University technology transfer: leveraging experiential learning and TTO’s resources

We examine how experiential learning from both previous technology transfer experience and the disclosure of discoveries impact universi-ties’ technology transfer outcomes, in terms of the number of spinoffs created and licenses granted, while acknowledging that TTOs provide specialized resources that support technology transfer processes. By using panel data models on a sample of public Spanish universities for the 2006–2011 period, our model introduces discovery disclosures as an instrument to account for the endogenous nature of the scientific research process. Results show that spinoff creation and license granting depend on their own experiential learning and on the number of discovery disclosures, which in turn depends on its own experiential learning. Technology transfer is influenced by two learning processes connected to technology transfer outcomes and discovery disclosures. Also, the effect of TTOs’ specific resources appears as not significant. In the context of Spanish universities, the findings support the view that, compared to TTOs’ staff specialization, accumulated knowledge from their own experience adds more value for generating technology transfer outcomes.

Plain English Summary The involvement of scientists, together with the resources deployed by technology transfer offices (TTOs), are crucial for the transformation of scientific knowledge and technology into commercialized products and services.The article explores how two distinct, interconnected learning curves related to discovery disclosures and technology transfer outcomes (i.e., spinoffs and licenses) explain the commercialization of new knowledge generated within Spanish universities.The findings support the notion that scientific research processes can benefit from experiential learning.Past experience with both discovery disclosures and technology transfer outcomes was found to positively impact universities' capacity to create spinoffs and grant licenses.On contrary, we did not detect that TTOs' staff specialization affects the generation of technology transfer outcomes.These results are relevant to university managers and social planners.This study provides a template for future work interested in enriching our understanding of both barriers and enables to technology transfer outcomes.

Introduction
Academics and policy makers increasingly acknowledge that universities-in their position within entrepreneurial ecosystems-have the potential to contribute to economic development by channeling different forms of knowledge to the society (e.g., Breznitz & Zhang, 2019;Cunningham et al., 2019;Lockett & Wright, 2005).Rooted in the dominant narrative of the entrepreneurial ecosystem literature which positions universities as key actors of this ecosystem (e.g., Brown & Mason, 2017;Cho et al., 2022), the primary function of universities in contemporary societies surpasses purely academic tasks and revolves around knowledge generation and technology transfer (e.g., Etzkowitz et al., 2000;Lafuente & Berbegal-Mirabent, 2019;Lee & Jung, 2021).Ideally, this knowledge-led function materializes in the channeling of new knowledge and technologies to local markets through different outcomes-e.g., spinoffs and licenses-which, in turn, facilitates knowledge spillovers and contributes to generate economic surplus (Aldridge & Audretsch, 2011;Brown & Mason, 2017;Caldera & Debande, 2010;Doblinger et al., 2019;O'Shea et al., 2005;Wagner et al., 2021).
The increased awareness of the importance of commercializing technology transfer outcomes has stimulated universities to create technology transfer offices (henceforth TTOs) that act as knowledge brokers connecting scientists and organizations (Fini et al., 2017;Perkmann et al., 2013).TTOs equip scientists with specific resources and services for technology transfer purposes (Huyghe et al., 2014;Lafuente & Berbegal-Mirabent, 2019;Lee & Jung, 2021;Macho-Stadler et al., 2007).Because of the strategic and economic relevance of universities' technology transfer function, policy makers have progressively encouraged university-industry collaborations in order to exploit research outcomes (Breznitz & Zhang, 2019;Cho et al., 2022;European Commission, 2022).In this sense, TTOs are identified as relevant intermediaries that help cope with potential tensions between academic and commercial demands (e.g., Aldridge & Audretsch, 2011;Lafuente & Berbegal-Mirabent, 2019).
Existing work dealing with technology transfer processes has primarily analyzed how universities support various pathways to technology transfer, and how TTOs commercialize the outcomes of these processes (e.g., Friedman & Silberman, 2003;Hsu et al., 2015;Lafuente & Berbegal-Mirabent, 2019;Lee & Jung, 2021;Macho-Stadler et al., 2007;Siegel et al., 2007).Nevertheless, underlying the design of most studies within this literature stream is the assumption that TTOs' knowledge base, which results from their technology transfer activities, remains unaltered over time.From an organizational viewpoint, TTOs universities are heterogeneous in terms of both available resources and the processes supporting technology transfer.It is therefore plausible to argue that these marked differences condition technology transfer outcomes.As part of the routines documented for many types of organizations (e.g., Argote et al., 2021), past experience in the form of accumulated knowledge will therefore affect the generation and commercialization of technology transfer outcomes.In the context of this study, the specific analysis of the role of learning as a trigger of technology transfer activities has been sidelined in previous studies.
This research focuses on how two distinct (and interrelated) forms of learning affect technology transfer outcomes, in terms of spinoffs and licenses, while acknowledging that TTOs are instrumental units providing specialized resources that support this process.Concretely, we evaluate the relationship between accumulated experience with spinoffs and licenses and the generation of technology transfer outcomes.In our model, discovery disclosures are considered the outcome of an endogenous process driven by its own experiential learning process.Furthermore, we evaluate whether TTOs' resources support the learning process that occurs within universities.
The empirical exercise uses a sample of Spanish public universities between 2006 and 2011.This setting is attractive because, similar to other European countries, Spain has undergone significant reforms in the university's regulatory framework seeking to encourage technology transfer (Berbegal-Mirabent et al., 2019;Caldera & Debande, 2010;Lafuente & Berbegal-Mirabent, 2019).The Spanish context offers the opportunity to analyze how universities capitalize on their own resources to produce technology transfer outcomes in a scenario where two types of experiential learning (i.e., discovery disclosure and technology transfer outcomes) have the potential to enhance universities' current technology transfer activity.TTOs are key intermediary units leading technology transfer processes (Lafuente & Berbegal-Mirabent, 2019;Macho-Stadler et al., 2007), and we have defined two variables linked to relevant TTO outcomes, namely spinoffs and licenses.Following the learning tradition, we have modeled experiential learning as the accumulated experience with past technology transfer outcomes (see, e.g., the comprehensive survey by Argote et al. (2021)).
In our model, discovery disclosures play a critical role in the technology transfer process.Because scientists lead the disclosure of discoveries within universities (e.g., Caldera & Debande, 2010;Lafuente & Berbegal-Mirabent, 2019), this process is modeled as a function of accumulated experience with past discovery disclosures.
Our research contributes to both the technology transfer and the organizational learning literature.Most research on learning has been carried out in the context of manufacturing firms (see, Argote et al., 2021), while few studies have examined learning curves in knowledge-intensive industries (see, e.g., Fong Boh et al. (2007) in software development firms, and Jain (2013) in biotech firms).Additionally, by proposing a model in which two learning curves coexist, the proposed study of universities' learning offers the opportunity to further understanding how different sources of experiential learning-linked to scientific research and technology transfer-that simultaneously occur within universities, together with TTOs' specific resources, impact relevant technology transfer outcomes in these public knowledgebased organizations.
From the technology transfer perspective, most studies assume that TTOs are equipped with homogeneous resources to develop their tasks.But, the mixed results reported in the literature only reinforce the notion that TTOs are heterogeneous in terms of both resources and capabilities (see, e.g., Gomez Gras et al., 2008;Caldera & Debande, 2010;Lafuente & Berbegal-Mirabent, 2019;Lee & Jung, 2021;Sansone et al., 2021;Soares & Torkomian, 2021).Learning effects and resource specialization have been overlooked in the literature.In this sense, by presenting an analysis through the lens of the organizational learning theory, our study offers insights into how universities can simultaneously benefit from both different learning processes and TTOs' specific resources in order to increase their technology transfer outcomes.
The study proceeds as follows.Section 2 reviews the literature and presents the study hypotheses.Section 3 presents the data and method.Section 4 outlines the empirical findings, while Sect. 5 offers the discussion, implications, and future research avenues.

Hypotheses development
Building on postulates of the organizational learning theory (Levitt & March, 1988;March, 1991), organizations create, accumulate, and disseminate knowledge that they acquire from experience.The resulting knowledge stock will likely enhance business performance by altering and improving existing practices or developing new organizational routines.
Learning begins with experience, and this experiential process is at the heart of exploitative learning described by March (1991).Exploitation is largely defined as a knowledge refinement process resulting from the systematic repetition of routines and tasks.The continuous exploitation of accumulated knowledge and experience is expected to increase learning and, subsequently, business performance (Argote et al., 2021).This phenomenon is referred to as learning curve or experience curve (Levitt & March, 1988;March, 1991).Initial evidence of learning curves was found in manufacturing (Epple et al., 1996) and new product development (Gopal et al., 2013).Recently, empirical evidence of learning curves is also reported in knowledge-intensive industries like software development (Fong Boh et al., 2007), biotech (Jain, 2013), the industry 4.0 (Tortorella et al., 2022), or sustainability of supply chains (Kalantary & Farzipoor Saen, 2022).
This study focuses on the role of experiential learning on technology transfer processes.Technology transfer is defined as the commercialization of intellectual property created in universities and government laboratories, and discovery disclosures obtained from scientific research are considered the triggers of this process (Feldman et al., 2002;Harmon et al., 1997;Siegel et al., 2007).A discovery disclosure is a confidential document issued with the aim to determine if patent protection applies for the scientific research result.After a technological development phase, universities give firms the right to exploit the invention through licenses.If the firm receiving the license is a new venture to which the university participates through equity, it is called a spinoff.Universities may prefer to create spinoffs rather to license to incumbent firms because it allows them to better structure deals, the possibility of holding valuable assets if technology transfer is not successful and the shorter time to get revenues compared with licensing (Conti & Gaule, 2011).
The technology transfer process described above in a simplified version might also produce organizational benefits in the form of knowledge-driven experiential learning.A well-documented phenomenon in industrial organization is the enhanced performance resulting from the improvement of internal processes and routines as firms acquire experience (Argote et al., 2021).
Like other technology-intensive activities, scientific research is prone to benefit from learning curve.Medical practice, a field of knowledge sharing many practices with scientific research, has procedures that benefit from the learning curve.Like medical teams, research teams can acquire experiential learning through teamwork and methodological standardization of practices (e.g., Kreiker et al., 2004;Li et al., 2009;Pisano et al., 2001).Researchers bring research results to TTO offices, which evaluate potential for innovation and eventually recommend issuing discovery disclosures (Jensen et al., 2003).This evaluation process is also subject to experiential learning, as TTO officers will tend to push forward projects from faculty with academic prominence (Hsu & Kuhn, 2023).As a consequence, we hypothesize that:

H1. Accumulated experience with discovery disclosures positively affects the number of discovery disclosures of universities
In the specific context of this study, spinoff creation and licensing are two distinct outcomes of technology transfer, which require different organizational capabilities (Caldera & Debande, 2010;Feldman et al., 2002).On the one hand, the drafting of license agreements needs specific knowledge on property rights as the objective of this process is granting a business the rights to exploit a specific invention developed by the university.Furthermore, TTOs not only provide expertise on property rights, but they also assess universities' discoveries in order to determine their potential and marketability.In return, the university obtains royalties that depend on the success of the commercialized technology (Feldman et al., 2002).
Spinoffs and licenses are knowledge-intensive activities, and previous empirical evidence supports the presence of learning curves in spinoff creation and licensing processes among firms to which the technology is transferred (Daghfous, 2004).Instead of focusing on the technology's beneficiary, the core of this study is to analyze whether the technology's providers (in this case, universities) also benefit from learning curves as their technology transfer process evolves.
In the specific context of this research, each outcome of the technology transfer process relies in its own knowledge stock.Following mainstream learning literature (see, e.g., Argote et al., 2021), we therefore propose that spinoff creation process and licensing will each have its own learning curve, which is determined by the university's own accumulated experience with these technology transfer outcomes.Therefore, we hypothesize the following: H2a.Accumulated experience with previous spinoffs positively affects the number of spinoffs created by universities.H2b.Accumulated experience with previous licenses positively affects the number of licenses granted by universities.
As discussed previously, the starting point of technology transfer is a result of scientific research that can have potential of innovation and therefore worth of issuing a discovery disclosure document.This is connatural with technology transfer that allows universities to specialize in scientific research, while firms specialize in commercializing discoveries.Technology transfer goes from universities to firms most of the time, and rarely the other way around (Harmon et al., 1997).In a recent paper, Lee and Jung (2021) have found empirically that academic and applied research are the most relevant inputs for technology transfer, and di Gregorio and Shane (2003) found that intellectual eminence of the university is an antecedent of licensing.We consider that the number of discovery disclosures is a common antecedent of both spinoff and licensing, as the access to an invention with commercial potential is previous to the decision of creating a spinoff or licensing to an existing firm.Therefore, we posit that: H3a.The number of discoveries disclosed positively affects the number of spinoffs created by universities.
H3b.The number of discoveries disclosed positively affects the number of licenses granted by universities.
While academics are concerned with creating new scientific knowledge, public administrations are increasingly encouraging universities to deploy specific resources for effectively carrying out technology transfer outcomes.For most universities, knowledge transfer tasks rely on TTOs, whose primary role is to channel new knowledge created by faculty, that is, discoveries, to the market through different outcomes, in our case spinoffs and licenses.The business development expertise and ability to attract venture capital required for spinoff creation are common to technology transfer projects, irrespective of the underlying technology (Lockett & Wright, 2005).Regarding licensing, TTOs take advantage of pooling innovation across university units (Lafuente & Berbegal-Mirabent, 2019) for an effective licensing process (Macho-Stadler et al, 2007).This way, researchers can delegate licensing activities to TTO staff.As these activities are heterogeneous, TTOs will need to dedicate staff to each activity separately.
As a result of their activity within universities, TTOs become the locus of learning from technology transfer processes.As this learning can take place at individual and unit level (Argote, 1993), the presence of a large proportion of TTO's specialized staff can moderate the learning curve for each activity.
Additionally, teams' specific factors related to job stability might prove effective in improving the propensity to learn (Edmondson, 2002).Specifically, teams developing their tasks in environments with more stable work conditions might increase members' self-reflection and willingness to learn from their own past experience (KC et al., 2013).This is particularly the case of public universities.
Arguably, the psychological safety that characterizes the work environment of TTOs affiliated to public universities facilitates the creation of a culture that encourages workers' effort and team cohesiveness (Edmondson & Nembhard, 2009).On the basis that technology transfer might be seen as a form of product innovation, the intra-group properties described above might support risk tolerance and, consequently, learning processes that enhance the development of future technology transfer outcomes.From these arguments and evidence, we hypothesize the following: H4a.The proportion of TTO staff dedicated to support business creation processes is positively associated with the number of spinoffs created by universities.H4b.The proportion of TTO staff dedicated to create and manage license agreements is positively associated with the number of licenses granted by universities.H5a.The proportion of TTO staff dedicated to support spinoff creation processes positively moderates the relationship between the number of spinoffs created and accumulated experience with previous spinoffs.H5b.The proportion of TTO staff dedicated to create and manage license agreements positively moderates the relationship between the number of licenses granted and accumulated experience with previous licenses.
The logic supporting our theoretical framework is graphically presented in Fig. 1.

Data
To examine the antecedents of technology transfer in universities, we have gathered data about technology transfer in Spanish universities from two sources.First, profile details as well as information on the faculty working in Spanish universities were collected from the reports made available by the Spanish Association of University Rectors (Conferencia de Rectores de Universidades Españolas, CRUE).Second, all information on the results of universities' technology transfer and TTOs' resources and structure was obtained from the annual reports generated by the Network of Spanish Technology Transfer Offices (RedOTRI).The dataset covered from year 2006 to 2011 and includes all 47 Spanish public (state-owned) universities.For the sake of methodological rigor, we have removed three universities from the dataset (Universidad de las Palmas de Gran Canaria, Universidad de León and Universidad Politécnica de Cartagena) as information about technology transfer outputs was unreliable and incomplete.Therefore, the final dataset includes 44 universities (254 university-year observations).

Variables
In this study, the dependent variable is technology transfer performance which is measured as the number of spinoffs and licenses reported by each university (i) during the studied period (t = 2006,…, 2011).
Concerning the independent variables, we use accumulated experience with past licenses and spinoffs to proxy organizational learning.In line with previous research, for each university and year, the technology transfer values reported in previous years are discounted by a factor equal to the square root of the years passed since the experience was acquired (Argote et al., 1990;Argote, 1993), that is: The discount factor ( ) accounts for knowledge depreciation coming from organizational forgetting or from the mismatch between the accumulated knowledge and the current environment or existing business practices.To compute the accumulated experience values for the first year of the data series (i.e., 2006), we used the number of spinoffs and licenses created since 2004 (i.e., 2004 and 2005).
The values of spinoffs and licenses and their respective accumulated experiences allow us to obtain preliminary evidence of learning curves.In Fig. 2, we present the average number of spinoffs and licenses across universities from 2006 to 2011.We observe that both variables grow monotonously with time.Then, we have plotted the value of the variable versus its accumulated experience for the first and last year of the series.For both spinoffs (Fig. 3) and licenses (Fig. 4), we observe that while in the first year of the series the value of the variable can be larger or smaller than its accumulated experience, in the last year, the accumulated experience is always larger than the actual value.
The second set of independent variables deal with the output of scientific research.On the one hand, scientific research is proxied by the number of discovery disclosures reported by universities during the analyzed period.On the other hand, similar to the variables linked to past experience with technology transfer outcomes, accumulated experience with past discovery disclosures is computed as ∑ The third set of independent variables includes TTOs' resources.The effect of TTO resources on each outcome is measured via the proportion of TTO members working on spinoff creation and licenses in each university and year.Our model also adds a term accounting for the interaction between accumulated experience and proportion of TTO members working on each outcome.Finally, we have added control variables that account for organizational age and size, two factors that can influence technology transfer processes.Age is expressed in years for the university and the TTO, while size is measured by the number of faculty and TTO staff.The four control variables are defined for university and year.Notice that all dependent and independent variables are transformed logarithmically to account for skewness.
Table 1 presents the descriptive statistics and bivariate correlations for the study variables.First, we observe that the correlation between spinoffs and licenses, although positive and significant, is low.This indicates that these two variables account for two distinct forms of technology transfer.Second, we also observe strong positive correlations between spinoffs, licenses, and discovery disclosures and each of its accumulated experience variables.These results show preliminary evidence of the importance of learning curves in technology transfer and scientific research.

Statistical approach
Our goal is to estimate how cumulative experience with technology transfer outcomes-i.e., spinoffs and licenses-and TTO resources impact university's capacity to generate technology transfer outcomes.To this end, we propose the following full model: In Eq. (1), j are parameter estimates for the set of independent variables (j); the university control variables include size (ln faculty) and age (ln years), while size (ln staff) and age (ln years) are the TTO controls.Also, T refers to the set of time dummy variables that rule out potential effects of time on universities' technology transfer outcomes that are common to all higher education institutions.The term i is the timeinvariant effect that controls for unobserved heterogeneity across universities (i), and it is the normally distributed error term that varies cross-universities and cross-time.
Empirically, our approach assumes that scientists trigger the disclosure of discoveries within universities, which implies that discovery disclosures are the result of an endogenous process.To deal with this endogeneity issue, in Eq. ( 1), we use the accumulated experience with prior discovery disclosures as an instrument to model the endogenous nature of scientific research, that is: The proposed instrumentalized modeling allows to test hypothesis H1.
For the rest of hypotheses, we expect that  1 > 0 to verify that previous experience with the studied outcomes is positively correlated with the level of spinoffs (H2a) and licenses (H2b).Similarly, we expect that  2 > 0 to confirm that discovery disclosures are positively related to technology transfer outcomes (H3a, spinoffs; H3b, licenses).A positive result for the coefficient linked to TTO resources ( 3 > 0) will verify that TTOs' staff specialization positively affects the level of spinoffs (H4a) and licenses (H4b).Finally, we expect that  13 > 0 to confirm that TTOs' staff specialization positively moderates the relationship between accumulated experience with previous technology transfer outcomes and the level of spinoffs (H5a) and licenses (H5b).
Our model specification systematically controls for university-specific effects, thus absorbing observed and unobserved sources of heterogeneity that are constant over time.Moreover, we control for a number of timevarying characteristics that capture several organizational differences across universities.Nevertheless, universities differ in ways that are not directly observable to us.For example, prior work suggests that the management of discoveries and their disclosure represent the starting point of generation of technology transfer outcomes (Harmon et al., 1997;Hsu et al., 2015).As it was highlighted in the theory section, the disclosure of discoveries is a necessary step to engage in the technology transfer process by securing the new knowledge produced by scientists which will serve as the roadmap of any potential technology transfer outcome (in our case, spinoff firms or license contracts).
For estimation purposes, Eq. ( 1) is computed using a two-stage least square regression (G2SLS).In order to verify the appropriateness of our modeling strategy, the underidentification test (Kleibergen & Paap, 2006) is used to test if the chosen instrument is relevant, whereas the overidentification test (Cragg & Donald, 1993) will corroborate if the study's instrument is orthogonal to the disturbance term.

Results
This section presents the empirical results.The findings for the models evaluating the relationship between learning and the analyzed technology transfer outputs are presented in Table 2.
Before reporting the results, we carried out the underidentification and overidentification tests to validate the appropriateness of the instrumental variable approach used in this paper.The null hypothesis of the underidentification test (Kleibergen & Paap, 2006) evaluates if the instrumental variable is irrelevant in the second stage model, whereas the null hypothesis of the overidentification test (Cragg & Donald, 1993) states that the instrumental variables are orthogonal to the residuals of the second stage model.The results for the underidentification test indicate that the null hypothesis can be rejected for the baseline and the full model.For the overidentification test, findings show the null hypothesis of overidentification cannot be rejected, thus validating the chosen instrument.
To account for possible multicollinearity effects, we have examined the variance inflation factors (VIF) for the baseline (without interaction) and full (with interaction) models for spinoffs and licenses.For spinoffs, the mean VIF for the baseline model is 1.84, with a range between 1.26 and 2.83.For the full model, mean VIF is 2.40 (range = 1.56-4.95).For licenses, in the baseline model, the mean VIF is 1.83 (range = 1.18-2.65)and in the full model is 2. 77 (range = 1.61-6.95).Given the values of VIF, the collinearity between independent variables does not inflate the variance of estimators significantly.
Concerning the coefficients for the key study variables, for both technology transfer outcomes (i.e., spinoffs and licenses), the findings strongly support the positive and significant effect over the disclosure of discoveries of accumulated experience with discovery disclosures (Table 2).In our interpretation, this result evidences that the disclosure of scientific discoveries benefits from a learning curve driven by experience.Therefore, this finding confirms our hypothesis H1.
The results in Table 2 show that experiential learning with previous technology transfer outcomes is a decisive factor driving the generation of both spinoffs and license contracts.These results are in line with our second hypothesis that states a positive relationship between accumulated experience with technology transfer outcomes and the number of spinoffs created (H2a) and licenses granted (H2b).The coefficients for the relationship between discovery disclosures and both technology transfer outcomes are positive and significant.This therefore supports our third hypothesis stating a positive relationship between discovery disclosures and spinoff creation (H3a) and the number of licenses granted (H3b).
Looking at the findings for TTOs' staff specialization, the coefficients for the variables linked to specific resources are not statistically significant.This result is consistent for both technology transfer outcomes, thus failing to support our fourth hypothesis which proposes a positive relationship between TTOs' staff specialization and the number of spinoffs created (H4a) and the number of licenses granted (H4b).We also observe that the coefficients of the interaction terms between accumulated experience with previous technology transfer outputs and TTOs' specialization are not significant.Consequently, hypothesis H5a, which proposes that the proportion of TTO staff dedicated to spinoffs positively moderates the relationship between spinoff creation and accumulated experience with previous spinoffs, and hypothesis H5b, which states that that the proportion of TTO staff dedicated to manage intellectual property rights and licenses positively moderates the relationship between licenses and accumulated experience with previous license agreements, are not supported.
To further asses the validity of our results, we performed an additional robustness analysis focused on the potential effect of TTOs' scalability on the generation of technology transfer outputs.To achieve this, we added to the regression model (Eq.( 1)) an interaction term between accumulated experience with previous technology transfer outputs and TTO size.Results presented in Table 3 of the Appendix indicate that TTO size is not a factor that directly or indirectly drives the generation of technology transfer outputs.For all model specifications, the parameters of experiential learning remain positive and statistically significant, whereas the coefficients for the interaction terms between experiential learning and TTO size are not significant (see The contribution of this study is not limited to the analysis of universities' learning patterns in a model that incorporates two learning curves that simultaneously occur within these publicly funded institutions.In today's economic and social landscape, universities are much more than graduate factories.By adding technology transfer to their objective function, universities have become relevant actors supporting the development of entrepreneurial ecosystems by deepening the ecosystem's technology base and enhancing markets' capacity as the breeding ground for entrepreneurial action (Cunningham et al., 2019;Lafuente et al., 2020;O'Shea et al., 2005;Wagner et al., 2021).
This study therefore contributes to better understand how different learning processes impact the functioning of TTOs, with the objective to provide insights that help universities to continue to incubate and channel their technology transfer outputs to the local economy which, in turn, contribute to propel the entrepreneurial ecosystem primarily through the expansion and accumulation of knowledge at the local level (Aldridge & Audretsch, 2011;Cho et al., 2022;Lafuente & Berbegal-Mirabent, 2019;Perkmann et al., 2013).
Prior work often assumes that the specific knowledge linked to technology transfer embedded within universities, and by definition TTOs, remains unchanged over time.For the analyzed universities, our findings showing that two learning curves coexist are in line with the notion that scientific discovery triggers technology transfer (Aldridge & Audretsch, 2011;Harmon et al., 1997;Lafuente & Berbegal-Mirabent, 2019).These results suggest that successful technology transfer involves TTOs that actively engage in mediating university-industry collaborations as well as the commercialization of scientific knowledge.In this sense, the design of policies that align scientists' interests (often closer to research activities) and the market-orientation of TTOs is a critical strategic action that universities should adopt if the generation of economically meaningful technology transfer is the desired goal.
On contrary, it was found that more specialized TTO-i.e., proportion of TTO staff with specific knowledge related to each analyzed outcome-is not instrumental in explaining technology transfer outcomes among Spanish universities.In our view, this finding might be the empirical manifestation of other unobserved effects not evaluated in this study.Specifically, prior work studying the relationship between the profile of TTO staff and technology transfer outcomes underlines that TTO supporting employees' specialization are in a better position to overcome potential barriers emerging from the conflicting objectives of TTOs and scientists (Conti & Gaule, 2011;Goble et al., 2017;Soares & Torkomian, 2021).
Building on the logic underlying these studies, our results suggest that the background of TTO employees can be outdated or is not fully exploited by these units.These problems might translate into information asymmetries and communication problems between the TTO, academic scientists, and industry agents.The inability of TTO employees to transform the (sometimes abstract) knowledge of scientific language into marketable projects is an example of such problems.Obviously, employees' specialization and continuous updating constitutes a relevant strategic avenue for TTO managers interested in the effective exploitation of scientific ideas while minimizing potential contingencies that may arise during the technology transfer process.
These actions may constitute a source of competitive advantage for TTOs in their attempts to strive for superior performance.

Implications
The findings of this study offer relevant implications for scholars and policy makers.First, TTOs engage in different strategic actions to successfully commercialize scientific knowledge.From an academic viewpoint, the results highlighting that two learning curves simultaneously shape universities' technology transfer process fuel the scholarly debate on both the multidimensionality of TTOs' objective function (Lafuente & Berbegal-Mirabent, 2019;Olaya-Escobar et al., 2021;Siegel et al., 2007).Additionally, the results emphasize the need to match TTOs' goals with those of the different stakeholders that participate in technology transfer processes, for example, scientists and industry representatives (Cunningham et al., 2019;Lee & Jung, 2021).
Second, our core findings have implications for the learning literature.Specifically, by distinguishing different learning mechanisms that energize the commercialization of scientific knowledge, our analysis has produced novel evidence on how two learning curves-linked to past experience with both discovery disclosures and technology transfer outputs-work together to boost universities' technology transfer performance.In a context where experiential learning drives technology transfer outcomes, universities can enhance their learning process from past experience with scientific research by encouraging the development of projects that, by supporting research on existing or new fields as well as the creation or consolidation of research teams, add valuable knowledge to scientists and TTO professionals (Argote et al., 2021).
Additionally, organizations learn from failures as much as from successes (Bennett & Snyder, 2017;Madsen & Desai, 2010).Therefore, instead of focusing on successful experiences only, we suggest to university and TTO managers to turn their attention to the specific characteristics of all inventions, in order to draw learning lessons from potentially valuable inventions that failed to be commercialized.Third, by definition, TTOs are innovation vehicles, and the successful commercialization of inventions requires specific investments.The results of our study underline the value of internal analyses.By conducting a profound analysis of available resources, TTO managers will be in a better position for promoting a more meaningful (in terms of valued added) strategy making of TTOs.Concretely, we reported a null effect of the level of specialization of TTOs' staff on TTOs' outcomes.This finding suggests that, among TTOs affiliated to Spanish public universities, the creation of efficient organizational structures and the recruiting of skilled staff with specific capabilities that support the transfer of new knowledge to the industry should be prioritized by TTO managers in order to improve both internal learning processes and the evaluation of the potential market value of inventions (Lee & Jung, 2021;Soares & Torkomian, 2021).

Limitations and further research
As with any study, our research has some limitations which, in turn, can provide relevant avenues for further research.First, it should be noted that in the Spanish context research funding purposely devoted to technology transfer activities primarily comes from government agencies or public-private collaboration agreements (e.g., Kantis et al., 2023;Lafuente & Berbegal-Mirabent, 2019).In this sense, future work should examine whether the allocation of university slack resources impacts technology transfer, and whether the different sources of funding condition the learning capacity of scientists and TTOs for generating technology transfer outcomes.Second, technology transfer is a complex phenomenon whose processes are heavily reliant on other factors related to the university and the local environment, including, for example, the university's entrepreneurial culture (e.g., Herrera-Valverde et al., 2020;Miller & Acs, 2017;O'Shea et al., 2005), the access to research facilities (i.e., science parks and accelerators) (e.g., Breznitz & Zhang, 2019;Fini et al., 2017;Rothaermel et al., 2007), and the entrepreneurial ecosystem where the university is located (e.g., Doblinger et al., 2019;Lafuente-González & Leiva, 2022;Meoli et al., 2019;Villanueva & Martins, 2022).Further research on the universities' technology transfer function should integrate these variables into the analysis in an effort to identify potential complementarities (or substitutability effects) with experiential learning from past technology transfer processes.
Finally, research on learning in scientific research should focus on the study of how learning takes place at the unit and team level.Specifically, future work should track over time the outcomes and progress of research teams and labs with the objective of better grasping how experiential learning at different organizational levels (i.e., individual, team, and TTO) explains the generation of technology transfer outcomes within universities and other research-oriented institutions.Appendix Table 3 ) .As we indicate below, keep in mind that the latter term enters in the final model as an instrumental variable.

+ 4 Fig. 2
Fig. 2 Average number of spinoffs (left) and licenses (right) across Spanish universities for the 2006-2011 period

Fig. 3
Fig. 3 Evolution of number of spinoffs created as a function of the accumulated number of spinoffs created in Spanish universities for 2006 (left) and 2011 (right)

Fig. 4
Fig. 4 Evolution of number of licenses created as a function of the accumulated number of licenses created in Spanish universities for 2006 (left) and 2011 (right)

Table 1
Descriptive statistics and bivariate correlations for the study variables For the bivariate correlations, absolute values below 0.1004 are not significant, values between 0.1005 and 0.1208 are significant at 10% level, values between 0.1209 and 0.1559 are significant at 5% level, and values greater than 0.1560 are significant at 1% level Vol.: (0123456789)