at STAGE_1 and STAGE_2 of total foundations were respectively 76.8% in the non-science, and 64.3% in the science, fields. In addition, the share of founders from all academic workers at STAGE_1 were 12.9% for non-science and 10.9% for science, while the latter had a higher overall proportion of firm founders (i.e., 14.1% vs. 11.6% in non-science).
Based on the results in
Table 3, academic founders at all career stages generally possess more academic assets (with the only exception being PhD degree possession at STAGE_2 and STAGE_4). Despite larger academic asset values for entrepreneurs, the statistical tests documented in
Table 3 indicate only a few statistically significant differences in academic asset values between founders and non-founders. Thus, a similar result was expected from the multivariate analysis. The same information as provided in
Table 3, but for the science and non-science fields separately, can be found in
Appendix A,
Table A1 and
Table A2.
Table 4 documents logistic regression models composed for the four career stages in the whole population, while science and non-science workers were viewed individually as well. By applying the significance threshold
p < 0.05, the significant variables by career stage are as follows (“+” indicating a rise in variable value increasing the likelihood of being a founder and “−“ otherwise): at STAGE_1 SUPEXP “+” and PUBLIC “+”, at STAGE_2 SUPEXP “+” and IP “+”, at STAGE_3 SUPEXP “+”, at STAGE_4 IP “+” and PHD “−“.
We conducted a robustness test by considering only two career stages, namely ≤40 years old (i.e., STAGE_A) and >40 years old (i.e., STAGE_B). At STAGE_A, the only significant (p < 0.05) variable was SUPEXP, in the earlier analysis being significant at STAGE_1 and STAGE_2, which summed as STAGE_A. At STAGE_B, the only significant (p < 0.05) variable was IP, while it was significant at STAGE_4, but at STAGE_3 its significance (0.123) slightly exceeded the largest reasonable p-value threshold of 0.1. The latter more consolidated results provide additional proof that pooling various career stages together can, to a certain extent, distort the stage-specific findings.
4.2. Theoretical Conceptualization
This paper extends the extant literature by providing an empirically validated integrative concept of resource-based and life-cycle theories in an academic entrepreneurship setting. Specifically, based on the empirical evidence we formulated the following five theoretical propositions of how academic assets and career stages are interlinked:
Time varying property (#1): The relationship between academic assets and firm creation by academic workers is non-linear, i.e., assets that are significant determinants at one career stage often might not be so in another. Thus, a career stage specific role of academic assets in firm creation is apparent.
Uniqueness property (#2): Only a few academic assets at specific career stages have a significant association with firm creation. Thus, at each career stage, only certain academic asset types portray the individual characteristics needed for starting a company.
Field-specific property (#3): The relationship explained in points 1 and 2 is academic field specific, i.e., substantial differences occur between science and non-science fields. In science field, a larger variety of academic assets tend to determine firm creation.
Scale property (#4): Academic assets are usually associated with firm creation by following a “more is better” logic, i.e., the larger the portfolio of a specific asset, the higher is the likelihood of firm creation.
Sophistication property (#5): At later career stages, somewhat more sophisticated assets are associated with launching a firm. Specifically, publishing and supervision, which are easier to achieve, do not play a role at later stages, where, for instance, the much more difficult patenting activity obtains a major role.
6. Practical Implications
The practical implications directly originate from the earlier theoretical conceptualization and empirical findings. Technology transfer offices at universities should ensure that incentives and mechanisms for academic entrepreneurship follow a life-cycle approach, i.e., the design of incentives should account for the specifics of different life-cycle stages (including academic assets and academic field). The latter is in line with Perkmann et al. [
2] and Hossinger et al. [
11], who concluded that policies and incentives should be tailored to support academic workers’ entrepreneurial activities.
Generally, at earlier stages academic workers who are more productive in simple academic tasks (e.g., supervision or publishing) seem to be more likely candidates for firm creation, while at later stages, more specific assets such as the presence of intellectual property are the cornerstones for starting a company. Potentially overlooked are the older academic workers without a PhD, who might have enough accumulated knowledge for commercialization.
Statistically speaking (see also
Table 2), the field of science seems to offer more commercialization opportunities, except for in the earliest career stage. However, the latter postulate is based on actual longitudinal information about firm creations, and thus, might not disclose unexploited potential. Namely, the population of academic workers with similar characteristics to firm founders, but having not achieved this, is remarkable. The second career stage, which is characterized by workers having obtained a PhD and moved above basic academic positions, seems to be the best springboard for firm creation. Such individuals have not reached a breakpoint in their career, where too many academic activities might start hindering entrepreneurial opportunities.
7. Conclusions
The paper aimed to find out which academic assets associate with firm creation at different academic career stages; while the extant literature has mostly remained silent about the potential dynamics of entrepreneurship determinants. A study design integrating different theoretical perspectives was implemented, while the dataset included the longitudinal whole population of Estonian academic workers. Logistic regression was implemented to test which academic assets were associated with business creation by academic staff at four career stages, determined by an individual’s age. Additional models were composed for the two main academic disciplines.
We conceptualized the empirical findings into theoretical propositions reflecting the time varying, uniqueness, field-specific, scale, and sophistication properties of the interconnection of academic assets and academic entrepreneurship (specifically, firm creation). The results indicated that at different academic career stages, the academic assets associated with firm creation differ, leading to the time variation context. However, at each stage, the frequency of significant determinants was low, and moreover, they tended to be unique. Namely, while at earlier stages more general assets (e.g., supervision experience, number of publications) associate with company creation, at later stages they become more sophisticated (e.g., possession of intellectual property). The associations between the two phenomena are generally positive, meaning that a larger scale of assets leads to a higher likelihood of becoming an entrepreneur. The results varied between two academic disciplines, i.e., science and non-science fields, thus providing a field-specific context. In addition, the findings include some statistically significant relationships between academic entrepreneurship and academic assets that static studies have not found before. For instance, grants (only for non-science field) and PhD status decrease the likelihood of firm creation at the last stage of academic careers.
As a major avenue of further research, the concept provided in the paper could be further developed and tested. The few limitations, also giving rise to ideas for future studies, are discussed as follows. First, follow-up research could resolve the limitation of this study that only one proxy for each academic asset type is used. The latter limitation originates from using factual population level information from a database, i.e., such sources are not as rich as interviews or questionnaires. Second, the context of the founded firms could be elaborated further, as there could be a high level of diversity among them [
60,
61,
62]: e.g., by evaluating for their financial performance, exact technology transfer mechanism from academia, and corporate governance (i.e., roles, background and relationship between managers and owners). Third, the population applied in this study could be even futher enlarged by counting those academic workers who have permanently left academia to be entrepreneurs. Last, as the paper is directed to studying only the association between academic assets and firm creation, future studies could focus more on causality questions: namely, if and how these assets exactly contribute to knowledge commercialization through firm creation.