POLITECNICO DI TORINO

Master of Science in Management Engineering

M.Sc. thesis work

Team composition influence on decision- making process: scientific and effectuative approaches

Supervisor: Prof. Emilio Paolucci Co-superv. Daniele Battaglia

Co-superv. Andrea Panelli Candidate: Davide Campo

March - April 2021 session

“Fortuna audaces iuvat”

1

Content

1. The Decision-Making process ...... 8 1.1 Innovation and entrepreneurship overview ...... 9 1.2 Scientific Method ...... 11 1.3 Heuristic Search & Effectuation ...... 17 1.4 Team heterogeneity and other effects ...... 19 1.4.1 The role of network ...... 21 1.4.2 Psychological dimensions ...... 22 1.4.3 Industry effect ...... 24 1.4.4 Gender effect ...... 25 1.4.5 Academic level and background effect ...... 27 1.4.6 Prior experiences weight ...... 28 1.4.7 Stevens’ scales Theory ...... 29 1.4.8 Blau’s index ...... 30 1.4.9 Research questions ...... 31 2. The research program and its scope ...... 38 2.1 Sampling and design experiment ...... 40 2.2 Why RCT design ...... 41 2.3 Gauss-Markov Theorem: the OLS efficiency ...... 43 2.4 Marketing and sponsorship ...... 44 2.5 The data collection ...... 46 3. Sample analysis and preliminary observations ...... 52 3.1 Team level aggregation and database creation ...... 60 4. Results analysis ...... 67 4.1 Preliminary evaluations ...... 69 4.2 Regression presentation ...... 71 4.3 Robustness check ...... 82 4.4 Final conclusions ...... 85 APPENDIX ...... 93 References ...... 103

2

3

Abstract

A number of scholars agree that process fostered by a good task conflict entrepreneurial activities lead implicit management, meant as the concept high risks and therefore would deserve a introduced by Jehn et al. (1997, 1999) in deep study (Forlani & Mullins, 2000; Ping, opposition to the relationship conflict 2004; Eisenmann et al., 2013). A crucial management. Right such last point arises issue which must be faced nowadays is the from the evidence about control variable high failure rate and strong difficulty in called as Internal Network, which has expansion by scaling the business. given evidence that preexisting ties among According to US Census Bureau, team members can positively affect the Kauffman Foundation and US Labor decision-making process through a better Department is reported that 87% of the relationship conflict management, thus startups abandon the entrepreneurial idea mitigating divergences deriving from within seven years by the born (Fairlie and heterogeneity. Instead, the Miranda, 2017). Yet, the Thompson ambiguousness of the findings regarding Venture Economics showed that the 55% gender differences underlined the of startups receiving first financing round primary doubts claimed in the wide gets failure with economic loss, whereas literature (Phillips & O’Reilly, 1998) only the 6% is able to achieve financial produced on the argument, anyway the return five times bigger the initial prominent relationship conflict investment (Kerr et al., 2014). In order to management weight suggested by a put boundaries around the elevated number of different authors (Jehn et al. uncertainty, the researchers advise the 1997; Klotz et al., 2014; Fitzgerald et al., strategic agility utility which can derive 2017; Zhang, 2019) is here confirmed, both from structured (Ries, 2011; Camuffo thanks to what found on the internal et al., 2017) and more flexible approaches network variable and its mitigating (Sarasvathy, 2001), so that in this thesis effects. Yet, the startup team numerosity is project the decisional-making methods instead an evident proxy of how many named Scientific approach and external ties the team is potentially able to Effectuation will be explored. achieve: the more the team members are, the more they could merge their different Into the experiment of interest analyzed in contacts with the aim of creating a stable the following work, age heterogeneity and loyal business network. Referring to resulted to be positively and highly linked what written in the recent study of to the variables of interest, which are both Alessandri et al. (2018), the work linked to the decisional process. It showed engagement has been tracked and resulted to be more influencing on scientific as statistically significant throughout all approach rather than effectuative the regression models, with positive behavior: this is to confirm that the influence on both natural levels of experience diversity factor is deeply scientific and effectuative approach. crucial for the scientific paradigm, since it Additionally, as discussed by Amit (2001), puts roots on the continuous feedback also the prior experience in business plan exchange and test comparison, that is a 4 writing demonstrated to be influencing, (2019), too high academic achievements thus proving a real tie between the make become resistant to the capabilities practical business plan experience and the needed in the effectuation framework: this capabilities needed in the effectuative is what arised from who already obtained framework. Also differences in decisional a Master post-lauream or PhD, since these processes were expected when keeping two dimensions negatively related with the offering kind and industry type as effectuation. The top university levels reference (Schleimer & Shulman, 2011), seem to negatively influence the natural maybe due to the physical nature of effectuative approach maybe because the product vis-à-vis the more abstract service latter puts roots on flexibility, execution, concept. Here the offerings appear being and practical experience, namely features positively correlated with the sole being often stiffened and obstructed by scientific approach when they are services. extremely high levels of standard It can be due to the fact that the flexible academic education. In the end, the nature of the services makes easier to accumulated experience is difficult to create continuous feedback exchange with replicate and allows entrepreneurs to market and customers, so that phases of understand competitive structure and test and validation come to be facilitated market strengths, quality standards and thanks to the possibility of completing the most profitable trends. It improves entire process by only using online means, capability of performing more precise with consequent less time wasting and predictions, along with diminishing the more efficient learning. Yet, Mitchell and usually observed effects of Shepherd (2010) proved that decision- discouragement that emerge after the making tends to be excessively risky and initial instants. The outcomes from STATA inconsistent when acting in more dynamic indicate that correlation exists and always industries: our results seem to partially presents positive verse. By using the team corroborate their findings, because as reference, the average amount of belonging to the software industry different industries in which members negatively related to the sole innate have cultivated direct experience is effectuation degree. strongly linked to higher values of scientificity and effectuation; such is Excellent significance is guaranteed consistent with what described in (p<0.01) by the boolean indicating when literature, namely the effectuative team majority is engaged with a M.Sc. - it approach is someway strictly linked to the is positively linked to both innate experience dimension. Yet, the fact of scientificity and effectuation, therefore an already having established some firms or high academic level in progress can experienced business plan writing provide the needed technical insights to positively correlates with the sole innate conduct appropriate test phases and use effectuation degree. Again, this is to give right data validation tools (scientific evidence about the prominent experience approach), along with offering the amount weight on the effectuation framework, of experience necessary for balancing while showing its inferior influence on the abilities of execution, network, control and natural scientific approach. flexibility. Yet, according to Chatterji et al.

5

The Blau’s gender index finally showed startups, with a total of 542 entrepreneurs; significant negative correlation with the the following project will contribute to the natural scientific approach, looking like poor literature already existing about this such difference leads worse relationship topic, so that further analyses will be conflict management which outperforms encouraged in order to provide useful the benefit coming from a better task suggestions for scholars, with the final conflict management. The effect does not objective of fostering the Italian keep the same within the specification entrepreneurial scope and economic where effectuation is treated: this is to system. In light of the outstanding and corroborate that heterogeneity has largely superior performances registered by more influence on the structured startups applying scientific and dynamics of feedback exchange which effectuative approach – as furnished in characterize the scientific paradigm, while literature above all by Camuffo et al. are not a central driver for the natural (2017) and Sarasvathy et al. (2001, 2003) – effectuation level. Indeed, the effectuative this thesis work brings to conclude that the method pays more attention to enough institutional bodies as well as more abstract abilities such as network, control, technical entities (e.g. government, execution, and flexibility, less suffering universities, accelerators, incubators and the effect brought by task conflict and so on) should promote the most balanced relationship conflict management. These entrepreneurial team compositions, so as results perhaps confirm the intuitions to foster the positive impacts that are proposed by Carr (2010) and Tinkler likely to be observed on decisional (2016). The first paper found that decision- activities. making process can be affected by concerns about stereotypes and identity devaluation, rather than attribute gender differences to innate and stable factors.

Tinkler instead decided to investigate venture capitalists’ funding decisions in high-growth and high-tech entrepreneurship, so discovering that women were supposed to be less competent and having less leadership ability when available information was insufficient. The findings thus suggest for gender differences not linked to actual personal features, but rather to prejudices and social inequality; in other words, they warned about a sort of self-fulfilling prophecy at basis of the phenomenon.

The totality of the analyses executed in this thesis work took into consideration a sample composed of 305 mainly Italian

6

7

1. The Decision- circumstance is critical to adapt proactively the corporate attitudes and Making process features to the market advancements, in terms of offering, marketing activities, as During the last years, the entire economic well as quality requirements, design, world participated in the continuous usability, cost structure and price. The increase of new ventures launches, a trend entrepreneurial activities lead implicit surely supported by the lately introduced high risks and deserved a deep study working way in the high-technology dedicated by a number of academics. industries. Indeed, every day many firms Forlani and Mullins (2000), for instance, arise to follow the current agile growth stated that the main components affecting rate of the global entrepreneurial entrepreneurial risk perception lie in three ecosystem, trying to acquiesce to the high aspects, that are the investing sum of fund, standards and competitive habits dictated potential loss and uncertainty of by market demand. Thus, a so quickly anticipated outcome (i.e., the more capital growing system implies an ever- the new firm demands, the higher increasing complexity in various aspects: variability in expected result is assumed, extremely high competitiveness along therefore greater odds of economic losses several segments of customers, huge and higher risk degree perceived by complexity in understanding and entrepreneurs. It is relevant keeping in applying technical discovers, ambiguous mind that certain types of personality outcomes deriving from ability in traits can account for the majority in a gathering and interpretating data from team, but strategic decisions require for heterogeneous and diverse sources, not-individual processes: interactive difficulty in elaborating the psychological discussion is needed (Ping, 2004), and personal features coming from each however considering that single individual personality and, therefore, influences (complying to common rules) finally the role that every CEO (or affect overall outcome. In the recent founder, or co-founder team, or decision- literature two different currents maker in general) has in the corporate established in the decision-making success or collapse. In light of the above, framework, in order to better interpret and this study - as many others present in draw the entrepreneurial attitudes: literature - aims to investigate more heuristic search (McGrath, MacMillan, deeply the potential patterns and 1995; Shepherd et al., 2012; Sarasvathy, frameworks existing in the way by which 2001) - among which emerged the new firms (i.e. startup) arise, develop, and Effectuation approach suggested by possibly die. Sarasvathy (2001) - and a more rigorous One of the most relevant factors many and systematic one, named scientific scholars and researchers characterized as method (Eisenmann et al., 2013; Ries, 2011; influencer of enterprise dynamics is Camuffo et al., 2017). This thesis project mainly the decision-making process, will focus on the two different standpoints involved in the birth and enhancement of just cited, since they represent the attempt the businesses, since such key to manage and accommodate a continuously evolving and complex 8 economics, respectively by incessantly phenomenon almost never calls for a solo modelling (and adapting) the effort, so that it requires innovation entrepreneur’s strategies to the scope on management, along with startups’ one side, and by implementing a rigorous environment care and financing. Right (and faultless) method on the other side. A such last aspect is often the main reason recent research of Barron and Amoros why it is crucial creating a targeted (2019) highlighted how entrepreneurship approach and structured co-ordination, education for scientific communities is fighting against the criticality represented increasing as importance and common by the lack of proven business track attitude, thereby turn research activity records, which is a constraint making into marketable products. Their article strongly reluctant the funding groups in analyzes the entrepreneurial program giving trust to the new enterprises; a named NoBI offered to scientists in further issue is the need of expertise in Mexico, thanks to the support of business running the new emerging technologies and technical experts; it was aimed to and startups, other than the necessity to validate business ideas (exactly as done in generate an actually unique competitive scientific approach) and construct a global environment. consistent business model framework.

The final outcome observed was the improvement due to the program along 1.1 Innovation and both the two dimensions. The great attention assigned to certain entrepreneurship dynamics is only a facet of a spreader overview movement affirming in the new emerging Every entrepreneurial activity, in light of economies: many studies, as those the considerations expressed in the above realized by Koekemoer & Kachieng’a paragraph, is marked by a strong presence (2003), underlined the importance of of risk, uncertainty and complexity. These boosting the entrepenurial ecosystem factors are more crucial if we consider a given its critical role in the national particular type of entrepreneurship welfare, other than for the sociological, present in the current landscape: the so philanthropic and financial implications. called “start-up” ventures or, more in In addition, recent researches general, the high technology-based demonstrated the prevalence in weight enterprises. Indeed, neglecting of the legal (for the emerging economies as well as definitions of such entities (i.e., in terms of countries like Italy) of startups and small “innovative” formal definition, average & medium enterprises (SMEs) as main education level gained by the co-founders economic drivers in the local context: team, income flows, business’ years of life, during the last decade of the 20th century, original patents exploitation, and so on), at several scientific parks, innovation hubs the centre of the analysis proposed are and incubators arised, that is, renewed located all the new ventures able to attention to technology innovation and concretely contribute for the technological new services/products development has development, through the use of new incited the growth of new companies; such methods, thinking ways, applications and 9 innovative products/services in order to the wide Fairlie & Miranda (2017) study, meet the ever increasing market needs. realized by getting information from U.S. Given the definition of the object of study, Bureau of Census, Kauffman Foundation it is easy to understand that the scenario in and U.S. Department of Labor databases, which it plays is characterized by still a sample of around 581K startups was growing uncertainty, the major cause of analysed: they found that as much as 85% failure nowadays: so that, Effectuation of the total new ventures decided to and Scientific methods raise as possible abandon their own entrepreneurial idea solutions, providing strategic agility and within seven years (of the activities begin) pre-set decision making, which aim to and never hiring employees. Yet, as quickly adapt to the conditions mutation reported into an article of 2017 on Forbes although by using very different website (which grasped data from CB proceeding ways. Insights and Statista), and later cited by EU-Startups (2018), there is common agree A number of eminent works (e.g. Van de that the main reasons of startup failure are Ven, et al., 1984) deeply analysed the key lack of market demand, inconsistency success factors at the basis of startup between team and faced business, cash success. A whole comprehension of the burn ignoring. startups’ development phases is considered as fundamental (Pugliese, et Instead looking again at new ventures but al., 2016), thereby better handle and in more advanced growth phases, Kerr et coordinate them toward the success all (2014) studied a sample, by exploiting moment. Nowadays, the centrality of the the Thompson Venture Economics entrepreneur role within the society is dataset, composed of only startups which proven by a number of data, coming from already have received a first funding different sources reachable on the web, round. The researchers found that 55% of that confirm the importance of nourishing the total closed the activity and made it the business fabric. Many academics with economic loss, whereas only the 6% operative in the sector gathered a lot of gained financial returns of 5 times the evidences about economic indicators; e.g., starting obtained investment. Urbano & Aparicio (2016), and so also Furthermore, such 6% accounts for the Valliere & Peterson (2009), stated the new majority of the gross income registered ventures, through the creation of through the entire study, signalling an workplaces, prominently contribute to extreme and excessive disproportion in (not only) economic wellness of both the the distribution of economic success, countries and local areas where they arise namely of the quality about and progressively expand. Yet, it is entrepreneurial ideas and, still more interesting to highlight how the failure important, their execution. In light of the reasons, referring to the startup world, are above, the overview presented on the polarized towards few elements. In fact, innovative entrepreneur role, and the field the analysis was carried out only on US where he plays, is enough complex and startups, but the results appear to be difficult to interpret. The contribute consistent with what is expected also in brought to the economic system is other geographical clusters. According to recognized, nevertheless the criticality

10 encountered in the patterns analysis and (1921), who defined the “risk” word as a in measuring various attitudes and quantifiable phenomenon, of which we behaviors make challenging a whole have no certainty about the fact it will comprehension of the phenomenon, so happen, but its frequency is computable in that - seen the huge innate and intrinsic statistical terms through the means uncertainty conditions too - provoke the offered by the probability distributions. high failure rates observed by using the Instead, the “uncertainty” is an event of governmental data as means. non-quantifiable nature, namely of which it is impossible to make conjectures about probability distributions too. Thus, the evident difference is that, in the first 1.2 Scientific Method situation, the difficulty can be mitigated The author and American entrepreneur thanks to imprecise esteems extracted Eric Ries attempts to find an answer by from the historical series; meanwhile, declaring that each startup, given the looking at the uncertainty definition, we uncertain and unpredictable scenario observe the added requested effort where it must compete, has to introduce a brought by the lack of previous experience new and revolutionary approach in the in equal (or similar) activities and equal decision-making process for its own (or similar) boundary conditions. routines. Indeed the success (or failure), as The classic managerial techniques based result of long period performance, is an on ex-ante and deep study of problem and outcome deriving from a pre-set and scope, so trying to derive the probability systemic strategy, that is, it never arises distribution and better predict the right only thanks to genetic heritage or lucky development to follow as well as the contingencies. Ries (2011) tends to stress potential results: therefore, they act in more and more times on the concept of situations of risk and are ideal to such uncertainty, that each entrepreneur frameworks; yet, in case of issues (differently from managers e.g. CEOs, characterized by really high uncertainty, directors, admins) has to face starting from an ex-post approach would result useless his venture’s first day of life. According to and impossible by definition on one side, what already said, the statement done by but by the other side the classic techniques Freedman (1992) many years ago makes require for making too strong sense: the uncertainty factor plays a assumptions, leading to solutions founded crucial role in the choice and management on erroneous basis. Finally, the answer to of every-day entrepreneurial activities, so the largely uncertain entrepreneurial that it requires deeply diverse decision- scenario can be found in a more adaptive making tools, if compared with the method, the so-called strategic agility canonical ones usually used in certainty suited for quickly adapting to the events and risk conditions. Consequently, it mutation. Consequently, the possible appears of fundamental importance to solutions proposed in this thesis project make distinction between risk and are two: heuristic search (McGrath, R.G., uncertainty. One of the first academic MacMillan, I.C., 1995; Sarasvathy, S.D., helps in this direction is offered by Knight 2001; Shepherd, D.A. et al., 2012) - thought

11 to better mitigate and deal with situations experienced by the same by exploiting specific ad hoc tools - and entrepreneur as well as the scientific method (Ries, E., 2011; predecessors who worked in Eisenmann et al., 2013; Camuffo, A. et al., similar industry and/or under 2017), which looked at routines used by similar boundary conditions; researchers and academics. 2. creation of reliable and falsifiable The Scientific approach puts its own roots hypotheses suited to be onto the methodologies cultivated by successively tested, with the goal of scientists and researchers (Camuffo et al., trying to understand and finding 2017; Ries, E., 2011; Eisenmann et al., confirmation/refusal on specific 2013), just like scholars and several types and precise aspects of the business of science professionals, who use it in and/or entrepreneurial idea, so as order to better observe, analyze and to avoid wastes in terms of understand the phenomenon taken into resources and time, highly limited consideration. Eisenmann et al. (2013) by definition in the analyzed scope; showed the superiority of scientific approach adapted to business experimentation rather than other 3. testing such hypotheses, by using methods; Kerr and colleagues (2014) the right methods and tools to believed that scientific experimentation interpretate and elaborate the data, was central within the entrepreneurial focused on introducing no decision-making, as knowledge is not distortions into the experiment fully accessible in advance and cannot be design, so that the outcome is deduced from set of first principles. Such consistent and solid as much as widespread mindset, highly diffused in possible; academic environment, is appropriate to be exploited as a general research 4. continuous and iterative validation criterion, therefore also relevant in the (that is, probably the most decision-making sphere. It is based on the important side of the process) of the systematic path followed throughout feedbacks gathered from market, every scientific experiment, beginning target customer, offering kind and, from a set of initial conjectures (in this more in general, from the whole case, the actual starting point is the competitive ecosystem, paying entrepreneurial idea) with the purpose of attention to potential both finding acknowledgement from the real perceptual and psychological world and, if needed, starting again after biases - very common even if at having gathered the feedback and setted high academic levels - which may again the flow. Analogously, it is possible move the results toward a direction to recognize the four primary components far from the exact reality of the process, namely: interpretation. 1. formulation of a solid theory, created thanks to deep studies of the scope and previous events, 12

Since 2017, because of the RCT focusing on the minimum resources (Randomized & Controlled Trial) consuming. Then it comes to pivot the conducted by Camuffo and his team on an idea or proceed with Measure, the central early-stage startups sample, the research phase in which information and data are community can count on a new generated. Furthermore, here the fundamental contribute about the entrepreneurs has to define metrics and scientificity applied to the decision- validation threshold. Then, into the Learn making paradigm, given the fact he step, Ries describes the traits of a critical realized the empirical measure of its interpretation based on evidences: the impact: Camuffo et al. (2017), from final outcomes drive to pivot the idea, Bocconi University of Milan, confirming the business model or demonstrated the positive correlation dropout. between scientificity level and i) frequency Other scholars, like Gans et al. (2017), in abandoning the fallacious initial underlined the central role of the entrepreneurial ideas, ii) numbers of pivot entrepreneurial uncertainty in both (i.e., logical consequence of a negative decisions about business selection and outcome resulting from the validation observed failure reasons. Zenger (2015) process, signalling the Scientific method too elaborated that strategies cannot be has been properly followed by the mere trial-and-error processes. Thus, the entrepreneur, placing the real evidences validation phase has to be still more prior his own convictions) carried out, and fundamental in the business conducting iii) level of startup performance. Such (McGrath, R.G., & MacMillan, I.C., 1995), outcome is consistent and due to the so the literature provides entrepreneurs greater precision and accuracy provided with scientific approaches to decision- by the scientific mindset cultivated making processes, among which the Lean throughout the entire development, with Startup is one of the most recognized the appropriate methodologies and tools: (Ries, E., 2011, & Eisenmann et al., 2013), this makes easier to detect inadequate founded on the Galilean thinking way projects after reduced time bucket, regarding test and validation of ideas, improving the odds to pursue the with the purpose of learning maximisation successful ones. More exactly, it increases in high uncertainty conditions. In odds to pursue ideas with false negative particular, Vining (2013) defined it as an returns. iterative process used for studying an Ries (2011) described the scientific unknown situation (e.g. a new largely approach as a Build-Measure-Learn cycle, innovative venture) and its intrinsic ending with product improvement, insecurity, by exploiting a set of changing strategy or changing vision; the procedures repeating throughout the decisions however bring new time. The provisions to which submit are implementation, experiment and then divided in four steps and, as mentioned in learning. The Build phase puts its own the above paragraphs, they are theory roots into the hypotheses formulated by formulation, falsifiable hypotheses model the entrepreneur, and is usually realized generation, testing with respective data through MVP (Minimum Viable Product), analysis and finally the

13 rejection/validation, always keeping into experiments, by using an appropriate consideration the goodness and veracity objective criterion that can quantitatively of test and theory. The possible refusing measures an apparently abstract provides us with the right insight and dimension. Since the falsification counts as feedback needed to proceed for the a kind of central moment, the test utilized complete review, so that the entrepreneur should be rigorous and carefully setted, can be ready to start with a new other than realized on reliable and high development cycle. In the scientific quality data, gathered according to precise approach, the steps conceived to be strictly and agreed rules; the most classic tests, followed are the core discriminating that usable by every type of entrepreneur and distinguishes between good and bad all the possible business scopes, are enterprise path, while limiting the personal interviews, survey and poll frequency of failures along the administration, as well as A/B entrepreneur’s route: so that, the way by comparison and MVP (Minimum Viable which they are designed and mostly Product). The use of MVP version executed plays a critical role in the success empirically demonstrated huge utility, of the idea. Looking with particular focus given that it allows of quickly testing at them, it appears appropriate to specify different versions about the same offering, that, during the theory creation phase, the therefore avoiding preventable entire team should be able to generate commitment in terms of time and financial precise frameworks and elaborations additional resources. The Minimum about certain business facets deserving Viable Product is a mindset setting, dedicated attention. For instance, the initially suited on the business theory may be referred about a number of background but even relevant in other business model components, like target contexts, seen that it provides a solid tool customer, product, cost structure (and so resulting the more useful (Ries. E., 2011) final price), marketing channels, the more is the uncertainty level exhibiting distribution means, key activities, selling in any knowledge side. During the last mode and main commercial/strategic step, following the testing process, the partners; the more such procedures are entrepreneur’s focus is suggested to be well defined and clearly structured, the towards various biases: indeed, their more the entrepreneur will be ready to existence’s documentation in literature is minimize the number of failures while widely reported in a number of scientific maximizing the learning level (Zenger, papers, both in psychological and socio- T.R., 2016). Coming to the hypothesis’s economic studies. They include formulation, yet the conjectures must acquiescence (i.e., distortion in data comply the criteria of clearness and collection due to respondents which tend falsifiability: the second one criterion is to give answers depending on particularly needed to avoid bias and entrepreneur beliefs rather than other kinds of distortion, such as false complying to a truthful criterion) and positives perceptions during the selection bias (Kahneman, Slovic, Tversky, decisional process. Falsifiability is defined 1982; Kahneman, 2011; Clark & (Eisenmann, 2013) as the possibility to be Wiesenfeld, Harvard Bus. Rev.), rejected if resorted to the right specifical confirmation (Gilbert, 1991) and 14 overconfidence (Cooper et al., 1988; 1999; Adner, R., Levinthal, D., 2004) and Koellinger, Minniti, Schade, 2007; Klein, similar to the decision tree, a diffused tool 2007) bias, optimism bias (Kahneman & in project management and other scientific Lovallo, 1993). The existence of biases is disciplines. These provide help in case of the more likely the less defined are the highly uncertain financial situations: the selected metrics; in his book “The Lean idea passes from being a whole project to Startup”, Eric Ries proposed a number of becoming an authentic decision tree quantitative thresholds, usable in very branching, where each branch represents diverse application fields: such numerical a real option. Designing and limits become essential when the implementing rigorous experiments allow entrepreneur comes to the decision to avoid option trap, that might hinder whether proceeding with tests, dropout or generate escalation and abandoning the idea or finding minor overcommitment. changes. In case he chooses not to proceed Thus, the validation process conducted by with the same initial setting, two ways are the entrepreneur may be interpreted as a in front of him: definitive desertion or real option purchase of partial business pivot. The second one consists of re- idea, that is, when the experiment finishes considering the business idea and theory, the entrepreneur can decide to pursue the in light of the conjectures’ falsification or project exercising its own option, new inputs deriving from external otherwise abandon the idea waiting for environment. The review of his own initial option expiration date. In analogy on what convictions has to be seen as a good reported in the real options theory, the starting point, rather than as failure and entrepreneur appears as in front of a shame; in fact, the cyclical loop generated decision node every time after an from such thinking approach results as experiment conclusion: here he has to exactly the principle on which is based the select among pivoting, pursuing, and scientific method, suggesting for an abandoning the idea. Camuffo et all (2017) iterative process in incessant evolution showed how a scientific method is and, therefore, admitting errors and re- positively correlated with amount (and begin points. Furthermore, the continuous extent) of entrepreneurial pivot and exit mutual feedback interchange among the decisions. The several choice criteria are business protagonists aims to deal with all ascribable to the one based on the the high uncertainty, again, controlling for perceived value of the business idea and the economic investment on not-added- the consequent expected profit; the term value activities and resources, but a “perceived” deserves attention since it reliable measure of the value is not distinguishes if compared to the actual common knowledge in ex-ante scenario, value of the idea, due to bias during the so that it looks necessary initially assessment or interpretation deriving exploring the business ecosystem with from internal distortion effects or someway initial limited commitment. The environment conditioning. Then, a further unstoppable feedback exchange found its point is the influencing power of the roots in the real options conceptualization, scientific mindset on the perceived value a widespread framework largely adopted comprehension, aimed to isolate the into the finance world (McGrath, R.G., 15 disturbing variables for getting more derive only by the causal variables, accurate information about the expected without environmental influences. In income (i.e. without the frequent order to verify the natural differences overestimation tendency). The scientific between the participants, a monitoring approach in decision-making routines is a program was established, including recently introduced methodology, slowly telephonic interviews with once a month diffusing nowadays among entrepreneurs frequency, aimed to detect intrinsic and scholars; indeed, a certain number of scientificity level at decision nodes and relevant studies are moving toward this type of conducted activities. So, the whole direction, and with a heterogenous pool of participants was divided into two geographical distribution: e.g., the homogeneous sub-groups through a Russian studies realized by Veretennikova randomization criterion, with the aim of & Vaskiv (2018) confirmed the increasing avoiding the disturb effects introduced by interest in such thematic at global level, internal and intrinsic heterogeneities while highlighting its centrality at initially in the sample. Hence, while one managerial view occupied in launching group experienced the scientific treatment new innovative products/services, (so named Treatment Group), the other namely realities characterized by largely served as controlling means (so named agile and complex dynamics. However, Control Group): by this way, when the the international literature still lacks outcomes deriving from the experiment evidences about the effectiveness brought come out, it becomes easier to understand by scientific thinking, and one of the whether the observed result belongs to greatest contributions comes from ICRIOS one effect rather than to another, isolating research centre of Università Bocconi in the causal variables since it became doable Milan. In particular, the RCT experiment linking the evidences to the respective conducted by Camuffo, Cordova & diversities in treatment. A course Gambardella (2017) investigated whether composed of ten lessons on market the scientific routines applied on decision- analysis and feasibility testing was given making activities had an observable effect to both treatment and control group, if measured in only early-stage startups: introducing the only difference that in the focus has been over the consequences treatment group the execution of activities come to light in terms of performance, in terms of scientific view is taught. In the amount of abandoned ideas and amount end, the hypotheses were confirmed: the of pivots. The conjectures were that the treated early-stage startups, compared to application of treatments on early stage the early-stage startups belonging to startups - where the external effects are control group, systematically reported minimum, and contingencies have yet superior amount of economic turnover, influenced at a low level - would have along with greater number of both resulted in 1) greater economic revenues, pivoting and abandoning rates; this result 2) major number of left entrepreneurial confirms the starting intuitions, and ideas and 3) greater number of pivots. This underlined how a more rigorous and kind of study has been conducted as a structured approach can lead Randomized & Controlled Trial, because entrepreneurs to better recognize i) the of the necessity to be sure the effects could most profitable paths on which develop 16 the idea (pivot), ii) the most efficient Nevertheless, the high uncertainty degree means to better maximize the profit necessarily imposes of narrowing it down (higher revenues), and finally iii) the cases by using assumptions as means, so that where the initial business idea is totally their utilization is fundamental and wrong, given the boundaries conditions. cannot be neglected; hence, the greatest objective is to let entrepreneur can really

understand the huge impact brought by 1.3 Heuristic Search & his assumptions on the final outcome. In particular, the first 3 phases accurately Effectuation described the entrepreneurial idea also in Alternatively to the scientific method just light of the assumptions making it doable. illustrated, in the existing previous Instead, the fourth and fifth phases act literature the possible solutions offered with the aim of validating the conjectures’ with the purpose of tackling (i.e., goodness, other than measuring the mitigating its effects and negative possible impact of their variation on the impacts) the entrepreneurial uncertainty enterprise scope. Basically, the primary are the heuristic search versions, and they goal refers to the fact that it is highly better are mainly three: effectuation (Sarasvathy, to avoid resources commitment in case of S.D., 2001), discovery driven planning fallacious initial ideas, and this is the more (McGrath, R.G., & MacMillan, I.C., 1995) efficient way: starting by deeply and confirmatory search (Shepherd, D.A., investigating as in a sort of feasibility Haynie, J.M., McMullen, J.S., 2012). If analysis, before to proceed in pursuing considered the solo discovery driven further investment of financial, time and planning, it is possible to distinguish a set human resources. Yet, Shepherd et al. of distinct phases to strictly follow. (2012) propose of using the confirmatory Indeed, McGrath & MacMillan propose search heuristic on every assumption, five diverse items, in order: business with the objective of confirming (or framing (planning definition aimed to rejecting) its goodness. It can be conducted locate the best initiatives to pursue), by resorting to three different sub-types: benchmarking (referring to both market positive search (i.e., by testing the and competitors), and translation in terms conjecture truthfulness when the veracity of functional strategies along with focus is expected) and, as opposite, the negative on requirements about the operations. Yet, search (i.e., by testing the assumption assumptions documentation and, finally, goodness when the contrary is expected), the phase in which we have to clearly other than the mixed search (i.e., by identify the milestones. The totality of merging the previous ones). Based on such such rules provides the tools suited to search versions, it is possible to proceed in restrict the impact of the assumptions, generating several frameworks, which which each entrepreneur inevitably makes vary depending on a number of factors during the setting of its own business (e.g., kind of assumption, odds to since, as already hinted, the building of underestimate its impact compared to the systems born on fragile bases is a very overrate probability, cost associated to the frequent and common problem. esteem error). All the presented procedures try to test the whole of the 17 uncertainty sources, in order to limit heuristic search aims to avoid an excessive distortions due to existing biases (e.g., level of investment on the not-best confirmation bias) and errors on the cost solutions, so that the resources dispersion esteem linked to wrong initial minimisation is made possible thanks to assumptions. four primary principles (to strictly comply if the desired effect is of concentrating Finally, it comes to Effectuation heuristic investment on few, precise and search: Sarasvathy (2001) describes an remunerative aspects) : 1) focus on approach based on effectuative processes, reduced losses rather than profit differently from what usually used (i.e., maximisation; 2) preferring the strategic causal ones) by both entrepreneurs and alliances and partnerships, rather than managers. Indeed, the causal method pursue competitive strategies based on mandates to realize the means selection by market analysis and trend predictions; 3) beginning from a given effect: this path of focus on continuous development of course leads the risk of having a fallacious expertise and, more in general, of business (due to the high entrepreneurial experience, instead of looking at the uncertainty) starting point, consequently historical series and at the previous bringing erroneous conclusions. As entrepreneurial experiences; 4) trying to effective solution, she defines the keep control over the future and the effectuation standpoint as something possible event evolutions, in place of which puts the roots into the available making unreliable predictions and basing resources (e.g., time, cash, immobilized the long-period tactics on conjectures economic flows, real estate, human about what it is impossible to experience resources, network, and so on), so that at the moment. consequently choses the best alternative among the possible usable business The final purpose is still that of placing scenarios. However, this is not a linear boundaries around the assumptions’ process since is follows iterative steps: the effect, preventing the use of wrong idea development is continuously resources, with excessive commitment at dynamic and has to be shaped according the wrong time or in the wrong place. to the new information acquired Saras Sarasvathy (2003), in her essay throughout the venture progress. named “Entrepreneurship as a science of Similarly, the management approaches the artificial”, wrote about some key nowadays are quickly diffusing at several concepts by drawing on Herbert Simon’s levels of the organizations (e.g., agile, article “Sciences of the artificial”. The scrum, complex matrix structures, essay wants to connect the following four elaborated cross-functional teams, etc…), ideas to recent paper on the proceeding along a “Bottom-Up” mindset entrepreneurial expertise: comparable to the one traced by i) the natural laws constrain, Sarasvathy’s theory, fostering a more without the power of dictating inclusive environment and making it able the implemented designs; to deal with the new challenging and hyper-dynamic current competitive ii) necessity of seizing every opportunities. In this case too, the opportunity, in order to avoid 18

the use of predictions in design Though, it appears equally relevant the process; importance of keeping in mind that such constraints represent a superable limit iii) we have to accept that locality someway, in the sense of circumscribing and contingency both govern the area in which startupper can move, but the whole science of artificial; without defining exactly the pathway to walk; that is, a tight leeway exists and waits for being exploited. iv) near-decomposability is a key

feature of the enduring designs. 1.4 Team heterogeneity The principles Sarasvathy provides in her studies stressed the centrality of and other effects maintaining focus on flexibility, in view of A really wide preexisting literature the contingency dominion all over human already established about topics as the activities. In conclusion, the entire set of influence of team heterogeneity on relevant key factors falls into five corporate performances. However, categories: 1) bird in hand, namely heterogeneous composition may refers to capability of founding and developing the a number of facets, ranging from age, entrepreneurial idea putting roots into the passing through gender and geography, initial personal resources owned by the until to personal behavioural traits. First of decision maker, in terms of individual all, the trigger for a such big research flow abilities, background and network; 2) was a series of dated essays (Resnick et al., affordable loss, standing for the measure in 1991; Nonaka, 1995), asserting that fruitful which the founder is ready to invest the discoveries in group dynamics are maximum affordable amount looking at brought by simultaneous presence of time as well as at financial investment; 3) diverse standpoints and perspectives by crazy quilt, gauging the extent to which the one side. Meanwhile by the other side, entrepreneur succeeds in mitigating more recent works, as those proposed by uncertainty by adopting proactive Jehn et all (1997), demonstrated that team behavior (i.e., creating contact network characterized by highly different with all the components of the business education experiences had relevant more chain, like clients, suppliers and difficulties in defining activity progress, competitors); 4) lemonade orientation, thus presented a prominent problem in indicates how much the founders the analyzed framework: heterogeneity exploited unexpected events to create new appeared to bring, along with disruptive opportunities basing on the improvements in performance, a certain available resources and flexible aptitude; degree of difficulty into the management 5) pilot plane, when the focus is on what of internal conflicts. Van Knippenberg and entrepreneur can control and does quite Schippers (2007) defined team well, so that execution becomes a key heterogeneity as the perceived difference success factor at the expense of wait and in attribute features among team predictions. members. The essay of Mell Van

19

Knippenberg (2016) deserves to be the second one had negative influence on mentioned, as he underlined how many overall performance through scholars deeply contributed to the insights interpersonal conflicts. Klotz et al. (2014) found on team heterogeneity, focusing on proved that heterogenous compositions influence affecting firm’s results and lead to more diverse knowledge and decision-making processes. Early capabilities, thereby foster creativity and explorations (Phillips, & O’Reilly, 1998) innovation. However the phenomenon were based on social variables (e.g. encounters side effects, given that more gender, race, age), often leading to homogeneous teams sometimes achieved penalizing effects on performance and better results thanks to higher cohesion, internal satisfaction. Successively, also that significantly improves conflict counting on a strong Asiatic contribute on management and efficacy (meant as speed this side, a deeper research began and in reaching for purpose). shifted toward less superficial Other papers published by Jehn again heterogeneity factors: new attention was faced with such topics, highlighting how given to discern between demographic the team conflict was due to the perception features and expertise factors, such as of team members on intra-team cognitive professional background and university differences and goal incompatibility. At achievements, which could reflect the the begin, scholars used to believe accumulated knowledge, standpoint and conflicts leaded to lower effectiveness, ideological tendencies. Kristinsson et all later Jehn helped to realize that task (2015), drawing on information decision- conflicts were positively correlated with making theory from perspective of resulting performance, while cognitive resources diversification, relationships conflicts were negatively declared that different team members correlated with overall efficiency. Also consider issues from different viewpoints Fitzgerald et al. (2017) employed effort in according to their experience, while at the investigating the above themes, declaring same time the integration of different that diverse professional backgrounds and insights is more likely to improve experiences produced different opinions disruptive creativity and team decision- on the same point, which was bound to making capability. Great contribute was improve the degree of team task conflict offered by Jehn et al. (1999), which in “Why management. Consequently, such results Differences Make a Difference: A Field Study suggested for correlating convergence of of Diversity, Conflict, and Performance in diverse opinions among members of Workgroups” presented a study aimed to different expertise with enhancing in criticize prior approaches and widespread entrepreneurial innovation and decision- ways of thinking. They stated there are making quality. This brings to a relevant many factors leading to diametrically conclusion: task conflict plays mediating opposite evidences, so that was necessary role into the relationship between team a division between 1) information, 2) heterogeneity and performance (i.e., the social classification attribute and 3) value positive influence of heterogeneity affects heterogeneities; the conclusion was that venture outcomes through task conflict). the first type positively affected Drawing on a very recent work of Zhang performance through task conflicts, while 20

(2019), aimed to reveal the contribute external actors of market (Adler & Kwon, guaranteed by expertise heterogeneity 2002). Arif T. (2015) studied social (i.e., vocational and professional networks of computer engineering background) on team performances department within the Indian Institute of through mediating effect of team conflict, Technology, getting quantitative insights it seems to exist strong evidence of such about any behavior. Also organization relationship. The mediating role of team theorists coming from universities in east task conflict and attitudes toward of Europe, Lithuania in this case, invoked heterogeneity is prominent and central. the social capital theory application to social network paradigm (Jurkevičienė et Deserves to be cited another recent work al., 2018). It provides theoretical evidence (Hoogendoorn et al., 2017), based on a for links between entrepreneurial field experiment involving 573 students, heterogeneity and performance, by where exogenous variation was ensured answering that benefits are due to in (otherwise random) team composition complementary in psychological traits by assigning students to various teams and diverse attitudes which - when according to their measured abilities. supported by right mix and overlapping - Hoogendoorn demonstrated how lead improvements in outcomes. Aldrich corporate performances subtend under and Kim (2007) confirmed how three macro-categories (i.e., opportunity entrepreneurship requires social recognition, problem solving and connections to several professional implementation) and thus studied their groups, usually not accessible to a lone variations. They found that, within a team, entrepreneur. Again, invoking literature a balanced combination of higher and on the social capital theory, Reagans et al. lower cognitive abilities levels is largely (2004) suggested the compositional more productive as some people can be diversity provides benefit thanks to the assigned to mundane tasks (usually range of diverse ties that each member implementation) while others to the ones guarantees externally to the team calling for greater cognitive capacity. environment, so obtaining access to Consequently, performance of teams first valuable extra resources and information. increases than decreases with ability dispersion, and average team ability is not A deeply large-scale Russian study carried related to team performance. out by Aven and Hillmann (2018), by analysing 9461 entrepreneurs and 2446 industrial enterprises, showed that variation among team members’ 1.4.1 The role of network brokering ability significantly predicted A great current of literature has been the starting capital gained by the firms. dedicated on the argument, by treating the The focus was on team members’ ability to topic both in qualitative and more act as network brokers. Furthermore, quantitative ways. Team social network another evidence was that when both definition usually focused on the existence average and variance in brokering of network based over informal - or/and potential were high for the team, then formal - ties between team members and firms raised greater starting capital. A 21 recent paper published by Butler et al. latter is based on team learning activities. (2019) put attention on new tendencies: The effect could be observed through the with the increasing penetration of digital influence of team composition, given that paradigms into the entrepreneurial the Geremias’ study was conducted on competition, the traditional need for only-students groups. proximity to specific locations or the necessity of huge funding for infrastructure building has diminished, 1.4.2 Psychological and of course such effect has been recently exacerbated by pandemic emergency. dimensions Thus, nowadays entrepreneurs pursue A certain amount of Asian papers (Su-li & locations that provide more opportunities Ke-fan, 2011) conducted extremely for founding rounds and greater social quantitative researches on abstract topics, networks. So, Butler and colleagues for example by adopting the MBTI grasped data from CrunchBase and framework to interpret risk decision- PricewaterhouseCoopers, while making activities within group scope. A intercepting network dimensions by set of variables described individual scanning social pages from LinkedIn. The behavior differences: (E)xtraversion vs results coming from 1418 analysed (I)ntroversion as source of spiritual entrepreneurs suggested that funding vigour, (S)ensing vs i(N)tuition as ways to rounds per year is a reliable indicator grasp information, (T)hinking and playing significant and positive role in (F)eeling in decision-making attitudes, affecting startup creation in certain finally (J)udging and (P)erceiving in locations: local social network density adapting to external contingencies. The generates stickiness to the local relevant results were that: environment, thus negatively influencing entrepreneurs’ willingness to relocate towards further places. There was also 1. teams with majority of J members evidence that midcareer individuals (like were more prone to make millennials) are more likely than early- optimistic estimates about career and late-career to create successful entrepreneurial opportunities than technology startups. A new work carried S majorities; out by Geremias, Lopes and Soares (2020) on a sample of 480 undergraduate 2. groups with more P individuals students gives evidence of the positive were likely to detect greater correlation between network centrality amount of opportunities and lower and internal learning in young teams. risks, while possessing higher risk Such result allows to understand the perception if compared to those importance of centrality in advise with J majority; networks among innovative teams: this suggests for positive correlation between such centrality (e.g. it can be measured by the Team Building variable tracked in IVL dataset) and scientific approach, since the 22

3. teams with more E members felt harmful. Again in terms of age, the work lower risks than who was labelled of Amit et al. (2001) gave contribute. as I type. Firstly, they found that entrepreneurs engaged in business planning activity

reduced the likelihood of venture The resulting outcomes helped to disbanding while increasing the odds of understand how it should be of central pursuing on the idea with success (so importance to know and interpret business plan boolean variable is used as personality traits, thereby prevent too control). Secondly, they demonstrated optimistic (or pessimistic) judgements and younger entrepreneurs showed more false negatives/positives through a overconfidence bias while running rational decision-making process. The ventures, along with lower decision already mentioned studies produced by comprehensiveness - similarly to what Clark and Wiesenfeld (2017) focused on written by Parker (2006) who showed how personal traits by the bias viewpoint: they older entrepreneurs were more anchored reported cases of companies which made to their own prior beliefs, given the major strategic decisions based on biased amount of prior experience, lacking in samples, that are more likely to capacity of reacting to market signals and corroborate initial hypotheses even if data evolving information. This current suggest the opposite. suggests for the use of seniority dimension as control variable in regressions, since it Other scholars (e.g. Parker, S.C., 2006) can be the effect through which is stated that entrepreneurs differ in the way mediated the scientific approach. which they anchor to their own beliefs, leading to ignore external signals and Further findings on personal traits of information coming from real market, entrepreneurs (Shah & Tripsas, 2007; Shah with such phenomenon more pronounced & Tripsas, 2012) took into consideration in older individuals. Then it means less the possibility that entrepreneur is even adaptivity in decision-making. Difference the user of the product he offers on in age also suggests for using seniority market: in those papers is proposed the parameter to control variations in natural idea of a sort of correlation between scientific degree, given that scientific personal motivations and outcomes. How method is exactly based on better in these cases do entrepreneurs exploit interpreting external signals. Several types opportunities (key aspect of effectuative of biases usually affect the human approach) based on initial inventions decision-making process, and they are thought for their own use? Do the social subtended to certain psychological and capital (e.g. involvement in user attitudinal aspects, even if a scientific communities, contacts to markets) weight approach setting is adopted. York et al. increase or decrease within these (2014) revealed them in a dedicated paper: scenarios? The most interesting emerging acquiescence, selection, confirmation (also question is: does such process depend on described into Gilbert’s essay published in nature of the community problem? That 1991), overconfidence and optimism means, for instance, the ‘inventors-users’ biases were the most frequent and in medical sector can be more prone to

23 share their discovers given the prominent develop new capabilities and knowledge, iii) I social impact, maybe emphasizing the like difficult lavorative tasks by which I can economic gain less at the expense of develop new competences, iv) The possibility higher interest for fighting others’ to can develop work abilities is so important to sufferance. Wang et al. (2019) tried to take a risk, v) I prefer to work in situations connect psychological capital with startup calling for high capability and talent. What performance, basing on reliable emerged was that mentoring supported interviews registered by famous increase in opportunity recognition only international entrepreneurs: the resulting for low LGO mentees. finding was that human (i.e. age and education, work experience, near relatives models), as well as relational (i.e. 1.4.3 Industry effect trustworthiness and relationships among Mitchell and Shepherd (2010) explored co-founders) and strictly psychological whether the environment in which (i.e. optimism, autonomy, hope, resilience) entrepreneur plays might change his key features showed being crucial for decision-making criteria, discovering that born, development and performance of the executives acting in more dynamic new ventures. The central relevance of the sectors, namely characterized by high psychological capital has been recently uncertainty in future predictions (e.g., underlined by Alessandri et al. (2018) too, innovative startups), are prone to take who created structural equation decisions inconsistent with market and modelling analysis and found that harmful for corporate performances. This absolute levels as well as increases in result makes think that more dynamic, PsyCap (hope, autonomy, resilience, uncertain and then risky industries (e.g., optimism) predicted work engagement ICT) can lead to a minor average degree of increases, which in turn predicted better scientificity. job performances. Anna et all (2000), looking at a sample Finally, a narrow set of specific variables coming from US states of Utah and has been adopted according to what Illinois, investigated whether relevant described by St-Jean and Tremblay (2020), differences in nature of industry mitigated who declared how mentors’ support can the gender effect. According to the SBA, 22 facilitate the development of a particular million of US small businesses generated psychological trait (i.e. the opportunity more than half of the total GDP and were recognition subtended to self-efficacy) the principal sources of new jobs. The through the moderating effect of National Foundation for Women Business psychological aptitudes, mostly the Owners reported that the amount of Learning Goal Orientation (LGO). Such women-owned ventures experienced last property has been computed as huge growth during the last fifty years. arithmetic mean of 5 different sub- Although the growth in quantity is variables tracked thanks to the following encouraging, the size remains small in questions: i) I tend to face with challenging terms of revenues and employees if working tasks which can teach me a lot, ii) I compared to male-owned firms. One continuously search for opportunities useful to explanation for the disparity is that 24 females tend to concentrate on retail and personal traits, indeed women-unanimity service industries, where markets are teams were less aggressive in pricing smaller in opposition to sectors like high- strategies, invested less in R&D but more technology, construction and in social sustainability. A more dated manufacturing. One of the most fruitful work of Gatewood et al. (1995) explored streams of research based on the self- differences in gender about motivations efficacy trait. The final outcomes showed meant as success key factors in how traditional-women business owners technological entrepreneurship – then had largely different key success factors concluding that women who started than non-traditional owners; the resulting having personal reasons (e.g. autonomy) conclusion appears to be appropriate for and men who started having external being extended at men level, keeping the incentives (e.g. solve a perceived market same differences among different need) were statistically more likely to industries. successfully terminate the initial phase of firm development. Shepherd (2012) also affirmed that gender effect can vary depending on the nature of industry (e.g., 1.4.4 Gender effect traditional or non-traditional) in which A largely explored current exists on entrepreneur plays, so suggesting that differences in gender in terms of decision- gender impact can be controlled, for making aptitudes. Please note that example, even through the kind of literature elaborated in this sense is not offering. Still, Shepherd invokes more aimed to state the supremacy of one attention on conditions under which category rather than the another; instead, gender differences are magnified (e.g. it aims to take notice of the existing non-traditional-for-women industries), differences, due to endogenous, evolutive thereby recognize and fight them. Finally, and natural factors, thereby formulate the he concludes that the focus could move perfect mix in team composition and from the amount to the nature of the diversities management. However, the experience (e.g. failed attempts of startup gender effect may be mitigated by the establishing), thus leading to consider dimension named female identity, as such aspect as control variable throughout highlighted by Shepherd (2012). the regressions. Apesteguia (2012) carried out a complex Findings in anxiety management (De and long field experiment, demonstrating Visser et al., 2010) confirmed relevant that: i) teams formed by women totality differences in genders: by a medical point are significantly outperformed by all the of view, decision-making and anxiety other gender combinations, both at share underlying neural substrates, so that undergraduate and MBA levels; ii) the variations in anxiety handling capability best performing group is two men more 1 provoke variations in decision-making women for MBA classes (suggesting for a and cognitive functioning. Indeed, women mitigating role of women presence); iii) were demonstrated to be more able in the differences in performance are anxiety management, resulting in higher explained by those in decision making and lucidity through complex decision

25 moments. In addition, the anxiety impact hormonal components, namely the was more evident for men during the early endogenous level of testosterone: they stage (i.e. exploration phase), whereas discovered that high testosterone levels stronger on women during the were correlated with greater risk-taking exploitation phases of the tasks. A key role attitude, and this was true both in men is played by the perfectionism trait, strictly and women individuals. The difference in linked to fear for failure and anxiety taking risky choices is demonstrated even generation (Frost, Marten & Lahart, 1990). at neuroscientific level: an extremely Masson, Cadot and Ansseau (2003) used a recent study of Wu et al. (2020) observed sample composed of 617 first year higher sensitivity to risk and betrayal in students from Liege university to track women, by exploiting magnetic resonance potential differences in gender and imaging to investigate neural signatures. experience of failure (i.e., repeating an The results illustrated in neuroimaging are academic year): the outcome said that girls consistent with the previously analyzed were more subjected to society exigencies theory. Similar findings are provided at of studying and consequent major sense of medical level by Orsini et al. (2016), who anxiety and incompetence, while boys gave explanation of such differences reported higher scores in self-confidence drawing on instinctual reasons, thanks to but even higher tendency to procrastinate, the evidence emerging from an because failure expectancies could be experiment on rats’ behavior. The deeply harmful for their self-esteem. registered effect was not due to differences These findings explained why the rate of in shock reactivity, body weight or estrous male dropouts after the first academic year phase, so that the effect of interest was was superior, also justifying the female well isolated from exogenous influences. superior in terms of performances at A revolutionary work is proposed by Carr university. In the same way, males were and Steele (2010), who provided the first more likely to declare as first choice of evidence that decision-making process studies something near to short and less can be affected by concerns about difficult paths. Another quite recent paper stereotypes and identity devaluation. (Kluen et al., 2017) suggests that gender Indeed, rather than attribute gender differences in anxiety management affect differences to innate and stable factors risky decision-making situations: acute (e.g. biological and hormonal reasons), anxiety escalates cortisol activity, which they discovered that women subjected to stresses the risk attitude in men but not in stereotype threat in academic/business women, thus driving to diverse behaviors settings were more loss averse (i.e. less under pressure. Also Van den Bos et al. risk taker behavior) than both men and (2012) underlined differences between women not facing the threat of been men and women toward risk appetite, judged in light of negative prejudices. stating that women were more sensitive to Instead, no gender differences in risk occasional losses and, as consequence, appetite were found in absence of needed more time before reaching the stereotype threat. same level of performance if compared to Yet, a recent paper (Lee, Ashton, 2020) men. Stanton et al. (2010) tried to justify analyzes a wide sample of 347.192 persons such differences in risk taking by invoking 26 from 48 different countries to conclude entrepreneurship. They found that that women averaged higher than men in entrepreneurs’ technical background specific psychological features, such as moderated the gender effect, and women emotionality and honesty/humility. This received higher evaluation by the venture insight suggests gender as control variable capitalist when the assessment moment into the psychological influence of happened with close contact. When in personal traits. Stoet et al. (2013) detected presence of technical background of both differences in gender about multi-tasking male and female entrepreneurs, the VC activities: men suffer more when there is evaluation did not register variations necessity of handling multiple among genders; anyway, in situations commitments; however, the academics where technical background was absent underlined the lack of empirical studies on and prior performance information was gender differences in multitasking, so ambiguous, the female entrepreneurs advising for caution against making received lower evaluations than male non- generalisations. technical entrepreneurs, meaning that women were supposed to be less Zhao and Zhang (2016) focused on an competent and having less leadership interesting point: they found that people ability when available information was tend to trust strangers of opposite gender insufficient. more than those of same gender, and females trust females much more than males trust males. This key finding can suggest that male teams (absolute absence 1.4.5 Academic level and of females to mitigate contrasts) suffer more during the coordination phase, and background effect it can be more relevant in view of the An interesting recent randomized field scientific method, in which orchestrating experiment conducted by Chatterji et al. the feedback exchange activity is a core (2019) has investigated the advice effect: issue. Block et al. (2018) found that young the sample was a pool of 100 Indian high- boys endorsed communal values less and growth technology firms, whose founders agentic values more than girls, suggesting received advice from other entrepreneurs that gender differences in core values about people management. The emerge early in personal development entrepreneurs who received advice built and predict children’s expectations, even on formal approach (i.e. regular meetings, without receiving the influence from consistent goals, frequent feedback external environment and society rules; it exchange) grew more and were less likely drives to consider gender control as to fail; in addition, they found that crucial when analyzing patterns in entrepreneurs with MBA, accelerator psychological traits. Finally, looking at experience and similar, did not follow the venture capital world and corporate general pattern, suggesting that formal reliability, Tinkler et al. (2016) adopted an training limited the spread of peers’ experimental design to investigate advice. Such result can be extended since venture capitalists’ funding decisions scientific method put its own roots into the (VCs) in high-growth and high-tech capacity of listening and observing

27 feedback from environment and market - contacts network, facilitating and it looks like academic level and development and firm growth. After, an previous professional experience can entrepreneur with prior experience in weaken this capability - so that they might tasks regarding the same industry is able become useful as control variables during to better identify opportunities and holds the regression analysis. Furthermore, more odds of finding market gap. Also the Miozzo and Di Vito (2016) demonstrated researches of Pugliese et al. (2016) that entrepreneurs having more scientific confirmed that startups having more education, when the business idea was expert teams (referring to a specific launched into the market, faced difficulty industry) are prone to experiment higher in cognitive distance respect to their odds of success. These beliefs are customers, other than suffering for lack of confirmed by several studies presented in management practices knowledge. This literature (Gimeno et al., 1994; Tornikoski, implies that background differences can 2007; Preisendorfer et al., 2012; Cassar, affect the ways of firm conduction. Yet, the 2014), since the accumulated experience is website faculty.washington.edu reports an a discriminant highly difficult to replicate, article of Greenwald and Banaji (1995), and allows entrepreneurs to understand observing how different types of cultural competitive structure and market and national heritage, as well as the kind strengths, other than quality standards of university background, affect the and the more profitable trends. Moreover, decision-making process since it improves the capability of accurately entrepreneurs experience implicit estimating revenues and costs and differences, coming from diverse moral therefore performing more precise values, natural creativity and motivations. predictions, a useful outcome that It is of interest to notice that, according to diminishes the usually observed effects of Toma’s (2020) results, along some discouragement and delusion emerging dimensions (i.e. revenue, dropout) the after the initial instants. Fern et al. (2012) condition of STEM-majority team discovered that entrepreneurs tend to followed the same pattern tracked by the overly rely on their own historical network variables (advice from industry experiences, when shaping initial accelerator and incubator), implying how corporate strategies; anyway, those with startups having STEM majority in more different experiences showed less composition required for greater support pronounced bias and this is why an on business side offered by incubators and artificial variable (i.e. number of industries accelerators. in which experience is cultivated) has been taken into consideration. Looking at team

level, further facets are worthy to be 1.4.6 Prior experiences considered: influence of the offering type (Schleimer and Shulman, 2011), maybe weight due to products’ physical nature (design As declared by Vliamos and Tzeremes and creation phases) which suggests for (2012), the teams having previous higher scientificity values, and experience in the same sector can count on abstract/flexible nature of services suggesting for higher effectuative score. 28

Finally, there are the group decision- Mohs hardness, air quality), while the making rules, since Kaplan and Miller latter characterizes the scale’s qualitative (1987) concluded that into no-information nature. As displayed in Appendix A, in condition, decisions under majority and which a table extracted by Franceschini et unanimity rule did not differ, whereas in al. (2019) can be analysed, the ordinal scale complete-information condition the refuses consistency in distances (intervals) unanimous decisions fell farther to the and ratios, a limitation due to its low right, namely closer to the position of the nobility degree. In fact, although the low most extreme member. level, such scale accepts a number of transformations: starting from the most

refined scale, we view similarity, 1.4.7 Stevens’ scales exponents, linear function (e.g. translation, max-min normalization), Theory increasing monotonic function (e.g. According to the theoretical basis statistic cumulate transformation) and provided by Stanley Smith Stevens (1946), permutation, respectively introduced into it is mandatory to consider the kind of the scales of ratio, logarithmic interval, scale on which the numbers are expressed. linear interval, ordinal and nominal. It can Indeed, this thesis project used data and be demonstrated the transformations numerical values deriving from personal feasibility on ordinal scale, given the judgements, since they come from following enunciation: y > x, where for entrepreneurs’ auto-assessments and instance x assumes value 4 while y esteems made by the research assistants, assumes value 7, easily becoming 7 > 4. so that they might be subjected to biases Similarity (by assuming for example α=5): and distortions. This aspect is captured by φ φ the measurement scales theory, which (y) > (x) provides some properties and conditions α ∙ y > α ∙ x to better interpret any type of quantitative → characteristic, taking onto consideration 35 > 20 thus accepted. both its limits and reliability. Referring to Linear transformation (γ=6; δ=10): the data of interest in this study, they are μ μ mainly measured on a particular kind of (y) > (x) system: the ordinal scale. Such is due to γ ∙ y + δ > γ ∙ x + δ the fact that the totality of values → considered for analytical manipulation 52 > 34 thus accepted. came from subjective perceptions, so that Increasing monotonic function (e.g. the unique possible level of expression is cubical power): only at the second degree of ‘nobility’, y3 > x3 meaning it supports just for two types of relationship: non-equivalence and 343 > 64 → thus accepted. ordering, where the first is in common Permutation (items inversion): with nominal scales (i.e. those utilized for categories such as names and gender, or y ↔ x

29

7 ↔ 4 ordinal values of 2 and 3. As cited by Gioia et al. (2013), the qualitative research has 4 > 7 → thus refused. been critiqued as too often lacking in In conclusion, an ordinal scale has a scholarly rigor, so that the authors restricted set of available indicators used answered by confirming the richness and for centrality: just modal and median potential for discovery of such approach, values, whereas any kind of average indeed proposing a systematic summary (arithmetic, weighted, harmonic, and grounding theory articulation in geometric) is forbidden. A similar order to bring qualitative rigor while behavior is followed by indicators of presenting inductive research. Also Bloem dispersion: the logarithm of number of (2018) discussed about validity of cardinal classes along with fractiles are permitted, treatment on ordinal variables, asserting but variance and percentual variation are how robustness depends on the specific not reliable on this scale. Finally, statistics details of individual specifications, since it as chi-squared and verse tests are doable, deeply varies from a statistical setting to while t-student and f-fisher forbidden; another. Spearman correlation exists on ordinal scale, instead the Pearson correlation coefficient is not computable. Please note 1.4.8 Blau’s index that through statistical analysis and regressions, such forbidden operations As explained into the previous (e.g. variance computation, average, and paragraphs, the indicator proposed by so on) are however applied to the dataset, Blau is utilized in this work, since it has even if not completely corrected by a been adopted in a number of eminent theoretical standpoint. Although the papers and researches, other than boundaries surrounding the use of the analyzed and adapted in several ordinal data gathered within the dissertations (e.g., Herfindahl, 1950). It is experiment of interest, the goodness of the useful in order to simply detect the manipulations is however high, given the diversity, namely heterogeneity, along a professionality and experience of who wide range of diverse variables. It can be assigned evaluations (namely the research defined as: team), able to guarantee good reliability in N 2 BI = 1 - ∑i=1 pi assessments and increasing with the passing of time - thanks to learning economies, evolving mastery of the where 푝푖 indicates the proportion, arguments and standardized tasks expressed as percentage, of a certain type repetitively executed. In other words, the of the category compared to the total strong assumption concerns the fact that numerosity, so that the ones’ complement each research assistant has been able to returns the extent to which each single assess intervals between values by using component shows distance respect to the steady criterion, while valuating the others, in other words the dispersion of a ranking to be used; that is, for instance certain property among group members. considering the difference between 4 and 5 equal to the one perceived between the 30

The Blau’s indicator reminds of the 1.4.9 Research questions Herfindahl index, used by designers of antitrust policy toward mergers in role of By resuming the previous considerations, statistical indicator, indicating whether a the research questions presented in this merger should be challenged or not, seen paragraph will be examined throughout the increase in market power during the the thesis project. First of all, the studies of M&A processes. Both the indexes aim to Jehn et al. (1997) underlined the side quantify the entropy within a system effects brought by diversity in education: looking at variability in the components, it appeared to lead a certain degree of even if in different scopes of study. difficulty in internal conflicts management, even if coupled with Thereby the Blau’s indicator is included in performance improvement. Such regressions proposed at the end of this ambiguity is reported by a number of thesis project, and referred to a series of authors (Resnick et al., 1991; Nonaka, variables: diversity in team gender, 1995; Phillips & O’Reilly, 1998; Kristinsson seniority*1, geographical origin*, et al., 2015), who mentioned as core issue maximum academic level achieved, the presence of a dual effect: diverse academic level of the current studies, viewpoints and insights integration imply presence of other commitments, quality of more disruptive creativity and better student member, academic background*, decision-making outcomes, meanwhile prior experiences as startupper and in social and moral heterogeneities are business plan writing. As suggested in causes of negative influence through literature, the aggregated Blau index has interpersonal conflicts. to be defined as sum of the sub-indicators rather than arithmetic mean, so that its In other words, only the diversity at formula has been expressed in the “information level” positively affects the following manner: performance and is due to task conflicts. Yet, other scholars (Klotz et al., 2014; Fitzgerald et al., 2017; Zhang, 2019) C claimed how opposite effects are likely to Aggregated BI = ∑i BIi ∈ [0,C] be observed, seen the creativity creation leading benefits along with more difficult where C is the number of different Blau conflict management: the innate team indicators, so as to have the value equal to cohesion can be represented by the C representing the superior limit of the artificial variable called Internal Network, overall index domain. measuring the degree at which the team members have connections preexisting among them thanks to previous experience, both at work and academic levels. An interesting point is highlighted by Hoogendoorn et al. (2017), since they found that a balanced combination of

* These are all clustered variables, with the aim of generating homogeneous sub-groups, and comparable in terms of numerosity 31 higher and lower cognitive abilities levels the recent papers of Aven (2018) and (e.g. diversity in academic level) is largely Butler (2019): the first one, by analysing more productive. The reason is that some 9461 entrepreneurs and 2446 industrial people can be assigned to mundane tasks enterprises, showed that variation among (e.g. implementation) while others to the team members’ brokering ability ones calling for greater cognitive capacity; significantly predicted the starting capital consequently, they concluded that gained by the firms; further, when both performance of teams first increases then average and variance in brokering decreases with ability dispersion, and potential were high for each team, then average team ability is not related to team firms raised greater starting capital; performance. Thus, the first research therefore it seems that diversity in question aims to investigate whether the network ability predicts credibility in several Blau’s indicators, if separately front of investors. Instead, Butler (by considered, are statistically correlated analyzing data from CrunchBase and with variations in scientific and PWC, while intercepting network effectuative decision-making innate dimensions by scanning social pages from degree, with such effect controlled by LinkedIn) discovered that funding rounds variables incorporating interpersonal per year is a reliable indicator playing conflicts management, such as Internal significant and positive role in affecting Network and experience proxy (i.e. age, startup creation in certain locations, since previous work activities, years of local social network density generates experience in practical tasks). stickiness to the local environment; secondly, he found that midcareer Again, Geremias et al. (2020) recorded individuals (e.g. millennials) are more positive correlation between advise likely than early-career and late-career to network centrality and internal learning in create successful technology startups, young teams; Jurkevičienė et al. (2018) suggesting for the Age dimension to be invoked the social capital theory used as control variable. To sum up, the application to social network paradigm, second point is about observing the effect providing theoretical evidence for ties on scientificity and effectuation by using between entrepreneurial heterogeneity variables controlling for network and performance, since benefits are due to capability, such as heterogeneities in complementary in psychological traits seniority, students’ presence, academic when supported by right mix and level and background, as well as team overlapping. The connections importance numerosity and psychological capital is reported also into other works (Aldrich traits, mostly the Internal Network and and Kim, 2007; Reagans et al., 2004; Arif, Team Building dimensions. Please note 2015), who linked it to a superior capacity that the LinkedIn contacts counting of getting relevant results by exploiting (already proposed in recent studies) external knowledge and network should be biased in this experiment seen economies, as well as the possibility of the nature of participants, mostly young obtaining access to valuable extra and with few professional experience, and resources and information. Another way this the reason why it has been completely to look at network centrality is offered by neglected as proxy of external ties. 32

Looking with a focus on psychological which they anchor to their own beliefs, traits, Su-li and Ke-fan (2011) - by with such phenomenon more pronounced adopting the MBTI*2framework - stated in older individuals, thus it suggests for that, in terms of majority within the team, using seniority parameter to control 1) J were more prone to make optimistic variations in natural scientific degree estimates about entrepreneurial (given that scientific method is exactly opportunities than S; 2) P were likely to based on better interpreting external detect greater amount of opportunities signals). Shah and Tripsas (2007, 2012) and lower risks than J; 3) E felt lower risks took into consideration the possibility that than I. Therefore effectuative people see entrepreneur is even the user of product more opportunities and less risks, so that he offers on market, so they proposed the it can be controlled by the Risk Appetite idea of correlation between personal variable into the experiment subject of motivations and outcome. Wang (2019) interest in this thesis project. Clark and asserted how human (i.e. age and Wiesenfeld (2017) reported cases of education, work experience, near relatives companies which made strategic decisions models), relational (i.e. trustworthiness based on biased samples, more likely to and relationships among co-founders) and corroborate initial hypotheses, so that it strictly psychological (i.e. optimism, looks like such aspects could be included autonomy, hope, resilience) variables in a set of psychological dimensions such showed crucial for performance. Finally as Self Esteem and Novelty. York et al. Alessandri et all (2018) in their recent (2014) mentioned how acquiescence, studies concluded that absolute levels as selection, confirmation (also described well as increases in PsyCap (hope, into Gilbert’s essay published in 1991), autonomy, resilience, optimism) predicted overconfidence and optimism biases were work engagement increases, which in turn the most frequent and relevant ones, and predicted better job performances; this these biases can be tracked through the suggests for psychological variables seen use of the PsyCap variables considered through commitment level control. To within this experiment. Amit et al. (2001) sum up, the third research question further explained that the entrepreneurs investigates whether a correlation makes engaged in business planning reduced the sense between decision-making approach likelihood of venture disbanding while (natural degrees of effectuation and increasing the odds of pursuing the idea scientificity) and psychological capital, if with success, suggesting for looking at BP controlled by experience in business experience effect on decision-making. planning, age, initial motivations, Secondly, they demonstrated that younger industry and commitment level. entrepreneurs showed more Mitchell and Shepherd (2010) found that overconfidence bias with lower decision executives acting in more dynamic sectors comprehensiveness, so the effect should are more prone to take decisions be observed through age control and inconsistent with market and harmful for psychological traits. Parker (2006) corporate performances, so concluding affirmed entrepreneurs differ in the way

*(E)xtraversion vs (I)ntroversion as source of spiritual vigour, (F)eeling in decision-making attitudes, (J)udging and (P)erceiving in (S)ensing vs i(N)tuition as ways to grasp information, (T)hinking and adapting to external contingencies 33 that more dynamic and then risky process and business development industries (e.g., ICT) can lead to minor activities. In addition, as reported on the scientificity and in general to less efficient website faculty.washington.edu, decision-making processes. Schleimer and Greenwald and Banaji (1995) discovered Shulman (2011) cited that products are that different types of cultural and more profitable and its physical nature national heritage, and above all the kind of (design and creation phases) could university background, affect the suggest for higher scientificity, while decision-making through the influence of abstract nature of services and flexibility diverse moral values, natural creativity suggest for high effectuative aptitude. and motivations; such leads to think that Anna et al. (2000), in their studies looking geography and academic background at only-female ventures coming from US influence decision-making through initial states of Utah and Illinois, found that the motivations. Miozzo and Di Vito (2016) size remains small in terms of revenues affirmed that entrepreneurs having more and employees if compared to male- scientific education, when the business owned firms, as females tend to idea was launched into the market, faced concentrate on retail and service difficulty in cognitive distance respect to industries (where markets are smaller in their customers, other than suffering for opposition to high technology, lack of management practices knowledge: construction and manufacturing), so that background differences can affect the gender factor could control for sector. ways of firm conduction, so as to influence Secondly, traditional women business the natural aptitudes in decision-making. owners had largely different key success Chatterji et al. (2019) studied 100 tech factors than non-traditional business firms, whose founders received advice owners: the resulting conclusion appears from other entrepreneurs about people to be appropriate for being extended at management; who received advice built men level, keeping same differences on formal approach (i.e. regular meetings, among different industries. The fourth consistent goals, frequent feedback research question aims to investigate exchange) grew more and were less likely whether the industry to which the startup to fail; after, they found that entrepreneurs belongs can affect effectuative and with MBA (or accelerators or similar) did scientific natural aptitude, and this not follow general pattern (i.e. formal relationship could be controlled through training limited the spread of peers’ gender and kind of the offering. advice). Such result can be extended since scientific method put its own roots onto An extremely recent work by Toma (2020) the capacity of observing feedback from concluded that startups having STEM environment, so it appears that majority in composition required for excessively high academic level and greater support on business side (e.g. previous professional experience can offered by incubators and accelerators), weaken this capability. Finally the fifth thus the STEM majority boolean variable research question investigates whether the could be an appropriate factor to consider type of background (boolean variables as discriminant in order to detect about majority of STEM, Economics and significant differences in decision-making Other) and the academic level correlate 34 with the natural levels of scientificity and capability of performing more precise effectuation, when controlled by predictions, diminishing the usually geographical origin, initial motivations observed effects of discouragement and prior professional experiences emerging after the initial instants, so that (Chatterji et al., 2019). we expect experience influences decision- making thanks to a better comprehension Yet, Vliamos and Tzeremes (2012) got of the issue. The sixth research question is evidence that teams having previous about the possible correlation between experience in the same sector can count on previous experience and natural levels of contacts network, so facilitating the scientificity and effectuative mindset business development: team’s prior through a set of controls such as seniority, experience affects decision-making startup sector, number of sectors (Fern et through network (e.g. controlled by age). al., 2012) and innate overconfidence. After, they found that entrepreneurs with Please note that the prior experiences are prior experience regarding the same measured through the use of the following industry better identify opportunities, dimensions: years of past experience, therefore prior experience into the same experience in same industry, experience as sector in this experiment is expected to executive, the already having been affect effectuative approach as based on startupper, the number of prior companies opportunity recognition and exploitation. established, experience in business plan Pugliese et al. (2016) reported how writing, economics and management startups having more expert teams courses attended, entrepreneurship (referring to a specific industry) are prone courses taken, vertical and horizontal to experiment higher odds of success, so it competencies (or aptitudes) acquired is possible to conclude that experiences according to the experiences cultivated predict success through startup sector. into work and academic world. Furthermore, Fern et all (2012) in their researches affirmed that entrepreneurs The last point touches the gender bias. tend to overly rely on their own historical Masson et al. (2003) assessed a large set of industry experiences; anyway those with students, discovering that girls were more more different experiences showed less subjected to society exigencies of studying pronounced bias: this is the reason why and consequent major sense of anxiety the artificial variable described as number and incompetence, while boys reported of industries (in which experience is higher scores in self-confidence but even cultivated) should well predict anti- higher tendency to procrastinate as failure overconfidence aptitude. In addition, a expectancies could be deeply harmful for number of authors (Gimeno et al., 1994; their self-esteem: this leads to consider the Tornikoski, 2007; Preisendorfer et al., 2012; PsyCap effect as controlled by gender. Cassar, 2014) underlined how Such evidence further explained why the accumulated experience is difficult to rate of male dropouts after the first replicate, and allows entrepreneurs to academic year was superior, also understand competitive structure and justifying the female superior in terms of market strengths, quality standards and performances at university. Yet, males most profitable trends; it improves were more likely to declare as first choice

35 of studies something near to short and less were demonstrated to be more able in difficult paths, and such factor might anxiety management, resulting in higher contribute to gender differences in lucidity through complex decision academic level (male one should be moments, and this could justifies gender inferior). Shepherd (2012) stated that differences in PsyCap. In addition, the gender effect can vary depending on anxiety impact was more evident for men nature of industry (e.g., traditional or non- during early stage (i.e. exploration phase), traditional) in which entrepreneur plays, whereas stronger on women during the thus the gender impact is controlled exploitation phases of tasks; such maybe through the kind of offering and startup could result in worse men scientific industry. He invoked more attention on performance while worse women conditions under which gender effectuative performance. Frost, Marten differences are magnified, namely looking and Lahart (1990) asserted that a key role at women in non-traditional-for-women is played by the perfectionism trait, strictly industries; finally he asked for moving linked to fear of failure and anxiety focus from the amount to the nature of the generation, so they further proved in experience (e.g. failed attempts of startup medical terms the magnitude of some establishing), so confirmed the utility of gender differences. Wu et al. (2020) considering previous established registered higher sensitivity to risk and enterprises as control variable. Gatewood betrayal in women, by exploiting et al. (1995) obtained evidence that women magnetic resonance imaging to investigate who started having personal reasons (e.g. neural signatures: they provided proven autonomy) and men who started having difference demonstrated at neuroscientific external incentives (e.g. solve a perceived level by such extremely recent study, market need) were statistically more likely suggesting for particular attention on the to successfully terminate the initial phase Risk Appetite variable. Orsini et al. (2016) of firm development: it can be useful to found similar findings by medical investigate whether initial motivation standpoint: they explained gender drives to different outcomes through differences drawing on instinctual gender control. Lee and Ashton (2020) reasons, thanks to the evidence emerging analyzed a wide sample of 347192 persons from an experiment on rats’ behavior, from 48 different countries to conclude where the registered effect was not due to that women averaged higher than men in differences in shock reactivity, body specific psychological features (e.g. weight or estrous phase, so that the effect emotionality and honesty/humility), of interest was well isolated from suggesting for gender as control variable exogenous influences. This confirmed that into psychological influence of personal gender difference in Risk Appetite are traits. De Visser et all (2010) offered explained at instinctual level too. Carr and evidence by a medical point of view: Steele (2010), in contrast to a great amount decision-making and anxiety share of scholars, sustained that women underlying neural substrates, so that subjected to stereotype threat in variations in anxiety handling capability academic/business settings were more provoke variations in decision-making loss averse (i.e. less risk taker behavior) and cognitive functioning; indeed, women than both men and women not facing the 36 threat, while no gender differences in risk the effectuation paradigm as based on appetite were found in absence of flexibility (whereas scientific approach stereotype threat. This revolutionary follows a more linear scheme). As paper provided the first evidence that opposite to Carr and Steele (2010), Block et decision-making process can be affected all (2018) gathered evidence that young by concerns about stereotypes and boys endorsed communal values less and identity devaluation, rather than attribute agentic values more than girls, suggesting gender differences to innate and stable that gender differences in core values factors (e.g. biological and hormonal emerge early in personal development reasons). Apesteguia (2012) instead and predict children’s expectations, even offered a series of findings: without receiving the influence from external environment and society rules. i) teams formed by women Therefore the gender control appeared to totality are significantly make sense, against that sort of self- outperformed by all the other fulfilling prophecy suggested by Carr and gender combinations, letting to Steele; by the way, a certain degree of interpret that women- uncertainty keeps high in interpretating unanimity teams are maybe less the results and the possible crossing of focused on mere economic different effects and omitted variables. outcomes Tinkler et al. (2016) investigated venture ii) the best performing group is capitalists’ funding decisions in high- two men more 1 women for growth and high-tech entrepreneurship: MBA classes, perhaps due to the women resulted to receive higher mitigating role of women evaluation by VCs when the assessment presence moment happened with close contact. iii) women-unanimity teams were When in presence of technical background less aggressive in pricing of both male and female entrepreneurs, strategies, invested less in R&D the VC evaluation did not register but more in social variations among genders, but when sustainability, so that technical background was absent and differences in performance prior performance information could be explained by those in ambiguous, the female entrepreneurs decision-making and personal received lower evaluations than male non- traits (e.g. initial motivations technical entrepreneurs, sounding like and startup industry). women were supposed to be less Stoet (2013) found that men suffer more competent and having less leadership when there is necessity of handling ability when available information was multiple commitments, however he insufficient. Such paper could let to intend underlined the lack of empirical studies on that entrepreneurs’ academic background gender differences in multitasking, so moderated gender effect. Kluen et al. advising for caution against (2017) claimed that gender differences in generalisations – this creates room to anxiety management affect risky decision- investigate gender difference in multi- making situations since acute anxiety tasking management, maybe relevant into escalates cortisol activity, which stresses 37 the risk attitude in men but not in women, established ventures, experience in same thus driving to diverse behaviors under sector), offering kind, startup industry, pressure; that is a strong evidence that academic background. Please notice that gender differences in risky decision- the effectuative approach should be making are explained as hormonal particularly interested as highlighted by reaction too. Van den Bos (2012) the studies (Van den Bos, et al., 2012) on underlined gender differences toward risk the sensitivity to occasional losses, since appetite, stating that women were more right such facet is a core component of the sensitive to occasional losses and, as effectuative framework which also consequence, needed more time before includes network, execution, flexibility reaching the same level of performance if and control abilities. compared to men: so he justified women’s minor Risk Appetite aptitude by using sensitivity to loss. Stanton et al. (2010) 2. The research justifies gender differences in risk taking by invoking hormonal components, program and its namely the endogenous level of testosterone: high testosterone levels scope seemed being correlated with greater risk- This thesis project draws on a RCT taking attitude, and this was true both in (Randomized & Controlled Trial) study men and women individuals, so offering realized at the turn of 2020 and 2021 in further hormonal explanation about Risk Italy and named InnoVentureLab (IVL), Appetite differences. Finally, Zhao and born thanks to the partnership between Zhang (2016) reported that people tend to ICRIOS research centre of Bocconi trust strangers of opposite gender more University, Politecnico di Torino and than those of same gender, and females Politecnico di Milano. As described in trust females much more than males trust Bacco et al. (2020), the IVL program males. So an emerged interesting point is focuses on how entrepreneurs make that male teams (i.e., absolute absence of decisions under conditions of high females mitigating contrasts) suffer more uncertainty. The purpose is to extend the during coordination phase, and is perhaps prior few works which showed how more relevant into the scientific method, entrepreneurs can improve ability to make in which orchestrating feedback exchange key decisions for business development activity is a core issue. In conclusion, the by adopting a set of practices labelled last and seventh research question aims to 'scientific approach'. As explained more investigate whether gender*3moderates times in other chapters, scientific PsyCap (focused on Risk Appetite), approach is a set of rules based on strong academic level and initial motivation emulation of what scientists do. When effects on natural levels of scientificity and entrepreneurs use this approach, they effectuation, while interacting with accurately frame the problem they face, experience proxies (e.g. previous complying to the following steps:

*Blau_gender along with boolean categories such as female majority, female unanimity, at least one women 38 articulating theories, defining testable components of an entrepreneurial hypotheses, and conducting well setted idea and its business model; tests while making thoughtful interpretations. A randomized controlled - effective ways to do polls, trial (RCT) was conducted to obtain robust interviews and surveys, other than evidence and test the impact of scientific the respective best practices, with approach vis-a-vis another popular the objective of making data approach to decision-making process, the collection without personal, socio- so-called effectuation. When using such so economic and psychological biases, different mindset, entrepreneurs are while choosing the right pool of expected to exploit a non-predictive respondents (i.e., absence of approach aimed to define the needed steps selection and auto-selection biases) by gauging what resources they have. and finally analysing results by Building on other two pilot RCT studies considering the most appropriate realized in Italy (and counting 116 and 250 criteria; startuppers), IVL has the objective to extend the results over as many as 500 - Minimum Viable Product (MVP), entrepreneurs of new startups, measuring essential means useful for pursuing the diverse effects generated by both only the worthy ideas since it effectuation and scientific approaches to allows to conduct tests and decision-making, while deeply minimize resource investment investigating if they differently act by while making easier the modifying the boundary conditions. information exchange between startup team and its own The program got started on 1st May 2020 environment; and scheduled to end on 28th February

2022: it provides entrepreneurs with eight - concierge/prototype (respectively different sessions of training that, seen the referred to service/product) pandemic situation, have been supplied conception, both aimed to facing with mandatory online mode. Each the real market needs by creating a session is composed of interactive lectures, pilot version following right along with coaching lesson by qualified procedures, timing and mentors and instructors, everyone requirement-feature balance. working with a sub-group of the entire sample. Both treated and control startups received the identical number of lessons Given the alike contents proposed in both about entrepreneurship, where the treatment and control group, the primary topics are: difference took place in the scientific routines applied to entrepreneurial - BMC (Business Model Canvas), decision-making framework taught only useful and widespread tool in the first treatment class, meanwhile a introduced by Osterwalder and specific teaching has been done on the Pigneur (2005), it is used to other treated group, by using the efficiently frame the various flexibility provisions suggested by

39 effectuative method. The control group modifying the core value could have the possibility to receive proposition. learning insights neither about scientific nor about effectuation approaches. A final The whole research examines other sub-group of entrepreneurs did not potential dynamics too: e.g., gathering receive any training. All the groups values on variables strictly related to received exactly the same number of hours timing and means utilized in in training, in order to ensure entrepreneurial decision-making process. comparability in terms of pathways and In particular, the precision in predictions therefore of results. On American and esteems results as a key aspect, so that Economic Association registry website - it is observed in the starting instant of time socialscienceregistry.org - it is also possible and even throughout the course to consult official details and future development, since the expected and developments about the InnoVentureLab desirable effect is its pronounced program. There, the major outcomes and improvement due to scientific treatment. research questions (final purpose of the Furthermore, other co-variables have been whole research study) are specified: gathered about several features, useful to be employed in role of covariates interacting with treatments: gender, i) income flows - represent the psychological traits, perceived main dependent variable and competitiveness, information sharing, are measured as € (euros); management practices, wellbeing, insight

accumulated and developed, passion and

communication, as well as common ii) dropout – binary variable, knowledge within the team, mentors’ assuming value 0 until the firm influence, prior experience and abandons the learning program work/academic background. and entrepreneur ceases the startup activities (severe controls have to be made in order to ensure the actual 2.1 Sampling and design venture ending), while assuming value 1 right when experiment the startup concretely drops The RCT study in progress at the turn of out; 2020 and 2021 considers uniquely the nascent entrepreneurs, meant as those starting a new business at the moment, iii) pivot – meant as cumulative while there are no restrictions about the amount of times in which an industry belonging. IVL was promoted on entrepreneur makes relevant numerous digital channels, in terms of a changes to business model, general course able to provide participants where “relevance” arises every with useful insights on creation of new time he moved from original to innovative enterprises. In addition, both another business idea by sign up and participation are completely 40 free of charge, having the purpose of 2.2 Why RCT design catching entrepreneurs with limited economic resources. Multiple instructors As underlined into the prior paragraphs, received specific training before the the whole IVL program experiment bases program delivery since the teaching on a randomized and controlled design, materials they used were accurately because of the particular purpose: the designed for that precise objective. Each observations have to be accurate and able instructor teaches three classes (i.e. to discern between the different effects, so scientific treatment, effectuation that a rigorous setting is needed to provide treatment, control sub-set). Meanwhile the robust results not affected by distortions research team - of which I took part in role and disturbing influences. Indeed, the of research assistant (RA) - has been in dataset naturally exists as not- charge of designing the time evolution of experimental data (i.e. not ideal shape), the activities, while coordinating and since the most frequents working overseeing them, so as to ensure the conditions are under observational correct progress of taught modules and setting: the analyses come from real coaching sessions, avoiding technical and behavior observation, so an empirical management issues whereas instructors study leading to a number of risks, namely carry out their lessons. The sample omitted variables, simultaneous causality composition was made recurring to pure (ambiguous verse in cause-effect relation) parallel randomization model, supported and presence of correlations not by STATA statistical software. The necessarily implying causality. definitive resulting sets are four: scientific Such complications call for precise treatment, effectuation treatment, control procedures to follow (Stock & Watson, (i.e. neutral training) and pure control (i.e. 2012), first of all at experiment level. A no training). After startups being causal (and not a casual) effect can be randomly assigned into one of the four effectively well measured only in groups, multiple sub-groups of 35 startups restricted situations, such as the following were randomly matched with the coaches, features: so that each coach got three sub-sets of startups. The reference unit during the • experimental design - the randomization process was the solo participants have not the possibility startup, while randomization to choose the treated individuals, discriminants were comparison among ignoring the existence of different mean values and t-test across groups; groups under different conditions; minimum detectable effect size was observed for the main outcomes. The • controlled – the research team, overall sample extent counts 500 startups, including RAs, is in charge of homogeneously divided into the assigning the treatments while following groups: 125 in the scientific selecting the control and pure group, 125 in the effectuative group, 125 control sets, aiming to gauge startups in the control group (neutral differential effect between training) and the last 125 in the pure treatment and absence of control one (absence of training). treatment; 41

Watson (2012): they proposed the • randomized, even named situation in which the experiment consists casualized – to be meant not as of observation on several cultivated fields, literally casual, that is a common where the treatment is about the fertilizing misinterpretation, but in terms of power; therefore, the farmer, in role of equiprobability, thus the treatment research team, does selection with the received assignment by respecting purpose of deciding whether a certain random criterion in order to avoid field will be treated with fertilizer or not. systematic correlation between Thus, the issue arises in the moment when characteristics which are external to some lands, while others not, are strongly the observation, often linked to hit by sunlight during the morning time: environment and specific this is a classic example of variable due to belonging to group (e.g. such group belonging, having the power to barrier provokes that, in case the affect final outcomes, given that research team could not choose the improvement in field productivity and distribution of startups among sub- other outcomes keep relevant groups, it would have been ambiguousness about which is the real impossible to compute the pulled reason provoking the differences in the standard deviation of the reference used metrics, making more difficult to population, given the potential infer in robust statistical terms. belonging of startups to diverse An effective answer to the presented populations); problems suggested for precise

procedures to follow during an • pseudo ideal – meaning everyone econometric experiment: the scholar, first follows the program protocol, by of all, must chose the hypotheses to verify, entirely complying to the imposed that is construction of economic theory rules, delivering correct reports, even drawing on prior evidences following the scheduled timing and emerging from scientific literature; here, so on. the independent variables are selected along with their internal relations. So, it comes to specification of econometric Such care is due to necessity of being sure model, consisting of assumptions on to measure the effect generated by real the regressors’ nature and their bond to regressors, that is, with no external residual error, leading to functional form influences which may interfere into the generation too, and conjectures about relationships of interest: these are nature and probability distribution of the undesired components affecting the residual errors. Indeed, it is useful to experiment, and may derive from other remind that the residual component is an features other than the observed ones, like aleatory variable, needing of be defined by dynamics tied to belonging to a group expected value and statistical distribution, rather than another, and any other while considering it could include omitted environmental factor. An explicative factors affecting the dependent variable, instance is offered, again, by Stock and 42 other than errors born in measuring the obtained by minimizing the sum of the values of dependent variable. squared errors, where error stands for difference between actual value (observed Yet, the next step is data collection, with thanks to empirical experience) and the aim of calculating unknown predicted value, inferred by the regression population parameters by exploiting model. OLS is not distorted and observations on sample and discerning consistent. To have correct OLS between the different types of dataset (i.e. estimators, the necessary and sufficient cross-sectional, panel and time series); condition imposes truthless about the so- then, the scholar proceeds to the called least squares’ assumptions. Firstly, the quantitative esteems of econometric residual error’s probability distribution model by utilizing the available means, conditioned to the independent variable such as the Ordinary Least Squares must have null mean value, namely the method (OLS), in order to obtain the estimator is not statistically distorted, that estimators’ values while assuming simple is always true in an ideal randomized casual sampling. Finally, it comes to firstly controlled experiment (e.g. the research specification assessment, controlling for team decides which participants belong to consistence among assumptions and which groups); in other words, the economic data, namely verifying that no residual errors are independent respect to relevant regressors have been excluded, regressors, given that they are definitely whereas controlling ex-post for the nature random and then get minimal extent, so of functional form, residual error’s nature null average value. and regressors’ meaning. Secondly, the last control is done on the correlation Secondly, regressor and dependent verse, being consistent with economic variable for each observation are theory, and in conclusion the econometric identically (units selected from the same model is ready to test the desired population) ed independently (units hypotheses, as well as make predictions casually selected so that regressor and and simulate alternative scenarios. dependent variable of different units do not get mutual influence) distributed - and

this is surely true when using simple random sampling. Such assumptions can be never verified if the experiment 2.3 Gauss-Markov registers variations of the same unit through the time, in presence of time- Theorem: the OLS invariant factors. Finally, the outliers both efficiency in regressor and dependent variable must be rare, that is in technical terms having An Ordinary Least Squares model (OLS) is finite fourth moments: if not verified, the adopted into the linear regressions estimator of the population parameter is reported in this thesis project since - other inconsistent, so becomes important to than being the most popular and carefully understand whether the widespread means used in econometric distortion is due to wrong measures, data field - it appears to be the best in such not belonging to the right dataset, experiment design. The OLS estimator is 43 codification error, and so on. In addition, teachers, instructors, academic researchers two further assumptions usually add to and others. The involvement of a such the basic set, though they represent great amount of individuals called for unlikely cases observable into the real high structured activities, along with well- world: homoscedastic residual error defined procedures and use of dedicated (Frisch, 1926) and residual following tools. Additionally, the whole program Gaussian probability distribution, unified the efforts of three prestigious characterized with parameter σ2 as universities: Bocconi University, variance and null in average. Politecnico di Torino and Politecnico di Milano, all avantgarde centres in several Thus the Gauss-Markov theorem asserts fields of knowledge, from economics that - if the first four assumptions are passing by technological subjects such as considered true within the design engineering and architecture. As in any experiment - then the OLS estimator large organization, the coordination is a becomes the most efficient one among all core point needed to the effectiveness of the linear estimators, where efficiency gets the overall work. The program has been the meaning of minimum variance, and carried out in different cities and countries linear estimator stands for linear function throughout the time, lately in Italy and UK of dependent variables for each (London), whereas this year two parallel observation. Instead, if the totality of five flows are conducted at the turn of 2020 assumptions is true, then OLS obtains the and 2021: one in Italy and another lowest variance among all the consistent simultaneously in India. One of the first statistical estimators (even if not linear), if activities requiring a well-orchestrated the sample counts infinitely great coordination was the marketing journey, numerosity. Anyway, the OLS model thought for the IVL program on different would be highly sensitive to outliers, if digital channels; obviously, the vis-à-vis compared with other estimators, where advertising in other athenaeums, or in sensitivity stands for major variance; so high school environments and that, in presence of numerous outliers it is professional clusters, was forbidden given advisable recurring to alternative position the challenging situation of global indicator (such as median value) in order pandemic. to significantly observe less variance, reminding that the more average and So each kind of promotion activity was median are near, the more unlikely are the conducted online, avoiding direct contact outliers. during the initial phase as successively during data collection: all the people

involved have been divided into sub- 2.4 Marketing and groups, with the aim of using different social networks and online channels to sponsorship push the program toward the future The InnoVentureLab program has been entrepreneurs, and the utilized means conducted thanks to the coordination and have showed effectiveness seen that the support of a number of people, including amount of entrepreneurs participating to RAs as well as PhD students, university IVL resulted sizable. The program 44 gathered a large pool of participants since particular segment (early-stage startup), it was free of charge, hosted ventures through online platforms as Facebook, belonging to any industry and called for Instagram, LinkedIn, specialized websites, entrepreneurs coming from south, centre direct contacts, incubators, accelerators and north of Italy (even if only early-stage and co-working centres established all startups), a condition made possible by the along the Italian territory. Particularly, pandemic situation, which allowed to the reference social channels have been overtake the barriers once represented by chosen according to carefully assessment, physical presence and travel time. High based on social network’s nature and kind sampling numerosity and heterogenous of enrolees. composition are both a key factor for the LinkedIn was strongly considered because effectiveness of a RCT experiment, of its working nature and, above all, remembering that the main objective is to because its enrolees usually use it mainly indagate correlation between internal-to- for professional objective, so that, the team features and scientific decision- contacts kept on such platforms and type making as well as effectuative decision- of shared contents are in line with the making, and their influence on final scope of IVL program and marketing outcomes in performance, while campaign. LinkedIn was exactly born with controlling for potential differences by the objective of making network varying boundary conditions. The development, able to tie entrepreneurs marketing campaign got started in the deriving from any background with summer of 2020, seen that the program workers, professionals, employees, begin was scheduled in October 2020. The accomplished firms and institutions. marketing activities were first of all Furthermore, Instagram was selected divided taking into consideration the RAs’ given the great popularity and massive belonging to one of the three involved presence of young people, including universities, as the coordination would students and new ventures founders, have been easier. Some cloud platforms which could be reached by our advertising and project management apps were in an effective way through the wide range needed during the campaign: Google of social tools provided by the platform, Drive and Dropbox for file sharing, such as promoted posts, Ig stories and whereas Doodle, Slack and Trello were targeted advertising campaigns. Finally, used as calendars and effective Facebook was the main channel to scheduling, other than task checking and promote the activities, since it gave the assigning. Yet, given the pandemic, some possibility of diffusing highly segmented online platforms have been fundamental contents on various dedicated groups, if for the program success, such as Skype, accurately located in the search bar. Zoom, Google Meet and Microsoft Teams, Indeed, the great amount of effort useful to organize lessons as well as employed in such activity lead to find a lot conference and meeting. The summery of groups about topics of interest, such as marketing campaign purpose was entrepreneurship, technology, innovation, achievement and activation of an enough but also early-stage startups, incubator more widespread target, compared to that environment, innovative accelerators, obtained in the prior years: it focused on a 45 ideas’ exchange, network research and, working as trait d’union platform between more in general, attempts to connect big companies and the best highly different points of view and human technological startups. Startup Grind is a backgrounds aimed to create new and further protagonist in the partnership disruptive power. network created by IVL: it resulted helpful given that represents one of the widest The social network campaign was carried communities grouping students, investors out by publicizing a series of dedicated and entrepreneurs. A large pool of contents on each platform, following a different types of collaborations also pre-setted scheduling; the used means includes Start Up Legal, which is involved were a set of social pages, named in legal and financial support addressing InnoVentureLab, expressly created in startup and young ventures, with order to promote the event and make the assistance throughout the entire enterprise name ascribable to something new: an lifespan, from born to funders entry. Yet, innovative program offering innovative VGen was selected as further partner since type of contents organized in innovative it is employed in open innovation way. The social pages were coordinated environment, while connecting great and in line with a common design (e.g., companies and young students through same use of house colour), because of the an innovative and continuously diffusing necessity of making immediately tool nowadays, that is the virtual recognizable the utilized template and internship. In addition, the IVL marketing brand. campaign was boosted also thanks to the The first step of internal organization was contribution offered by large online referred to create a certain number of task- authors and forums (i.e. 100.000+ force teams, in order to foster an easier followers on social pages), so that the right coordination in smaller sub-groups amount of visibility could be guaranteed divided according to university through digital channels during the belonging: in each team, the individuals reduced time horizon needed to gather the accounted for a specific segment on a necessary subscriptions. specific channel, so that avoiding overlapping and duplicated activities, which could be perceived as undesired 2.5 The data collection spamming, that is an effect to absolutely prevent. Among research subgroup The first step was referred to the measure members, the social network campaign of the startup natural orientation towards was divided. The great amount of contents scientificity and effectuation in decision- shared via digital marketing allowed to making process, other than potential reach great players belonging to variations due to the effect produced by entrepreneurship: for instance, Plug&Play boundary conditions. Then, the research enterprise was named as official partner of assistants proceeded to the collection of a IVL program, because of its fundamental number of different types of data, commitment at international level in referring both to team composition and encouraging technology development by leader’s features, psychological and following the most favourable trends, orientation traits. Yet, were tracked some 46 qualitative characteristics of venture and respondents: they range from leader’s perception - such as proposed demographical and registry items to offering, reference industry, working and questions focused on prior experiences, academic previous experience, academic from academic background to the background, potential prior direct dedicated effort on entrepreneurial tasks, collaborations among team members - as from the offering kind to the other well as quantitative ones - as team activities in which the team members are numerosity, hours invested on startup, involved, from previous collaborations perceived weight of each business aspect among co-founders up to the key features on the entrepreneurial success. in which the team is expected to Additionally, all the demographic distinguish if compared to competitors. features were detected, in order gender, Instead, the final long displayed list was average team age, detailed geographical created with the objective of capturing provenance. Every variable previously psychological and aptitudinal orientations explicited has been captured throughout of team leaders as well as startups at all the program, that is, in several overall level, by asking indirect and non- successive instants of time in order to track tendentious queries about their behaviors development. The used tools were mainly during the real-life situations of a typical survey, telephone interview and Qualtrics entrepreneur. The variables reported in platform, all items generated thanks to the Figure 1 provided personal insights on contribute of research team, and pursued participants, in order to make possible a by the whole pool of research assistants. successive analysis about internal The first data gathering started at the end differences among the demographical of summer 2020 counting on survey and aspects, other than using them as telephone one-to-one interviews. The first controlling variables for secondary effects tranche of information was got during the in case the environmental influences arise. finalisation of the registration step, using The team numerosity was detected to be online questionnaires and resume used as verification of composition and as templates. Then, after the selection phase, index about the components of the RAs contacted all the team leaders by heterogeneity indicators, other than in phone in order to gather information gauging the total commitment of hours about orientation towards scientificity and assigned to the entrepenurial occupation. effectuation in decision-making registered After, each entrepreneur should talk about at time zero, with the purpose of the kind of offering treated in his own understand and analyse the natural business, by choosing from a range aptitudes before of treatment, while filling including product, service and Other (in up the vacations left into the online this case, it was requested for further questionnaires. specification into the following question) As it is possible to view in Figure 1 in options. Radio button answer mode was order (from left to right, from top to often used to try of better categorizing the bottom), the first Pre-Survey proposed as gathered answers, namely create highly many as 121 questions aimed to intercept structured template for the experiment a lot of different information about the design. Yet, the entrepreneurs had to

47 provide the average amount of weekly specific subject between the classes hours invested by each team member on previously categorized; startup work, so data are simple numbers; moreover, the respondent gave • Doctorate - boolean and, if Yes was information about potential other the answer, it was asked about commitments added to the business idea specific subject between the classes treated in IVL, with the aim of exploiting previously categorized. such insight to measure the extent of commitment employed on Thus, it was asked to specify whether entrepreneurial activities. Then, a series of other occupations were present for each background parameters are asked: respondent (selecting between full-time and part-time too); so, further information about kind of work was requested to • current studies – to be chosen insert. Additionally, a number of between bachelor degree, master questions focused on prior work degree, master in business and experiences and their industry belonging, administration, doctorate and other along with those on prior roles of specializing courses post-lauream; executives, previous established firms and

relative business planning activities. Yet, • main subject of the studies – to be some points involved possible economics, chosen between architecture & management and/or entrepreneurship design, physics and maths, education courses already attended, while economics & management, ICT, controlling for the horizontal (or vertical) engineering, law, medicine and nature of academic studies. Six questions biology, political and social were about previous direct collaborations sciences, historical and among team members, both at university philosophical sciences, and Others and work levels, before of asking about the (in this case, successive key competitive and differentiating specification was mandatory); factors perceived by business owners. In

conclusion, the last 48 queries tried to • Bachelor of Science – boolean and, intercept some psychological and personal if Yes was the answer, it was asked traits of the team decision-makers, each about specific subject between the one underlying to specific more abstract classes previously categorized; factors to be analysed in view of

scientific/effectuative orientations. Such • Master of Science – boolean and, if last questions are all expressed on 7-points Yes was the answer, it was asked Likert’s scale, as suggested in the about specific subject between the approved and largely diffused Likert and classes previously categorized; Murphy (1932) essay, where zero value

stands for complete disagree while the • MBA & similar - boolean and, if Yes seven score stands for complete was the answer, it was asked about agreement.

48

In Figure 2, instead, the outcomes deriving personal resources such as network, from on-phone interviews made by RAs abilities, passion and background) until are reported: the variables are expressed pilot plane capability (controlling and on 5-points Likert’s scale, where value 1 executing instead of waiting for predicted stands for minimum level in the events), passing through affordable loss considered characteristic, while 5 (focus on the maximum available), crazy represents the maximum degree for the quilt (proactivity in keeping contacts with respective characteristic, and value zero is customers, suppliers, competitors) and chosen when the feature does not exhibit. lemonade (exploiting unexpected As explained throughout the prior situations counting on flexibility and paragraphs, the scientific approach preexisting resources) aptitudes. Still, for consists of four primary phases each of the 5 discussed macro categories, (highlighted with light green colour), the simple arithmetic mean has been being theory formulation, hypotheses computed and the overall average of the statement, test conduction and validation, averages is calculated, with the objective along with a precise and consistent of summarizing the whole information quantitative threshold useful at the crucial about effectuative propensity into just one moment of choosing between pursuing quantitative indicator. the entrepreneurial idea, pivoting and abandoning it. These are all steps to strictly follow, by using well-structured paths and rigorous tools, right like a scholar or scientific researcher. For each of the 5 illustrated macro categories, the simple arithmetic mean has been computed and, at overall level, the average of the averages is calculated in order to resume the totality of information on scientific inclination into a single indicator, to be compared with the other variables of interest subject of study in the experiment.

Meanwhile, also the variables underlying for effectuation approach to decision- making process are displayed (showed with red colour). Again, they are expressed on 5-points Likert’s scale, where value 1 stands for minimum level in the considered characteristic, while 5 represents the maximum degree and value zero is chosen when the feature does not exhibit. The variables range from bird in hand attitude (i.e., exploiting the owned

49

startup name PhD relevance in your business-offering usability I will found a firm in order to become rich

name PhD subject relevance in your business-offering design I will found a firm in order to move up into the business world

surname PhD subject-specify relevance in your business-offering other features I will found a firm in order to solve a specific problem faced by people with which I strongly identify relevance in your business-offering other features (specify and give I will found a firm in order to have a proactive role in shaping the activities of people with which I CF currently you are: mark) strongly identify gender work kind I can predict my corporate's market demand I will found a firm in order to solve a social problem which private firms are not usually able to face

birth year work kind-specify I can carefully predict when bigger competitors will enter my market I will found a firm in order to have a proactive role in changing the way by which the world acts In role of founder, it will be very important for me to manage my firm basing on robust management birth country working experience years I can make my corporate successful even if others could fail practices In role of founder, it will be very important for me to deeply analyse financial predictions of my domicile country working experience years - startup industry I like to experience bold actions dealing with unknown business In role of founder, it will be very important for me to supply offering useful to people with which I domicile region working experience industry I would invest time and/or money on initiatives with potential high yield strongly identify In role of founder, it will be very important for me to show my customers that I agree with their domicile province working experience industry - specify I tend to act with bravery in high risk situations opinions, interests and values domicile city working experience years as executive I like to experience new activities but not necessarily risky In role of founder, it will be very important for me to be highly conscientious world citizen When involved in projects I prefer testing unique approaches rather than REP boolean already established other companies before entering in such startup In role of founder, it will be very important for me to make the world a better place reconsider the already used ones In order to learn I prefer testing personal ways rather than those used When I will handle my corporate, it will be very important to focus on what my firm can obtain if REP name #companies established before entering in such startup by the others compared with competitors I prefer a new problem-solving approach rather than approaches When I will handle my corporate, it will be very important to establish a strong competitive REP phone experience in Business Plan writing already used by myself or others advantage on competitors When I will handle my corporate, it will be very important to focus on people with which I strongly startup #members attended academic courses about economics/management I usually act in order to avoid future issues, needs or changes identify When I will handle my corporate, it will be very important to support people with which I strongly offering attended academic courses about entrepreneurship I tend to plan my projects in advance identify I prefer to personally carry out the projects in which involved rather than When I will handle my corporate, it will be very important to focus on what my firm can do for social offering-specify my academic competences are highly specialized in a certain field waiting for someone else doing it welfare When I will handle my corporate, it will be very important to persuade others that private firms are weekly hours on startup during my academic path I developed equal competences in a number of fields I tend to face with challenging working tasks which can teach me a lot able to face social problems like those challenged by my firm thanks to my academic studies I am able to complete few tasks but with great I continuously search for opportunities useful to develop new Please indicate how much effort you employed in understanding decision processes of the other current other commitment mastery capabilities and knowledge team members Please indicate how much effort you employed in using devices appropriate to remotely work studying thanks to my academic studies I am able to complete many different tasks I like difficult work tasks by which I can develop new capabilities together on the business idea did you attend, even if in different periods, the same university of at least one of the For me, the possibility to develop work abilities is so important to take a subject Please indicate how much effort you employed in work together in presence on the business idea other team members? risk Please indicate how much effort you employed in generating a common dictionary with the other subject-specify who? I prefer to work in situations calling for high capability and talent team members did you work, even if in different periods, in the same corporate where at least one I deeply care about to demonstrate that I can achieve better outcomes if B.Sc. do you know other entrepreneurs candidates to InnoVentureLab? of the other team members worked? compared with colleagues I try to understand what is needed to demonstrate my capabilities when B.Sc. subject who? who? I work before working on this business idea, did you have work or study collaborations B.Sc. subject-specify I like when the other colleagues are aware of my well working Did a participant of TheStartupTraining or TheStartupLab suggest our program to you? with another team member? I prefer to be involved in projects where I can demonstrate my M.Sc. who? who? capabilities to others I prefer to avoid new tasks if I could appear incompetent if compared to M.Sc. Subject dimensions in which your business idea distinguishes if compared with similar Do you remember 3 books (about business and/or startups) which particularly influenced you? the others For me, it is more important to avoid showing low ability than learning M.Sc. subject-specify Other features of the offering - specify If yes, please indicate the top 3 list - number 1 something new I worry about starting a new working activity if my outcomes could Master relevance in your business-price/cost of the offering If yes, please indicate the top 3 list - number 2 demonstrate I have low competences Master subject relevance in your business-offering quality I prefer to avoid work situations in which I could get bad results If yes, please indicate the top 3 list - number 3

Master subject-specify

Figure 1: first pre survey - script

50

VARIABLE DESCRIPTION CODIFICATION VARIABLE DESCRIPTION CODIFICATION VARIABLE DESCRIPTION CODIFICATION

Data intervista date Teoria_evidenza 0-5 Val_alternativa I dati raccolti aiutano a stimare il valore della 0-5 Data della chiamate La teoria ha dei dati a supporto componente alternativa a quella testata Intervistatore Nome e Cognome della persona che fa la string Teoria_modulare La teoria scompone il problema in sotto- 0-5 Val_negativa I risultati negativi dei test permettono di capire 0-5 chiamata problemi da risolvere nuove possibilità di esplorazione Startup string Teoria_gerarchia 0-5 Decisione_soglia Se la decisione di 0-5 1. continuare o abbandonare il progetto è stata presa confrontando la stima del valore dell’idea La teoria aiuta a prioritizzare i problemi da Nome startup con una soglia minima risolvere 2. modificare il progetto è stata presa confrontando la stima del valore dell’idea con una soglia minima Referente string Ipo_esplicite Elenca le ipotesi che intende testare in modo 0-5 Decisione_soglia_calibrata La soglia tiene conto della qualità dei test e del 0-5 Nome referente (persona intervistata) esplicito tipo di dato raccolto Numero di ore lavorative Numero di ore medie che ciascun membro del number Ipo_coerenti 0-5 Bird_in_hand_whoare Misura in cui sviluppano l’idea partendo da chi 0-5 Le ipotesi sono derivate dalla teoria team dedica alla startup settimanalmente sono, ossia dalle proprie abilità e capacità Clear_definition_roles 1-5 Ipo_precise 0-5 Bird_in_hand_whoknow Misura in cui sviluppano l’idea partendo da chi 0-5 Le ipotesi sono formulate in modo da testare Hanno una chiara divisione dei ruoli? conoscono, ossia dalla propria famiglia, amici, una cosa alla volta network lavorativo Definition of milestones 1-5 Ipo_falsificabili Sono in grado di stabilire una condizione 0-5 Bird_in_hand_whatknow Misura in cui sviluppano l’idea partendo da 0-5 Hanno obiettivi chiari secondo cui organizzano (soglia) in base alla quale le ipotesi possono cosa conoscono, ossia dal proprio background il lavoro? essere considerate supportate o meno e esperienza Decision_maker string Ipo_testabili Le ipotesi sono formulate in modo che le 0-5 Affordable_loss_max 0-5 variabili da testare possano essere L’imprenditore ha usato il massimo delle risorse Chi è il principale decision-maker (ruolo, nome) operazionalizzate (=trasformate in misure) che può permettersi di perdere correttamente Gerarchia (SI/NO) boolean Ipo_alternativa Le ipotesi erano mirate a falsificare una cosa e 0-5 Affordable_loss_risk 0-5 Prendono le decisioni seguendo una L’imprenditore non ha aggiunto risorse (anche a supportarne un’altra come conseguenza gerarchia? soldi) a quelle disposte inizialmente diretta (alternativa) Unanimità (SI/NO) boolean Test_coerenti 0-5 Affordable_loss_focus L’imprenditore ha focalizzato la sua attenzione 0-5 Prendono le decisioni insieme e solo se tutti Il test è coerente con le ipotesi (permette di a non perdere più di quanto può permettersi d'accordo? testare le ipotesi) invece di focalizzarsi sul valore atteso Maggioranza (SI/NO) boolean Test_validi Specificità: il test è fatto nel vero contesto in 0-5 Crazy_quilt_competitor 0-5 cui opera la startup Prendono le decisioni in base a quello che Validità: Il utilizza metriche coerenti con il Se l’imprenditore ha stretto partnerships o pensa la maggioranza? costrutto teorico alleanze con possibili competitor Affidabilità: il test utilizza misure ripetibili con un basso errore di misurazione Fase_startup 1-5 Test_rappresentativi 0-5 Crazy_quilt_supply Se l’imprenditore ha ridotto l’incertezza 0-5 In che fase si trova la startup (vedi lista nello Il test coinvolge un campione con le stringendo accordi con fornitori che hanno script) caratteristiche del reale target della startup mostrato interesse prima della commercializzazione FEEDBACK boolean Test_rigorosi Usano il test giusto e con le procedure giuste 0-5 Crazy_quilt_client Se l’imprenditore ha ridotto l’incertezza 0-5 Se hanno ricevuto feedback da esperti, (es. domande aperte nelle interviste; o hanno stringendo accordi con clienti che hanno mentori, altri imprenditori, amici, familiari, ecc. una baseline di confronto o un contraffattuale mostrato interesse prima della oppure no nel test) commercializzazione Feedback_expert Feedback_expert 1 - hanno ricevuto feedback 1-0 Test_causalità 0-5 Lemonade_surprise 0-5 Il test misura un nesso di causalità tra le 2 Misura in cui hanno cercato di sfruttare eventi da esperti (ad es. Mentore, imprenditore variabili testate (se Variabile1 allora effetto su inattesi (nuove informazioni, nuovi incontri, esperto, altro esperto nel loro campo, ecc.), Variabile2) sorprese) altrimenti 0 Feedback_negative Feedback_negative 1 - hanno ricevuto un 1-0 Test_bias Il test è realizzato su un campione con bias 0-5 Lemonade_adapt Misura in cui adattano le loro scelte alle risorse 0-5 feedback negativo, 0 altrimenti ridotti di selezione e autoselezione a disposizione e non viceversa Feedback_change Feedback_change 1 – se cambiano qualcosa 1-0 Val_dati I dati raccolti non si basano su esperienze 0-5 Lemonade_opportunity Misura in cui hanno approfittato di nuove 0-5 sulla base del feedback riportare 1, 0 altrimenti individuali o sensazioni opportunità che sono emerse Competitor_close riportare numero, se imprenditore è incerto number Val_misure I dati raccolti misurano quello che 0-5 Lemonade_flexibility 0-5 Misura in cui considerano la flessibilità come un chiedere se più o meno di 10 o stima teoricamente l'imprenditore vuole misurare e valore da preservare approssimativa sono dati affidabili Competitor_broad riportare numero, se imprenditore è incerto number Val_sistematic 0-5 Pilot_plane_control Il focus è su quelle attività che l’imprenditore 0-5 c'è un modello di metriche, uno schema, chiedere se più o meno di 10 o stima conosce bene e può controllare, invece di qualcosa che categorizzi la raccolta dati approssimativa affidarsi a previsioni Teoria_chiara 0-5 Val_esplicativi riesce a connettere i vari risultati e a 0-5 Pilot_plane_exec Il focus è sull’execution invece che aspettare di 0-5 La teoria è comprensibile (falsificabilità) rielaborare la propria teoria conseguentemente vedere cosa succede Teoria_elaborata 0-5 Val_stima Se gli imprenditori hanno una misura di 0-5 Contingency_plan 0-10 Quanto è dettagliato il loro pensiero su cosa performance in base alla quale stimano il La teoria va nel dettaglio (falsificabilità) fare in questo caso (se una grande azienda valore dell’idea al fine di prendere la decisione dovesse entrare nel loro mercato) finale (Continua/Pivot/Exit) Teoria_alternative 0-5 Val_componente Evidenze dei test (relativi a specifiche ipotesi) 0-5 La teoria considera aspetti alternativi sono tradotte in una stima del valore della (generalizzabilità) componente del modello di business testata

Figure 2: first datapoint by telephone contact - script

51

3. Sample analysis and different but homogeneous groups in terms of numerousness, as they may be preliminary comparable this way. Obviously, due to the IVL nature and scope of the project observations (i.e., young enterprises as well as highly innovative backgrounds and ideas), the This thesis work is based on data gathered average seniority measured in the sample at the initial step of the program, so that was minimum, so that the three clusters the reference sample includes 542 records: used relatively low thresholds: 24 years it grouped the team leaders of each old separates the first set, while 30 years startup, as well as all the other team old is the higher threshold. The most of members who accepted to answer participants are very young, declaring to interviews and surveys. Then, first the be 24 or less; the set of the oldest ones (i.e. following analysis will include the starting from 31 years) counts 173 descriptive statistics according to what individuals, so becoming the less emerged from such pool of respondents. populous of the three. Again, if considered the demographical features, it is important to analyze the participants’ origin. As observable in Figure 4, the experiment design and pandemic scope allowed to attract entrepreneurs from all the national territory, registering the majority of subscriptions within the regions of in order Lombardia and Piemonte. Anyway, there is strong presence of regions belonging to south and centre of Italy: 90 of 542 individuals have domicile in the south of Italy, against 96 from the centre and 306 from the north (50 are blank values, namely who did not declare his own domicile address).

Figure 3: gender and age distribution

In Figure 3 it is possible to note that there is heterogeneous distribution of the gender, seen that male individuals are almost four times greater than females. Instead, looking at the age declared by every entrepreneur, is has been necessary to proceed with clustering by age range, with the purpose of generating three Figure 4: geographical distribution 52

The clustering process executed on the age grouped between 19 and 35 years, while was needed since defined as preparatory the second and third more frequent values to the successive computation of diversity recorded into this sample are respectively indicators (e.g. Blau’s index). Indeed, the 27 (41 times out of 542) and 28 (37 times heterogeneity measure calls for division in out of 542). A strong polarization around categorical sub-groups, being carefully the twenty-four value can be due to the selected according to keep internal fact that it represents the average age at homogeneity within the group in order to which students usually achieve an better detect external heterogeneities. academic title equivalent to master degree However, a more precise overview on the or, more in general, at which the age distribution is guaranteed by the university students conclude their own probability distribution (expressed in academic path. The criterion adopted to terms of relative frequency) in Figure 5, choose the second threshold is that such where the ratio of each seniority level is limit (thirty years old) exactly located on computed as well as displayed in the 68th percentile, that sounds like percentual terms, referred to the totality of something similar to the established 542 participants. It appears evident how partition utilized when a density the modal value corresponds to the 24 distribution is decomposed in three parts, years old level, that is the reason why it according to three-sigma criterion applied had been chosen as first threshold of the on percentile figures - which usually cuts three different ranges. The right tail is the probability area subtended under the longer and confirms presence of a reduced curve, in order to put the limits in pool of older entrepreneurs, a natural correspondence of precise values which phenomenon given the nature and the can differ according to the standard marketing channels of InnoVentureLab. deviation and to the adopted convention. Almost the total amount of individuals is

Figure 5: age density distribution

53

Furthermore, an analysis on the majors What emerged is that an unexpected faced during the academic courses has number of participants was engaged (or been launched; so it comes to the had been engaged) in several types of outcomes deriving from the question courses external to InnoVentureLab, about the kind of subject studied during always talking about entrepreneurial university classes while involved into the arguments and economic themes. In startup’s entrepreneurial activities. Such particular, referring to the whole pool of kind of record resulted in 227 total 542 individuals, just few people stated to responses, nearly amounting to the not have interests involved in such kind of totality of who answered to be student at classes. Looking at external courses the previous question, that is as much as treating about entrepreneurial methods 234. and tools, seven people are identified as blanks (i.e. they did not answer the

corresponding question), while as many as 205 entrepreneurs (about 38% of the totality) already took (or were taking) entrepreneurial education (e.g. business model canvas, minimum viable product usage, customer analysis, feedback exchange tools, and so on).

Instead, watching at more traditional courses - that means referring to lessons

on economic insights and managerial Figure 6: currently attended course concepts - the number was even higher. The results are shown in Figure 7, where the two diverse categories are proposed: The results were clustered into three the two sub-sets are separately splitted, macro categories and are showed in and the belonging to one sub-set is Figure 6. There is a certain degree of represented as boolean value. The value 1 polarization towards STEM field, which stands for event manifestation, otherwise counts for 51% of the whole pool. After, zero value indicates absence of expression economic studies emerge, weighting for in that dimension. Reading the chart, as 37% about. Finally we find the residual many as 346 people (64% of the whole) category, including fashion industry, affirmed to have received education about agrarian techniques, art, media and management means and economic communication, law, political science and knowledge (e.g., macro-economic theory, literature. Yet, it was asked whether the micro-economic theory, corporate finance, entrepreneurs had taken some courses game theory, general financial education, treating of specific topics, such as stocks market, and so forth). entrepreneurial education as well as economics and management.

54

The density distribution clearly identifies the value 10 as modal one, anyway the overall pattern is less polarized than that observed for other dimensions (e.g. seniority distribution). The second mode is in correspondence of twenty hours per week (per member), equivalent to almost 3 hours per day. The overview allowed to observe heterogenous distribution of the Figure 7: distribution among external courses 542 entrepreneurs along such variable, and this is due to other potential

commitments (e.g., further work activities, At every step of the program progress and university, sons/daughters, chance of for each startup, the research assistants are separating entrepreneurial tasks among requested to ask entrepreneurs for another team members, pandemic emergency, interesting dimension: the amount of lifestyle, and so on). However, majority of hours dedicated to the startup activities, people is concentrated on left side of the during just a week and referred to every distribution. Indeed, the 50th percentile is single team member. The outcome is quickly reached: it is individuated just displayed in Figure 8, where a probability before the modal value, namely where is distribution is plotted. Please note that, the value 8, that corresponds to the due to scale issues, the second dimension commitment of a few more than 1 hour per (i.e. percentile value at each abscissa level) day for each team member. is measured on the second axis, located on the right in the graph.

Figure 8: weekly commitment per member

55

Figure 9: number of companies established before pursuing the startup venture (left)

Figure 10: boolean variables about prior entrepreneurial experiences (right)

Figure 11: years of work experience declared by each team member

Figure 12: years of work experience into the same industry of startup declared by each team member

Figure 13: number of industries in which experience is cultivated

Figure 14: working experience years in role of corporate executive

56

When focusing on previous experiences the class assignment, since the research declared by participants, it is possible to team considered it as discriminant and view Figure 9 and 10: in the second one, potential control variable. In Figure 15 the two boolean dimensions are displayed in presence of each progress phase is order to indicate how many people displayed: for the most part (201) teams already experienced Business Plan writing stayed into the first phase at zero time, that (276 of 542, about 51%) and venture signalled belonging to the analysis step establishing (95 of 542, about 18%). In (i.e. survey and interview to customers addition, in a set of charts ranging from were still in progress, website yet did not Figure 11 up to Figure 14, the density exist as well as landing page or prototype). distributions expressed in absolute values are shown about four different variables. For graphic cleanliness and clarity about numbers, the blank values are not drawn since they are numerous especially into the Figures 12-14, in which they assume hundred order of magnitude, making difficult to read the values reached by the variables of interest. In all of the four scenarios, the modal value corresponds to a very low number (two times is 0, two times is 1), with an overall tendency of the curves toward the left side. Please note Figure 15: progress phase that the third variable (number of industries in which experience has been done) is an artificial variable, created by The other phases stand for: 2) a product counting for each individual the number basic version existed, 3) the prototype had of sectors indicated into the answers being tested toward the client, 4) the rather than looking at the nature of such offering worked while incomes had not sectors, so that the aim is measuring yet been generated, 5) the revenues had heterogeneity in prior experience, and started to arrive. Instead in Figure 16 are flexibility degree achieved thanks to the reported the other - to be meant as in previous work activities. It is surprising addition to the commitment dedicated on the amount of years in work experience startup workload - commitments declared with role of executive: if compared only to by each of the 542 participants at the who gave answer (135 of 542 people), the program starting: it is curious to note that percentage of those having at least 5 years only the lowest percentage is represented of experience as corporate executive is by people with full-time interest on near to 30%. startup activity, while almost the half of During the formation of sample and sub- respondents worked while involved into groups of treatment, control and pure the entrepreneurial tasks. control, the progress status achieved by the team was taken into consideration for

57

measured the maximum academic level reached throughout the entire lifespan, so including both the highest study title obtained before starting the program and the current attended course.

Table 1: maximum academic level

Academic_lvl_corrected B.Sc. 105 Figure 16: current other commitment M.Sc. 105 Master 88 nothing 84 Looking at the studies attended while PhD 21 working for the startup, as many as 233 studying B.Sc. 127 individuals answered to follow academic studying M.Sc. 6 courses during the examined period; in studying Master/MBA 5 particular, the range in which fluctuate the studying other post-lauream course 1 several types of university levels is wide 542 Academic_lvl_corrected (years codification) and composed as showed in Figure 17. It Average Max Min SD VAR appears very popular the choice of 1,8672 4,0000 0,0000 1,1715 1,3723 beginning an entrepreneurial venture in the same period corresponding to bachelor studies, maybe due to incentives Into the second part of Table 1, it is deriving from the new university viewable the split which reports the experience. The ones studying in a numerical main descriptive statistics of doctorate course counted only for five such variable, which is treated as items, that is just 2% out of the total. codification from categorical items to numbers, with the aim of computing finite quantitative values as those shown above. In addition, in Table 2 a statistical description is proposed again: this time, is referred to the psychological variables which in phase of preliminary experiment design were considered useful to be analyzed, in view of potential correlations with decision-making approach. So, they have been quantified and here are Figure 17: academic level presented the typical statistical properties: mean value, maximum and minimum

limits (so that it is possible to extract even Successively, a more inclusive indicator the range), standard deviation and then was created in order to get a complete variance. framework about the academic levels: the new variable (please view it in Table 1) 58

REP boolean: Yes & No Variable Average Max Min SD VAR

become rich 3,9021 7,0000 1,0000 1,6591 2,7527

move up into the business world 4,1563 7,0000 1,0000 1,7804 3,1699 I will found a 5,6008 7,0000 1,0000 1,5500 2,4026 firm in order solve a specific problem faced by people with which I strongly identify to have a proactive role in shaping the activities of people with which I strongly identify 5,3710 7,0000 1,0000 1,5781 2,4904 solve a social problem which private firms are not usually able to face 5,4840 7,0000 1,0000 1,6737 2,8012

have a proactive role in changing the way by which the world acts 6,0038 7,0000 1,0000 1,3161 1,7321

manage my firm basing on robust management practices 5,9190 7,0000 2,0000 1,0916 1,1915 In role of deeply analyze financial prospective of my business 5,9473 7,0000 1,0000 1,0820 1,1708 founder, it will 5,8343 7,0000 1,0000 1,4112 1,9914 be very supply offering useful to people with which I strongly identify important for show my customers that I agree with their opinions, interests and values 5,8079 7,0000 1,0000 1,3034 1,6989 me to be highly conscientious world citizen 6,0188 7,0000 1,0000 1,2330 1,5204 make the world a better place 6,1601 7,0000 1,0000 1,1859 1,4064

focus on what my firm can obtain if compared with competitors 5,5292 7,0000 1,0000 1,2858 1,6534 When I will establish a strong competitive advantage on competitors 5,7740 7,0000 1,0000 1,2641 1,5979 handle my 5,1224 7,0000 1,0000 1,5859 2,5152 corporate, it focus on people with which I strongly identify will be very support people with which I strongly identify 5,1186 7,0000 1,0000 1,5945 2,5425 important to focus on what my firm can do for social welfare 6,0094 7,0000 1,0000 1,1598 1,3452 persuade others that private firms are able to face social problems like those challenged by my firm 5,4068 7,0000 1,0000 1,5674 2,4569

SELF ESTEEM 4,4934 7,0000 1,0000 1,2121 1,4693

RISK APPETITE 5,5521 7,0000 2,2500 0,9178 0,8423

NOVELTY 4,8793 7,0000 1,3333 1,1582 1,3415

PLANNER 5,9165 7,0000 1,0000 0,9484 0,8995

LEARNING GOAL ORIENTATION 6,1308 7,0000 1,0000 0,7875 0,6201

PERFORMANCE AVOID ORIENTATION 4,3990 7,0000 1,0000 1,3924 1,9388

dimensions psychological PERFORMANCE GOAL ORIENTATION 2,5637 7,0000 1,0000 1,2784 1,6344

TEAM BUILDING 5,6059 8,0000 1,0000 1,9432 3,7759

Table 2: descriptive statistics

59

3.1 Team level typology has been assigned, according to what declared by the team representative. aggregation and database creation The variables illustrated within the above chapter have been deeply analysed in order to generate a new set of dimensions at team level, starting from the initial dataset. First of all, the team leaders (i.e., representative, spokesperson) were used as reference fellow in considering some categorical variables, since they offer a good idea of which are the aptitudes and Figure 19: kind of offering practices characterizing the whole startup.

In the pie chart above, it appears how the startups delivering services are the most numerous, amounting to around 64% of the totality. The belonging to one or another of such categories is useful as control variable, since we expect to differentiate the decision-making processes according to what is the main offering proposed by entrepreneurial team. Meanwhile, a core dimension Figure 18: team leaders’ origin employed for the regression analysis is the artificial variable named as internal network, since it measures the extent to The geographical provenience of each which cohesion and strong network team representative is tracked in Figure among team members are detected. 18, revealing that individuals mostly come from north of Italy, while the ones from south and centre are nearly in same proportion. This enough heterogeneous geographical composition has been made possible thanks to the fact that the entire program is delivered through online channels (due to pandemic situation), so avoiding necessary physical presence and travel costs, namely allowing subscriptions from all around the national Figure 20: internal network territory. Yet, to each startup an offering

60

Such dimension was shaped by looking at resulting level is 2, so that whether all the three different boolean variables - that three values display 1 then the level is 3. were: 1) did you attend, even if in different The consequent variable intensity is periods, the same university of at least one of plotted in Figure 20 by using diverse the other team members?; 2) did you work, shades of blue, and increasing colour even if in different periods, in the same intensity as level magnitude arises: the corporate where at least one of the other team overall intensity along such dimension is members worked?; 3) before working on this low, indeed 46% of the total teams business idea, did you have work or study registered minimum level, whereas only collaborations with another team member? – the 6% falls into the maximum one. Figure with the aim of obtaining a quantitative 21 permits to have a complete overview on indicator of ties preexisting among the distribution of average age of the members (i.e. previously to the startup teams, expressed in absolute terms and foundation). These questions have been not in percentages, where the mean value considered only when asked to the team is computed by applying arithmetic representative, in order to eliminate average on everyone belonging to team. redundance and double bonds or The curve is not particularly switched worthless complications due to matching toward the left side, signalling quite activities. When there is absence of all the heterogeneity along such parameter. At three sub-dimensions, the resulting half of the area subtended under the curve internal network level is zero. (50th percentile), it comes to a few less than Consequently, if only one boolean is equal 28 years old; that is an expected result, to one, it means the event manifestation is given that the most of participants are anyway observed, so that the final level young and declared less than thirty years, becomes 1. After, if two boolean items so that outcome appears to be consistent return positive outcomes then the with the previous statistics.

Figure 21: age distribution among teams 61

Figure 22: average weekly hours of commitment per member

Proceeding with aggregation of single Right about team numerosity, in Figure 23 data at team level, Figure 22 returns the is displayed the numerousness recorded outcomes referring to the commitment by every team. Please note that such invested on startup activities: the range of variable derived from what stated by hours comes from null value until 110 representative entrepreneur during hours a week, equivalent to about 16 hours compilation phase, therefore not taking a day per member. It is important to into consideration the actual answers underline that the total commitment received. The majority of the sample obviously depends on team numerosity, declared to be involved in unique-person other than the single amount of hours per teams (around 45%, 137 of 305 startups), person. The modal value is around 20 after there are dual teams (~ 25%) and so hours per week per member, forth, decreasing in frequency by corresponding to almost three hours per increasing in numerousness as it was day for each team member. expected: just few ventures consist of more than three people, due to difficulty

of finding compatible individuals with which sharing key decisions and strategies. The highest gap, in terms of frequency, is at the turn of 5-members and 6-members groups: the numerosity equal to 5 counts 14 startups, while numerosity equal to 6 fellows accounts for just two persons, equivalent to less than 1% compared to the totality of 305 enrolled

teams. Figure 23: team numerousness 62

Figure 24: current subject studied by majority

Figure 25: current subject studied by team leader

When it comes to the different majors mean fluctuates between zero as inferior studied during the classes attended while limit and four as upper limit. It is curious following the IVL program, it becomes to observe how the modes all locate in fundamental the creation of three macro- correspondence of finite values (i.e. no categories in which clustering all the decimals), that are 0, 1, 2 and 3. Such possible facets: STEM (science, phenomenon happens because of the technology, engineering and numerosity of teams: they are mostly mathematics), economics and other. composed of 1 entrepreneur, thus the During the aggregation at team level, two consequent average outcome belongs to different dimensions were separately the natural numbers’ set. Keeping in mind analyzed for each startup team: the the codification (PhD=4; B.Sc.=2; belonging of the majority within the team InProgress=1; other=0), the level 3 stands (please see Figure 24) and that of the team for master’s degree and similar: right the leader (Figure 25). What emerged is that in value 3 is the most frequent, sounding like both cases the most of participants were the average overall maximum academic STEM students, with a slight advantage level is equivalent to master. over economics students; the gap is slightly more pronounced at majority standpoint. The missing values represent a prominent slice: they account for 174 records in majority counting, while 188 records in representative counting. Remaining into the academic field, but moving towards the level instead of background type, we can look at Figure 26: the density distribution of the average maximum academic level (codification in years) is represented. The mean is computed starting from values ranging Figure 26: average maximum academic level achieved between 0 and 4, so that even the resulting 63

Figure427*: average years of work experience at team level

Figure 28*: average years of work experience within the same industry of startup at team level

Figure 29*: average number of industries in which experience has been cultivated at team level

Figure 30*: average years of experience as corporate executive at team level

* Value 0 has been not plotted since it gets huge numerosity and therefore causes scale problems in graph visualization, not allowing to well view and appreciate the other lower values 64

Figure 31: already established other companies before entering the startup team

Figure 32: previous experience in business plan writing

Figure 33: attended other economics and management courses

Figure 34: attended other entrepreneurial courses

Figure 35: dimension in which the startup differentiates (team leader’s perception)

65

Yet, looking at the variables referred to without losing information on the treated prior experiences deeply analyzed dimension. In each graph, the count of throughout the above chapter, a further each value indicated on abscissa axis must step is done by extending the analysis to return 305 as final sum, and that is verified team aggregate level: in Figures 27-30 are in all of the four scenarios. It is relevant presented the probability distributions, noticing that, for all the 4 variables, the expressed in terms of absolute values, of two most frequent values are exactly at the four variables already explained in this antipodes of density distribution: this is thesis work. Please note that the trend true for who already established other defined in i) years of work experience, ii) companies (222 startups registered 0% of experience in the same sector to which members vs 44 registered 100% of belongs the startup enrolled in IVL and iii) members), who already did business years of prior experience in charge of planning (113 vs 127 startups), who took executive role highlighted low average external economics and management overall values, since the entire distribution courses (85 vs 164 enterprises) and who switches towards the left side and the attended external entrepreneurship modal value falls always within the left courses (155 vs 83 different ventures). area. The above consideration is ever Instead looking at Figure 35: verified except for a variable: Figure 29 entrepreneurs (representatives are shows that the mode is again on a reduced considered as reference because of their value (1 sector in which work experience more reliable judgement) were asked of has been done), however all the residual which dimensions they believe were the frequencies dispersed through the whole most important within the competitive right side of the plotted area, signalling a environment in which they played. The great number of teams (startups) with several alternatives are 5, and it is highly high flexibility level - because it was interesting the resulting outcome: the measured as the average number of differentiating dimension chosen by different industries in which members participants more frequently is quality, in experienced work tasks (that is, a factor opposition to the less selected, that is underlying to agile mindset and flexibility price/cost focus. This result reminds of when applied to decision-making the economic theory, specifically to the process). discerning between the two more utilized In Figures 31-34, we can see the original and basic approaches in strategic boolean values (even in this case everyone competition decision-making: has been already explained in previous differentiation strategy (based on quality, chapters) which, at startup level, have or better to say “perceived quality”) been transformed in percentage of event versus efficiency setting (made possible manifestations (% of 1 boolean value) thanks to cost destruction and compared to the team total numerousness, consequently price competition). so that a quantitative tool can be kept

66

Figure 36: industries to which belongs each startup (N=305)

In conclusion, view in Figure 36: here are showed several industries ordered by 4. Results analysis increasing frequency, all including at least Finally it comes to conclusive outcomes. two teams except for the hardware sector. First, please note that as many as nine These are the main types suggested into different regression model have been the survey, with the aim of clustering elaborated, each one splitted for every possible combination. Obviously, scientificity and effectuation as dependent the most populous category is the residual variable and, in the end, each one has been one, which groups all the remanent duplicated for testing in the robustness industries not considered by the pre- check models. In Appendix B please view setted options: e.g. real estate, healthcare, an extract from the correlation matrix used environment, marketing, automotive, as orienting map: here only the correlation manufacturing, travel and tourism, beauty with the dependent variables are care, sport, design, photography, displayed - given the elevated sustainability and social network. The numerousness of variables (121x121 sum of the frequencies with which every matrix) - and a conditional formatting is industry appears is equal to 305, so that applied with the aim of highlighting the the entire sample is present, avoiding lack diverse intensities of correlation. Such of information in blanks. Please note that matrix has been initially used in order to food, software, fashion and health are the understand how moving in the large most prominent areas in weight, amount of dimensions, by quickly registering large gap if compared to the individuating the best relationships following ones. between independent variables, as well as between independent and dependent ones. Please note that the variables of interest (i.e. natural levels of scientificity and effectuative approach) have been also

67 considered in a weighted form, with the statistical results expressed into the weights reported in Table 3. First of all, for regression models do not differ if the each sub-category underlying to the two weights are included. This is the reason main variables the standard deviation has why the weighted dependent variables been computed among all the record will not be analyzed in such thesis project, (represented by the 305 startups), as of the seen that the evidences reached with the reciprocal is immediately available: the not-weighted variables of interest keep the total sum of reciprocals is needed as same within the weighted case. Indeed in normalizer, so that every single weight is Appendix B the equivalence is easily ready (and the sum of weight is 100% as readable: the correlation values are verification). The more the anti-variance is around the same number both for high, the more the sub-category gets weighted and not-weighted scientificity relevance in the final weighted mean. Such level, as well as for weighted and not- computation has been done since it could weighted effectuation level. be useful to isolate outliers and distortions A further verification is offered by the due to biased judgements from different diverse shades of colours adopted to research assistants – so when for a certain immediately distinguish the magnitude of sub-category the judgement is correlation: the various types of colours homogenous among all the variables then are always coupled, namely every couple a more reliable indicator is expected, seen of dependent variables – that it means there was no outliers by normal/weighted scientific approach and neither RA side nor startup side. normal/weighted effectuative approach – As opposite, should the standard always shows the same shade of colour (or deviation being too high for a certain sub- non-colour if the correlation is not category among the participants, then it sufficiently high) by getting a similar would signal that a set of startups degree of correlation intensity. Finally a registered abnormal values, with comparison between anti-variance values: ambiguity whether deriving from RA’s only a very slight difference exists error or venture performance. Therefore, between scientific (3,8531) and in the weight row from Table 3 are effectuative (3,9798) approach, thus we recorded the values by which the averages can conclude that the variability (standard are multiplicated in the final weighted deviation) observed for the two macro- mean. However please notice there are no categories is comparable, in other words significant differences if compared with no distortions from outliers is expected. the not-weighted values, indeed the

SD 1,077099551 1,189295822 1,597421542 1,350189426 1,394240433 1,112815896 1,540012152 1,16437591 1,504228115 1,101005991 WEIGHT 24,095% 21,822% 16,247% 19,222% 18,614% 22,579% 16,316% 21,579% 16,704% 22,821% 1/SD 0,92841929 0,840833695 0,626008836 0,740636818 0,717236408 0,89862124 0,649345526 0,858829173 0,664792786 0,908260271 1/SD sum 3,853135046 WEIGHT sum 100,000% 1/SD sum 3,979848996 WEIGHT sum 100,000% var. THEORY HYPOTHESIS TEST VALIDATION DECISION THRESHOLD BIRD IN HAND AFFORDABLE LOSS CRAZY QUILT LEMONADE PILOT PLANE

Table 3: weighted scientific and effectuative natural levels

68

4.1 Preliminary insights. Thus, obviously the natural levels of scientific and effectuative evaluations orientation were expected to be similar The regressions have been chosen and before the program beginning. In launched by using the numbers emerged addition, we must consider that a sort of in the correlation matrix (Appendix B) as learning economies exist among the reference, with the objective of research assistants, since the judgements maximizing the correlations with the they give may become more and more dependent variables while avoiding accurate and not biased by better knowing multicollinearity among independent the participants, as well as by doing variables, which would generate interviews and assimilating the theoretical ambiguity in interpreting outcomes and concepts. distinguishing effects, other than penalizing the coefficients’ statistical relevance and the adjustment goodness measures. In Appendix C are shown – for av_scient each model and for each variable – the significance and the verse of all the relevant statistical relationships. Please 4 note that every independent variable av_effectuat keeps a steady verse throughout all the 2 regression models, so that consistence 0 among them appears to be confirmed. 0 2 4 Firstly, deserves to be mentioned the Figure 37: scientificity-effectuation relationship well rounded extremely high correlation between the by linear function two dependent variables. Indeed, as displayed in Figure 37, it is possible to Instead, looking at the split of the observe how the variables of interest dependent variables, as described into the follow almost a linear function as kind of previous chapters we can discern 10 sub- dependence, so underlined the high level variables divided in 5 for scientificity and of collinearity which can be explained 5 for effectuation. It is interesting to with the following statement: the analysis highlight the dependences emerging treated in this thesis project only considers between the components which composed the initial phases of the IVL experiment. the main variables of interest. In Table 4 In other words have been included the levels of correlation measured through exclusively the data available at time zero the Pearson coefficients are presented, and before any type of lesson and technical then confirmed by the graphs illustrated suggestion by the instructors. This is the in Appendix D: each kind of colour reason why the levels of scientificity and indicates a diverse range of intensity, in effectuation can be considered as natural, order to quickly individuate the higher seen that they should be innate in the and lower correlations between all the individual aptitude and not influenced by sub-variables. program’s contents and interaction 69

av_scientth av_scienthp av_scienttest av_scientvalid av_scientdecthresh av_effectuatbird av_effectuatloss av_effectuatquilt av_effectuatlemon av_effectuatpilot

av_scientth 1 av_scienthp 0,704284422 1 av_scienttest 0,506565443 0,657486364 1 av_scientvalid 0,491528959 0,652540558 0,855542662 1 av_scientdecthresh 0,413211211 0,534501949 0,647792536 0,766889971 1 av_effectuatbird 0,457554298 0,389177635 0,344188954 0,307036275 0,221606057 1 av_effectuatloss 0,306015639 0,23003089 0,249409684 0,244854663 0,187059711 0,258145972 1 av_effectuatquilt 0,434946493 0,406066701 0,423856642 0,43514066 0,454217005 0,319859374 0,245655035 1 av_effectuatlemon 0,36855569 0,346920236 0,326551518 0,376760538 0,299710665 0,27209254 0,27391814 0,366587314 1 av_effectuatpilot 0,447303146 0,364339006 0,219727987 0,271142283 0,263063843 0,266956585 0,284346441 0,317680826 0,256306932 1 Table 4: Pearson coefficients computed for all the sub-variables of interest

The most elevated correlations (bordeaux equal to 0.1871 – here we need to zone) show in the scientific approach remember that the scientific threshold paradigm: the validation (av_scientvalid) measures the extent to which people took and test (av_scienttest) components decision by using solid criteria, and it registered the highest value which further assesses the criterion’s reliability amounts to 0.8555, whereas the second and consistency. Such point perhaps could one correlates appropriate threshold explain the low correlation with the loss during the major decisions variable: the loss aversion/appetite (av_scientdecthresh) with the validation measures also whether the individual is component and is equal to 0.7669 – the oriented or not towards the risk, rather result is in line with what expected since than the appropriate criterion chosen for the validation phase usually comes after taking orientation. So that an entrepreneur the test step but before the core decisional might be weak in initial economic moment, so that a sequential connection investment and when evaluating how exists along with the conceptual much is willing to loose, meanwhile being correlation between the different phases of largely scientific and able to select the best a same cognitive process. Yet, more in criteria during the core decisional general all the sub-variables underlying to moments. After, the second lower value in the scientificity world appear to be highly Table 4 is between pilot effectuative correlated among them, and such aptitude (av_effectuatpilot) and scientific evidence is no longer true if we watch the test phase, with a coefficient of 0.2197: this effectuative approach paradigm. There, can be justified by the fact that the pilot the Pearson coefficients are deeply lower plane feature is associated with who and it is surprising that such sub-variables prefers to execute (rather than recorded more correlation with the overthinking) and control what is well scientificity components than among known by not waiting for others’ themselves. predictions or possible future events. Such characteristic seems not to be related in The lowest value in Table 4 is in any way to the scientific capacity of testing correspondence of scientific decisional hypotheses, neither with positive verse threshold and loss orientation nor with negative relationship. (av_effectuatloss) by registering an intensity

70

4.2 Regression particular, belonging to regions of the south is negatively correlated with the presentation effectuative approach, even if the In this paragraph the nine different significance is enough low (p<0.1). The regression models will be presented and numerousness of the startups is discussed. Firstly, the model 5.b displayed statistically relevant too, given that in Figures 38 and 39 represents the best fit presents positive and strong (p<0.01) among all those that have been launched: coefficient into the relationship with the the adjustment goodness measures are the effectuative approach. After, also the highest ones, indeed the R-squared commitment invested on the startup work parameter is 0.4296 and 0.5433 is a core factor in the model 5.b: for each respectively when the dependent variable startup the average number of weekly is natural level of scientificity and hours employed by each member is effectuation; in other words, this is the positively and highly correlated with both model getting the highest degree of scientific and effectuative approach’s variance explained by the considered natural levels. In addition, it is relevant variables. Furthermore the Root MSE that the boolean utilized for the phase 2 of (Mean Squared Error), that is a value to the startup progress registered highly keep the lowest as possible since indicates significant coefficient: within the model the residual error dispersion (where error 5.b it is positively correlated with both is difference between actual and predicted scientificity and effectuation, and always value), is minimized by assuming value reporting the greatest level of significance respectively equal to about 0.939 and 0.636 (p<0.01). Please note that the omitted - thus we can conclude that 5.b is the boolean is the one referring to the startup model able to explain the greatest quantity phase number 1, so that the results of variance while recording the minimum highlighting positive correlation for all the error observed in this analysis. following steps are in line with what expected: indeed the 2-5 steps always Looking at the independent variables, the show positive correlation with the natural industry to which the startups belong level of effectuative approach, while this is appears as statistically significant: taking not always true for the natural level of into consideration the poor numerousness scientific method (i.e. in phases 3 and 4). of the most of industries, it is appropriate This outcome would be explained by the to focus on the more populated industries. fact that not all the entrepreneurs in For instance the software sector, having advanced development phase are there numerousness of 26 startups, shows thanks to particular entrepreneurial negative statistical correlation with the education or structured insights, there it effectuation natural level in the model 5.b looks like that being a good entrepreneur and at the maximum level of significance is not always aligned with acting like a (p<0.01), whereas no relationship scientist, but seems being always emerged with the scientificity natural associated with better abilities of degree. Yet, the geographical origin affects execution and control on which bases the the dependent variables in such model; in effectuative paradigm.

71

Figure 38: model 5.b – scientificity

72

Figure 39: model 5.b – effectuation

73

Another interesting point emerging from PGO effect. Finally it comes to the model 5.b is about the control variable of heterogeneity indicators: in 5.b model offering kind: taking into consideration only the Blau’s index referred to the that the omitted boolean is the offering student quality (bu_curstude) presents like product, the startups selling services statistical relevance – but however (rather than products) resulted to be minimal (p<0.1) – since it is positively extremely more scientific (greatest level of related to the effectuative approach but significance), but no relationship is found not to the scientific one. Therefore it seems with the effectuative approach. that the diversity in terms of students’ presence within the team positively affects The artificial variable created to measure the group abilities of execution and the number of diverse industries in which control. The right mix of current students the founders have had experience is and more experienced people lead to statistically relevant too. Furthermore, merge practical insights and technical also the experience deriving from education, so as a more pragmatic and previous tasks involved in venture empirical approach can be supported by establishment (i.e. boolean indicating solid academic insight and deep whether the individual has already knowledge about the established other firms) results to be business/technology topic. By the way, largely (p<0.01) and positively (verse +) please consider that such result is not correlated with the innate effectuative completely reliable since the Blau’s index approach (but not with the scientific one), here recorded just a few values not equal signalling that more pragmatic education to zero (36 out of 305, around 12%). and insight can be effectively useful in Finally, the boolean indicating whether improving the abilities related to control the team majority belongs to a B.Sc. class and execution. An interesting side effect is is in alert area: indeed bsc_maj is that another experience proxy, namely the negatively correlated with effectuation at number of firms previously established by an almost relevant significance degree each participant, is strongly (p<0.01) but (0.190%) blank values. recognized and appreciated approach The av_pao dimension, that describes the diffused in a number of prestigious Performance Avoid Orientation, is academic structures. It is something negatively correlated into the model 3.b; deriving from experience and execution, so that the aggregated variable trial-and-error processes as well as innate av_perforient, computed as simple sum of aptitudes, thus belonging to the lowest PAO and Performance Goal Orientation, academic level (i.e. Bachelor degree) presents again a negative correlation generally signals for low seniority and coefficient with the effectuation method, poor empirical experience, then even if the significance degree is low suggesting for inferior capabilities of (p<0.1) because probably mitigated by the execution, network, flexibility and control. 74

Figure 40: model 1.a – scientificity

Figure 41: model 1.a – effectuation

75

Now please look at Figures 40 and 41, experience suggests for higher seniority where model 1.a is displayed: despite the and empirical heritage and so a better low adjustment goodness measures (R- orientation towards effectuative squared <0.2) due to the small paradigm, which is indeed based more on specification size, this is the only model features deriving from empirical insight including the atleast1_stemeco. This rather than academic education. Finally variable results to be relevant as recording the artificial variable named Internal a positive correlation with the effectuative Network (intern_netw), that measures the approach (p<0.1), whereas such positive extent to which the startup team had influence suffers mitigation if related to previous ties in terms of work or scientificity (0.1

77 so a negative psychological aptitude appears to negatively affects the decisional processes, and such evidence is absolutely in line with what expected ex- ante. The last point touches the heterogeneity dimension: the seniority Blau’s index (bu_age) is positively related with both scientificity and effectuation natural levels, even if with superior intensity in the first case. Such last outcome is consistent according to the presented literature: differences in empirical experience and age bring diverse viewpoints and interpretation levels, other than the possibility of mixing more solid academic insights with previous practical experiences at work.

The exact balance makes possible to apply rigid insight to the testing phase in order to corroborate (or not) hypotheses in a scientific way, as well as to merge the skills born from experience (control, execution, flexibility, network) with the solid market knowledge and business passion coming from academic environment. Of course, in the effectuation paradigm such diversity effect is mitigated as the experience has predominant weight, in other words the absolute seniority level is more relevant than the diversity property.

78

Figure 42: model 1.b – scientificity

Figure 43: model 1.b – effectuation

79

Figure 44: model 2.a – scientificity

Figure 45: model 2.a – effectuation

80

Figure 46: model 3.b – scient.

Figure 47: model 3.b – effectuat.

81

4.3 Robustness check positively linked to high levels of scientificity: this is an expected result, A further control has been done by given that several courses focus on excluding the startups which declared a structured decisional methods and well- different number of team members (i.e. defined tools. The team building instead 111) if compared to the ones who actually measures the inclination of each responded the interviews/surveys. This individual towards the others, in terms of has been done in order to obtain a new less searching for the best means to dirty dataset with more reliable data, since communicate and collaborate, as well as at the origin some observed effects could avoiding conflicts and declaring mental be biased and distorted given the opening during the comparison with information loss: i.e., for a given record a external environment. It negatively high positive correlation has been related to scientificity, maybe deriving individuated between age heterogeneity from the poorly competitive nature of the and natural scientific level, but in fact the members, who lack of great stimulus and age heterogeneity (bu_age) bases on fake therefore are not able to focus on the data, so that real composition in seniority objective by setting structured paths and could even deeply differ if compared to strategies. By the way please consider that what tracked into the database. The its p-value is very high (more than 9%). findings coming from such change are Finally, the Blau’s indicators about gender completely in line with the original and business plan experience show models, except for two variables: negative correlation with the natural heterogeneities in gender (bu_gender) and scientific approach, looking like such in business plan writing experience differences lead worse relationship (bu_expbp). The adjustment goodness conflict management which outperforms measures (Figures 48 and 49) improve on the benefit coming from a better task overage thanks to the superior data conflict management – even though we quality: R-squared becomes 0.5737 (in have to underline that also in this case the place of 0.4296) and 0.5427 (in place of p-values are high, respectively equal to 8% 0.5433) respectively when dependent (gender) and 7.3% (business plan writing). variable is the natural level of scientificity The news arising from the model 5.b.RC and effectuation. In the model 5.b.RC it is do not keep the same within the relevant to note how a number of new specification where effectuation is dimensions become correlated with the dependent variable: this is to corroborate scientificity variable: previous experience that heterogeneity has largely more in attended entrepreneurship courses influence on the structured dynamics of (entrep_crs), team average degree of team feedback exchange which characterize the building orientation (av_teambuild), other scientific paradigm, while are not a central than the already cited heterogeneities in driver for the natural effectuation level, gender and in experience as business which pays more attention to enough planner. The entrep_crs indicates the abstract abilities such as network, control, percentage of members who already execution and flexibility, less suffering the attended entrepreneurship courses (before effect brought by task conflict and IVL) for each startup, and here seems to be relationship conflict management. 82

Figure 48: model 5.b.RC – scientificity

83

Figure 49: model 5.b.RC – effectuation

84

4.4 Final conclusions In this paragraph the resulting outcomes two regressors, and the scientific approach will be explicited in light of what studied innate intensity is highlighted with in literature. Before passing to the final different purple shades by exploiting the discussion, please note that there is a conditional formatting. narrow set of variables strongly Now we come to the research conclusions. influencing the dependent variables, and To sum up, the answer to the first research above all the av_numindustries_exp and question is that the aggregated Blau’s intern_netw showed to be statistically index never registers statistical significant in every regression model, correlation, and such is due to the general other than always record high degrees of low correlation between the sub- significance. Taking into consideration, for indicators separately considered and the instance, what happened in model 1.a, it is dependent variables. Indeed only the age possible to proceed with a sensitivity heterogeneity resulted to be positively and analysis on the natural scientificity level highly linked to the variables of interest. It by varying the values of the two just suggests for a strong contribute brought mentioned independent variables, while by merging academic insight and practical keeping the other regressors fixed to their experience in order to improve the build- domain’s superior limit (and obviously measure-learn cycles (i.e. scientificity) and the coefficients equal to what defined by the capability of focusing on execution regression model). In Figure 50 the while keeping flexibility and attention on analysis is shown, and above the table contact ties. However the bu_age has please note there is the functional form relevant statistical significance for both belonging to the considered linear scientificity and effectuation just in two regression. The average number (per models, that is 3.a and 3.b; indeed more in startup) of diverse industries in which the general it showed to be more influencing members got experience changes in its on scientific approach rather than own domain by varying the column, effectuative behavior. This is to confirm whereas the value of team internal that the experience diversity factor is network increases by going down along deeply crucial just for the scientific the first column. As expected, even though paradigm since it puts roots on the the number of regressors is quite great (10 continuous feedback exchange and test + intercept), the dependent variable comparison. deeply modifies with variation of these

av_scient 1 2 3 + xp y + + 1 xp p 5,0964 0 1 2 3 4 5 6 7 8 9 10 11 12 13 0 3,0747 3,1668 3,2589 3,351 3,4431 3,5352 3,6273 3,7194 3,8115 3,9036 3,9957 4,0878 4,1799 4,272 1 3,3495 3,4416 3,5337 3,6258 3,7179 3,81 3,9021 3,9942 4,0863 4,1784 4,2705 4,3626 4,4547 4,5468 2 3,6243 3,7164 3,8085 3,9006 3,9927 4,0848 4,1769 4,269 4,3611 4,4532 4,5453 4,6374 4,7295 4,8216 3 3,8991 3,9912 4,0833 4,1754 4,2675 4,3596 4,4517 4,5438 4,6359 4,728 4,8201 4,9122 5,0043 5,0964

Figure 50: sensitivity analysis in model 1.a – scientificity is variable of interest

85

Such is a process fostered by a good task 1998) produced on the argument, anyway conflict management, meant as the the prominent relationship conflict concept introduced by Jehn et al. (1997, management weight suggested by a 1999) in opposition to the relationship number of different authors (Jehn et al. conflict management. Right the latter 1997; Klotz et al., 2014; Fitzgerald et al., point arises too from the evidence about 2017; Zhang, 2019) is here confirmed, control variable Internal Network: it is thanks to what found on the internal statistically significant in all the regression network variable and its mitigating specifications, and with positive influence effects. in every scenario. Perhaps this is evidence Coming to the second research question, that preexisting ties among team members investigating the influence of external can positively affect the decision-making networks represent an issue given the lack process through a better relationship of such information into the database at conflict management, thus mitigating time zero. In addition, the LinkedIn divergences deriving from heterogeneity. contacts counting already proposed in A separate mention deserve the Blau recent studies should be biased in this indicators about students’ presence, experiment seen the nature of gender and business plan experience. The participants, mostly young and with few students’ presence appears as significant professional experience, so that a series of only in model 5.b and for sole effectuation, control variables was proposed at the other than with the lowest significance beginning of this thesis project. Among level (p<0.1) – therefore it does not the various control variables just cited, the represent a strong evidence, also seen that only ones which registered relevant diversity in commitment on study is correlation have been the students’ someway a proxy of diversity in seniority, presence Blau and the team so that a similar outcome would be numerousness. The first one is a expected (instead of no relationship with dimension on which also the Geremias’ the natural scientific aptitude). After, (2020) studies put a focus, but in this thesis diversities in gender and experience in project has resulted as not relevant given business plan are relevant only in models that influenced the sole effectuation by belonging to the robustness check; they registering also high p-value (p<0.1) – however affect only the scientific despite the poor correlation, in terms of approach and do it with too high levels of external network its relevance can be significance, respectively of 8% and 7.3% - explained by thinking that the presence of such is a very weak evidence, and in students and experienced people at the addition 1) no other relevant outcomes same time might guarantee access to have been emerged about gender boolean heterogenous set of resources and variables whereas 2) there is apparently no information, otherwise not reachable from reason why the previous BP activity an unique group as remembered by should be relevant for a dependent Aldrich and Kim (2007). This is clearly variable but not for another. The expected to affect the natural effectuation ambiguousness of some findings level, since the network dimension is right underlined the primary doubts claimed in one of its pillars. The startup team the wide literature (Phillips & O’Reilly, 86 numerosity is instead an evident proxy of work aptitudes surely lead to worse how many external ties the team is performances in decisional processes too. potentially able to achieve: the more the Referring to what written in the recent team members are, the more they could study of Alessandri et al. (2018), the work merge their different contacts with the aim engagement has been tracked through the of creating a stable and loyal business control variable av_weekly_hrsxmember network. The su_numer independent (average number of hours that each team variable is positively related to both member weekly invests on the startup) scientificity and effectuation in every and resulted as statistically significant regression model and always with throughout all the regression models, with extreme significance (p<0.01), except in positive influence on both natural levels of specification 5.b (no relationship with scientific and effectuative approach. Fixed scientificity degree) – that result is in line the dependence between hours of with its control functionality, since the engagement and decision-making capability of measuring the network process, it remains quite hard to link such extent was expected to be crucial more in engagement measure with specific effectuative than scientific approach, and individual traits since they did not appear this is what underlined by the exception being significant in any specification. model (5.b); anyway, please consider that Additionally, as discussed by Amit (2001), this is only a starting point in order to set also the prior experience in business plan more in-depth analyses in the future, writing demonstrated to be influencing: it given the lack of useful information in the has positive coefficient compared to the dataset at time zero. We must also innate effectuative approach (models 1.a consider that use of psychological traits as and 2.a, p<0.05), thus proving a real tie control variable has not been fruitful, since between the practical business plan no particular relationship was found and experience and the capabilities needed in this lack of evidences is in opposition with the effectuative framework. By the way, it a huge number of papers in literature is again not possible to link such variable (Reagans et al., 2004; Arif T., 2015; to the individual psychological capital Jurkevičienė et al., 2018). since no evidence emerged in that sense. In conclusion, no findings have arised Looking at the third research question, about the psychological cluster, in unfortunately no variable underlying to opposition to what suggested by literature the psychological capital (Wang et al., about traits such as risk appetite (Su-li & 2019) resulted as influencing the decision- Ke-fan, 2011), self-esteem and openness making process, except for the toward novelty (Clark & Wiesenfeld, Performance Avoid Orientation: 2017), bias presence (York et al., 2014) and according to the dimensions defined in a initial personal motivations (Shah & narrow group of studies (e.g., St-Jean & Tripsas, 2007-2012). Tremblay, 2020; Uy et al., 2017), the PAO variable in model 3.b negatively affects Instead, the fourth research question both scientificity and effectuation with aimed to investigate whether secondary enough good p-value (p<0.05), and that discriminants like kind of offering and was a predictable effect given that toxic industry can affect someway the innate

87 decision-making aptitudes. Indeed decisions can derive from errors in Schleimer and Shulman (2011) discovered feedback interpretation and test phase; by positive differences in profitability coming the way, we must consider that the final from product nature compared to the outcome could be distorted by seniority service world, so that also differences in effects, which can maybe introduce the used decisional processes were deviations in the evidences from software expected, maybe due to the physical sector given its dynamic and young nature of product vis-à-vis the more nature. Finally, Anna et al. (2000) abstract service concept. In the experiment suggested for including the gender feature analyzed by this thesis work, the offerings as control to investigate potential internal appear being positively correlated with differences into the industry effect. In the sole scientific approach (and always particular, they proposed to understand with optimum values of statistical whether the business owners in significance) when they are services. It can traditional-men (or symmetrically in be due to the fact that the flexible nature of traditional-women) industries show the services makes easier to create different decisional aptitudes if compared continuous feedback exchange with to non-traditional ones. Unfortunately, it market and customers, so that phases of has been not possible to scrutinize such test and validation come to be facilitated point given the lack of arised evidences thanks to the possibility of completing the about gender differences within the entire process by only using online means, treated experiment, both in terms of team with consequent less time wasting and composition and heterogeneity. more efficient learning. Yet, Mitchell and Now we come to the fifth research Shepherd (2010) proved that decision- question: better understanding the ties making tends to be excessively risky and between academic influence and decision- inconsistent when acting in more dynamic making process. Looking at the industries: our results seem to partially background type at level team (rather than corroborate their findings, because leader), in model 1.b when the team belonging to the software industry majority currently attends economics negatively related to the sole innate courses then the scientificity degree is effectuation degree, and the registered p- higher (p<0.1), while in model 1.a when value is robust (p<0.01) in the models the team includes at least one member explaining more variance (i.e. 5.a and 5.b), studying STEM and at least one member namely those recording the best R- studying economics subject then the squared values (whereas p<0.1 in models effectuative aptitude is greater. Such 1.b and 3.b). Please note that the software- difference born thanks to the academic sector boolean is the unique one having influence is in line with what told in enough numerous records (around 9% of literature (Greenwald and Banaji, 1995; the whole sample) in that category, so that Miozzo and Di Vito, 2016; Toma, 2020). the results should be sufficiently reliable. Furthermore, the after-codification A relationship would have been expected average academic level per team with the natural scientificity level rather (av_aclvlyrs_right) positively affects the than the effectuative approach, seen that scientific approach in model 1.a (p<0.1), excessively risky and inconsistent 88 but please note that the three just cited entrepreneurs to understand competitive variables registered poor levels of structure and market strengths, quality significance, that are always near to 10% in standards and most profitable trends. It p-value. Excellent significance is instead improves capability of performing more guaranteed (p<0.01) by the boolean precise predictions, along with indicating when team majority is engaged diminishing the usually observed effects with a M.Sc. while attending IVL program of discouragement that emerge after the (msc_ing_maj): it is positively linked to initial instants; thus, experience influences both innate scientificity and effectuation in decision-making thanks to a better models 2.a, 2.b and 5.b – therefore an high comprehension. The outcomes from academic level in progress can provide the STATA indicate that correlation exists and needed technical insights to conduct always presents positive verse. By using appropriate test phases and use right data the team as reference, the average amount validation tools (scientific approach), of different industries in which members along with offering the amount of have cultivated direct experience experience necessary for balancing (av_numindustries_exp) is strongly linked abilities of execution, network, control and to higher values of scientificity and flexibility. Yet, according to Chatterji et al. effectuation into all the regression models (2019), too high academic achievements analyzed. More in particular, even if the make become resistant to the capabilities significance level is nearly always robust needed in the effectuation framework: this (p<0.01), the latter seems being slightly is what arised from who already obtained higher for effectuation in models 5.a, 5.b, a Master post-lauream (master_maj) or 1.a and 2.b – such is consistent with what PhD (phd_maj), since these two described in literature, namely the dimensions negatively related with effectuative approach is someway strictly effectuation into the model 2.a – even if we linked to the experience dimension so that have to underline that the significance the effect of this variable was expected to value is not widely robust (0.05

(p<0.01) in models 5.a and 5.b, whereas though the variable fem_oneplus the second one is significant (p<0.05) in demonstrated to be the more relevant into models 1.a and 2.a . Again, this is to give all the regression models - as it could be evidence about the prominent experience expected according to the discoveries by weight on the effectuation framework, Apesteguia (2012) – they did not get while showing its inferior influence on the significant correlation with dependent natural scientific approach. Finally, variables in any statistical specification. As despite the positive influence on previously said in prior chapters, only the scientificity into a number of models (1.a- Blau’s gender index showed significant 1.b-2.a-2.b), an interesting side effect is negative correlation with the natural that an experience proxy, namely the scientific approach (model 5.b.RC), number of firms previously established by looking like such difference leads worse each participant, is strongly (p<0.01) but relationship conflict management which negatively related to the effectuative outperforms the benefit coming from a nature in models 5.a and 5.b . The better task conflict management. Anyway, phenomenon can be due to the poor we have to underline that also in this case numerosity of such variable, given that the the p-value is weak (8%). The effect does records in av_numestabl are mostly (>90%) not keep the same within the specification blank values. Instead, in opposition to where effectuation is treated: this is to Vliamos (2012) and Pugliese (2016) who corroborate that heterogeneity has largely invoked the role of opportunity more influence on the structured recognition into the effectuative approach, dynamics of feedback exchange which no evidences arise about the importance of characterize the scientific paradigm, while the industry to which the startup belongs, are not a central driver for the natural or the sector in which the entrepreneurs effectuation level. declared their experience years.

In conclusion, the seventh research question wanted to examinate potential gender effects onto the decisional processes, by following the large literature present on such topic (Frost et al., 1990; Gatewood et al., 1995; Masson et al., 2003; De Visser, et al., 2010; Apesteguia, 2012; Shepherd, 2012; Stoet et al., 2013; Zhao & Zhang, 2016; Block et al., 2018; Lee & Ashton, 2020). The drivers taken into consideration have been the gender heterogeneity, along with the boolean Figure 51: risk appetite stats according to gender variables measuring whether the team has female majority (fem_maj), female unanimity (fem_unan) or at least one Indeed, the effectuative method pays female individual (fem_oneplus); obviously more attention to enough abstract abilities the male equivalent is symmetrical. Even such as network, control, execution, and

90 flexibility, less suffering the effect brought These findings thus suggest for gender by task conflict and relationship conflict differences not linked to actual personal management. In addition, no less than the features, but rather to prejudices and risk appetite aptitude (Stanton et al., 2010; social inequality; in other words, they Van den Bos et al., 2012; Orsini et al., 2016; warned about a sort of self-fulfilling Kluen et al., 2017; Wu et al., 2020) resulted prophecy at basis of the phenomenon. as not correlated with gender, despite it According to what shown above, is has been studied so far by a lot of evident how a certain type of researches in medicine, so gathering characteristics systematically may have numerous evidences with accurate impact on natural scientificity and biomedical electronic means at diverse effectuation degrees belonging to each levels. They provided explanations by entrepreneurial team, with a great weight hormonal, instinctual, loss sensitivity and on the decision-making process. In light of neural standpoint. Please view Figure 51, the outstanding and superior where every statistical property about risk performances registered by startups appetite is completely aligned between applying scientific and effectuative female and male entrepreneurs (N=542) – approach – as furnished in literature thus again highlighting the lack of gender above all by Camuffo et al. (2017) and evidences throughout the experiment of Sarasvathy et al. (2001, 2003) – this thesis interest. This results perhaps confirms the work brings to conclude that the intuitions proposed by Carr (2010) and institutional bodies as well as more Tinkler (2016). The first paper found that technical entities (e.g. government, decision-making process can be affected universities, accelerators, incubators and by concerns about stereotypes and so on) should promote the most balanced identity devaluation, rather than attribute entrepreneurial team compositions by gender differences to innate and stable exploiting findings like that, in order to factors. Tinkler instead decided to foster the positive impacts that are likely investigate venture capitalists’ funding to be observed on decisional activities. The decisions in high-growth and high-tech totality of the analyses executed in this entrepreneurship, so discovering that: 1) thesis work took into consideration a women received higher evaluation by VCs sample composed of 305 mainly Italian when the assessment moment happened startups, with a total of 542 entrepreneurs. with close contact; 2) when in presence of Therefore, it could be interesting to technical background of both male and replicate the elaboration on a wider female entrepreneurs, the VC evaluation sample and consisting of more did not register variations among genders; international composition. Additionally, 3) when technical background was absent the work presents limitations and the and prior performance information variance explained by the considered ambiguous, the female entrepreneurs variables should be extended thanks to received lower evaluations than male non- following researches. For instance, the use technical entrepreneurs, therefore women of certain industries could be reviewed, so were supposed to be less competent and as to increase the numerousness of each having less leadership ability when sub-category and maybe assign the available information was insufficient. 91 startups to different clusters by exploiting the RAs is likely to increase with the IVL a new codification type. Yet, a more in- program evolution. depth investigation on gender effects could be implemented, trying to corroborate (or invalidate) the surprising findings here detected. Also the psychological capital influence deserves further detailed study, because the not statistical relevance we observed in this work could be due to biases introduced through tendentious questions (proposed in interview/surveys) or could derive from errors by research assistants in assigning marks at time zero, that is likely to happen given the extreme learning economies usually observed in difficult evaluation tasks like that. In conclusion the heterogeneity influence, as well as the ambiguous effects on innate levels of scientificity and effectuation deserve to be further investigated. The moderate and poor composition heterogeneity which was observed through the chosen sample has imposed a large set of unusable Blau’s indicators, because it is recognized that low regressors’ variance calls for major OLS estimators’ variance, so causing important problems for the OLS regression reliability. All this is mostly true in light of the not completely expected strong correlation between the two dependent variables at time zero. Indeed, as evident in Figure 37, the two variables of interest seem to fall on an almost perfect linear approximation, so provoking a kind of uncertainty about the existing differences and similarities. Anyway, it is predictable that the following analyses which are going to be executed throughout all the time touchpoints will lead to different relationships, given that the scanning accuracy level acquired by

92

APPENDIX

Appendix A: main features of Stevens’ measurement scales theory

-

-

verse

chi-squared

t-Student, F-Fisher t-Student,

SIGNIFICANCETEST

-

-

-

CORRELATION

Pearson coefficient Pearson

Spearman correlation Spearman

-

fractiles

H information H

percentage variation percentage

DISPERSIONINDICATOR

standard deviation, variance deviation, standard

-

mode value mode

median value median

arithmetic mean arithmetic

CENTRALITYINDICATOR

geometric mean, harmonic mean harmonic mean, geometric

similarity

exponents

permutation

increasing monotonic functions monotonic increasing

ADMITTED TRANSFORMATION ADMITTED

linear function (e.g. similarity & translation) & similarity (e.g. function linear

ordering

non-equivalence

equality between ratios between equality

EMPIRICALOPERATION

equality between intervals between equality

equality between both intervals and ratios and intervals both between equality

Ratio

SCALE

Ordinal

Nominal

Linear (Interval) Linear Linear (logarithm) Linear

93 bu_alrestabl bu_expbp bu_curacbg bu_curstude bu_othcom bu_lvlstu bu_maxaclvl bu_region bu_age bu_gender av_weffectuat av_effectuat av_wscient av_scient suind_agr 0,133151 0,034952 0,039004 -0,04928 -0,04613 -0,03014 -0,03162 -0,01025 -0,05158 -0,03275 -0,06682 -0,00453 -0,05141 -0,07357 ic suind_oth 0,005136 0,052193 0,018111 -0,00592 -0,04081 -0,03576 -0,04809 -0,03806 -0,05328 -0,08571 -0,03409 -0,05976 0,10135 -0,0505 suind_co -0,02247 -0,02044 -0,03713 -0,03997 -0,02282 -0,09959 -0,07432 -0,09631 -0,07643 -0,08981 -0,05401 -0,07729 -0,08047 mmun -0,0702 suind_ani 0,053516 0,025007 0,077972 -0,06704 -0,05654 -0,00538 -0,00463 -0,03089 #DIV/0! -0,02948 -0,04045 #DIV/0! -0,02268 mcare -0,0338 suind_pu 0,030942 0,033519 0,090773 -0,03089 -0,04183 #DIV/0! -0,02948 -0,04045 #DIV/0! -0,05166 -0,02268 -0,03246 blishing 0,08822 -0,0338 suind_ele -0,12211 -0,12307 -0,08586 -0,08242 -0,04909 -0,00587 -0,05714 -0,04685 -0,06428 -0,05877 -0,08209 -0,03604 -0,05159 -0,05371 ctro suind_en 0,047189 0,086298 0,053516 0,102658 0,093114 0,095275 0,133151 0,063341 0,077972 -0,03091 -0,03114 -0,02268 0,05328 -0,0338 erg suind_fin 0,020432 0,010052 0,026017 0,074639 -0,00329 -0,00912 -0,06622 -0,01618 -0,07643 -0,02126 -0,00119 -0,02929 0,08725 ance 0,0294 suind_fo 0,056458 0,079292 0,089806 0,017666 -0,04708 -0,05998 -0,05159 -0,04901 -0,03394 -0,00559 -0,04487 -0,08526 -0,06659 0,04607 od suind_ed 0,090758 0,080926 0,109917 0,112471 0,010797 0,170448 0,053312 0,024241 0,024066 0,061762 0,014273 -0,04631 0,15153 0,07762 uc suind_har 0,010926 0,020257 0,002911 0,002974 0,121914 0,085308 -0,02181 #DIV/0! -0,02081 -0,02856 #DIV/0! -0,01601 -0,02292 -0,02386 dw suind_ent 0,090614 0,092361 0,086331 0,085939 0,042861 0,038253 0,080655 -0,06368 -0,00714 -0,04554 -0,08402 -0,01726 -0,0043 -0,0396 ert suind_fas 0,016015 0,016444 0,033131 0,065995 0,072275 0,036681 -0,02416 -0,02404 -0,08149 -0,03371 -0,03872 -0,03602 -0,01054 -0,03708 h suind_he 0,051842 0,013363 0,007206 0,040468 -0,03648 -0,05886 -0,03394 -0,02796 -0,02685 -0,04487 -0,08219 -0,02835 -0,02916 0,04561 alth suind_ind -0,03069 -0,03035 -0,10696 -0,10744 -0,05386 -0,07293 -0,04648 -0,05141 -0,07053 -0,04782 -0,09008 -0,03955 -0,05893 userv -0,0566 suind_ho 0,006457 0,054097 0,056081 0,012782 0,018719 0,005721 0,004974 -0,07891 -0,05714 -0,02145 -0,06811 -0,05097 -0,04279 meserv 0,00567 suind_sw 0,032116 0,031587 0,147859 0,144359 0,117851 0,158023 0,249102 0,201885 0,258774 0,193224 0,066514 -0,05007 -0,00514 0,18671 suind_tra 0,058016 0,002502 0,010279 0,042251 -0,12465 -0,11445 -0,10154 -0,10173 -0,06622 -0,06811 -0,03318 -0,04869 -0,01848 nsplog 0,05247 0,021235 0,021321 0,120255 0,093209 -0,00985 -0,12082 -0,17626 -0,07651 -0,03534 -0,02278 -0,07809 -0,03321 -0,00935 av_age -0,0118 0,043301 0,127272 0,062522 0,022279 0,174359 0,044175 0,133275 0,055858 0,113016 0,080738 0,015151 -0,02009 Center 0,05099 0,13214 0,008265 0,009819 0,068068 -0,03589 -0,03926 -0,04041 -0,03126 -0,06308 -0,01395 -0,04346 -0,11438 -0,03634 -0,09443 -0,06161 North 0,043163 0,037683 0,040779 0,004203 0,025625 0,062691 0,013628 -0,09253 -0,09115 -0,06652 -0,06903 -0,03437 -0,07583 -0,08281 South fem_maj 0,034751 0,030467 0,130137 0,041291 0,068128 0,046428 0,207687 0,111864 0,047289 0,443253 -0,01906 -0,01914 -0,06975 -0,03106 fem_una 0,002661 -0,09964 -0,03407 -0,05908 -0,16683 -0,08137 -0,12664 -0,13268 -0,06804 -0,09868 -0,16395 -0,1015 -0,0361 -0,1123 n fem_one 0,039949 0,038134 0,088871 0,082718 0,270674 0,077066 0,171884 0,219822 0,037592 0,209147 0,658559 0,12784 0,15496 plus 0,2514 su_numer 0,415243 0,412356 0,373348 0,367048 0,260972 0,329156 0,243988 0,305799 0,356883 0,383626 0,449247 0,180408 0,362993 0,364106 y_hrsxme av_weekl 0,337068 0,283867 0,285792 0,105833 0,054194 0,053211 0,014359 0,094603 0,005947 0,140585 0,056835 0,021943 -0,03428 0,33176 mber su_phase 0,482647 0,483904 0,311159 0,081671 0,043628 0,041232 0,057274 0,070206 0,028771 0,180848 -0,05342 -0,02717 0,30842 0,1326 0,052792 0,061566 -0,52333 -0,52075 -0,37509 -0,37357 -0,07567 -0,12872 -0,10535 -0,07558 -0,08048 -0,07867 -0,10915 -0,14892 su_ph1 0,304835 0,300823 0,231469 0,231548 0,098021 0,112691 0,021133 0,040453 0,106432 -0,06325 -0,05569 -0,00536 -0,09509 su_ph2 0,01831 0,218785 0,219797 0,219444 0,131486 0,082254 0,052916 0,153491 0,120638 0,053382 0,104325 0,185342 0,061105 su_ph3 0,21458 0,03474 0,089532 0,095832 0,080229 0,047793 0,015604 -0,03281 -0,03251 -0,01059 -0,04648 -0,05141 -0,04782 -0,00235 -0,03955 -0,05893 su_ph4

Appendix B: extract from correlation matrix – part I

94

0,265479 0,268506 0,148959 0,145616 0,006583 0,034246 0,098698 0,001653 -0,04648 -0,03232 -0,03942 -0,04782 -0,06125 su_ph5 0,06914 prod_off -0,11055 -0,12027 -0,22311 -0,21915 -0,09112 -0,07071 -0,12329 -0,01987 -0,05844 -0,06599 -0,10486 -0,12426 -0,08325 -0,01465 er serv_offe 0,064211 0,073972 0,195805 0,191431 0,094154 0,101413 0,159593 0,022066 0,037949 0,104433 0,128443 0,156267 0,049288 0,07621 r bothcom 0,079133 0,078093 0,031231 0,034056 0,005549 -0,01475 -0,06644 -0,00619 -0,09109 -0,05596 -0,07398 -0,06863 b_offer 0,03242 -0,0886 av_weekl y_hrsons 0,304049 0,311536 0,252668 0,256027 0,019142 0,065275 0,053444 -0,04126 -0,00822 -0,03602 -0,02239 -0,04292 -0,05255 -0,05227 u rs_biased av_aclvly 0,068217 0,097068 0,106816 0,088367 0,063647 0,013723 -0,05514 -0,06366 -0,04429 -0,04603 -0,11908 -0,02278 -0,02551 -0,00699 maj_currs 0,081959 0,120393 0,190669 0,228562 0,126647 0,102686 0,023367 -0,04184 -0,04311 -0,03643 -0,05529 0,02032 -0,0373 tem 0,1229 maj_curr 0,026843 0,027645 0,064893 0,045855 0,011039 0,116828 -0,02111 -0,02752 -0,06301 -0,12874 -0,11433 -0,07405 -0,05525 -0,00659 eco maj_curr 0,018263 0,042419 0,183934 0,023707 -0,03811 -0,04099 -0,00995 -0,04748 -0,06035 -0,01028 -0,01923 -0,02462 -0,06464 0,0218 oth rep_currs 0,036634 0,040048 0,047749 0,113725 0,055097 0,048963 -0,11408 -0,11608 -0,12822 -0,12777 -0,07379 -0,01064 -0,0586 tem -0,011 rep_curre 0,011623 0,013743 0,022651 0,003489 0,046777 0,015893 -0,02978 -0,03523 -0,06164 -0,08091 -0,02277 -0,05768 -0,10657 -0,05383 co rep_curro 0,053969 0,051956 0,001396 0,030915 -0,03266 -0,03425 -0,04174 -0,04779 -0,04101 -0,01517 -0,03513 -0,00853 -0,05685 -0,0844 th av_aclvly 0,156501 0,146496 0,163315 0,155474 0,158038 0,175509 0,140478 0,230734 0,107586 0,166056 rs_right 0,13774 0,18424 0,17969 -0,0254 av_worke 0,023888 0,025756 0,004894 0,003791 0,086764 0,001002 -0,14352 -0,11657 -0,12956 -0,09131 -0,06706 -0,06938 -0,04348 xpyrs -0,08 av_worke xpyrs_sui 0,106861 0,109892 0,067642 0,066007 0,149305 0,046808 0,010425 0,028006 -0,01473 -0,12148 -0,05696 -0,00536 -0,05768 -0,02722 nd av_numin dustries_ 0,236582 0,226163 0,187407 0,188068 0,061769 0,008471 -0,03166 -0,07844 -0,06879 -0,02257 -0,06541 -0,01775 -0,00166 -0,06754 exp av_execy 0,102733 0,098749 0,089384 0,087582 0,067709 -0,10006 -0,09743 -0,10426 -0,02471 -0,04293 -0,11294 -0,06657 -0,04064 -0,04302 rs alr_establ 0,162336 0,152604 0,133433 0,134203 0,013775 -0,10189 -0,10822 -0,06301 -0,04502 -0,00462 -0,06262 -0,08667 -0,01139 0,25425 av_nume 0,124302 0,117532 0,157389 0,157027 0,205573 0,006317 -0,09534 -0,03398 -0,05601 -0,07416 -0,02202 -0,0879 -0,0065 stabl -0,065 0,220707 0,218736 0,151731 0,151786 0,089147 0,063498 0,044844 0,043821 0,052648 0,043138 -0,03397 -0,12094 -0,00145 -0,00521 exp_bp ecomgt_c 0,146507 0,136198 0,170593 0,164739 0,082746 0,108467 0,068162 0,101023 0,083053 0,052116 0,002485 0,074375 -0,14221 0,09294 rs entrep_cr 0,065123 0,067191 0,135311 0,134344 0,071367 0,036354 0,043554 0,042876 0,040206 0,005747 -0,00214 -0,10355 -0,03352 0,04506 s 0,064962 0,066797 0,036235 0,035525 av_vert1 0,090317 0,162191 0,078425 0,115011 0,129433 0,085592 0,129097 0,089754 0,153152 0,048535 av_horiz1 0,118846 0,117895 0,092232 0,086524 0,135402 0,147841 0,000749 0,083189 0,140779 0,149854 0,134266 0,085402 0,11741 0,12191 0,056154 0,060414 0,044118 0,039259 av_vert2 0,112111 0,103212 0,093901 0,024824 0,105715 0,064471 0,091162 0,089219 -0,03646 0,03627 av_horiz2 0,190645 0,185974 0,199655 0,193386 0,154133 0,108393 0,113132 0,115116 0,113979 0,147051 0,074023 0,166718 0,115744 -0,06186 intern_ne 0,313716 0,295848 0,288977 0,319648 0,270673 0,082855 0,218755 0,322769 0,263669 0,352122 0,165992 0,235392 0,238754 0,30875 tw diff_quali 0,087231 0,079782 0,075087 0,074068 0,008317 0,074559 0,050746 0,057953 0,001873 -0,02154 -0,00167 -0,09981 -0,0133 -0,0226 ty diff_desig 0,063085 0,063502 0,019559 0,014485 -0,03779 -0,03916 -0,06211 -0,10134 -0,02238 -0,00848 -0,02086 -0,0548 -0,0471 0,0158 n diff_usab 0,070252 0,072624 0,139324 0,134157 0,103884 0,056173 0,059184 0,068624 0,079731 0,067514 -0,01442 -0,02386 -0,00584 -0,00291 diff_prco 0,079942 0,066539 0,070259 0,064403 0,005747 0,033012 0,066768 0,025126 -0,02927 -0,04791 -0,00096 -0,01299 -0,01918 0,07926 0,022201 0,025482 0,036123 0,110628 0,050706 0,033572 0,059165 0,007502 0,081418 0,048453 diff_oth -0,02374 -0,02415 -0,00311 0,03924

Appendix B: extract from correlation matrix – part II

95

av_rlvncp 0,108299 0,106692 0,001424 0,040264 0,016178 0,032963 -0,00186 -0,06682 -0,00967 -0,09233 -0,03791 -0,04226 -0,01062 0,00184 rco av_rlvncq 0,006904 0,007986 0,016973 0,052805 0,053265 0,045245 0,044244 0,040322 0,061258 -0,00733 -0,00349 -0,05218 0,06447 0,01682 uality av_rlvncu 0,009487 0,020647 0,035097 0,095033 0,066037 0,048526 0,048608 0,062568 0,065481 0,047689 0,061075 -0,05859 0,03202 0,04973 sab av_rlvncd 0,076754 0,079529 0,000703 0,034278 0,037532 0,015552 0,019332 0,031606 0,051797 -0,00013 -0,01222 -0,08808 -0,07324 -0,00407 esign av_rlvnco 0,064322 0,069535 0,028262 0,029015 0,073992 0,036733 0,000916 0,039255 -0,01957 -0,01385 -0,03748 -0,00661 0,06305 0,00506 th av_selfes 0,177458 0,169794 0,101959 0,100744 0,094824 0,060546 0,037678 0,029765 0,048647 0,045888 0,020276 0,043427 0,014586 -0,03649 teem av_riskap 0,032681 0,028126 0,078315 0,077748 -0,00816 -0,00606 -0,03694 -0,02286 -0,04498 -0,08803 -0,06332 -0,02189 -0,02991 -0,04522 pet av_novelt -0,02255 -0,02829 -0,04063 -0,09089 -0,04487 -0,02741 -0,01926 -0,07468 -0,10431 -0,02661 -0,02036 -0,06715 -0,0398 -0,0603 y av_plann 0,013305 0,032277 0,054047 0,007066 0,028539 0,029643 -0,04187 -0,05145 -0,01088 -0,01265 -0,12484 -0,00787 0,02728 -0,0342 er av_firstlin 0,021472 0,019147 -0,05158 -0,06026 -0,02288 -0,04293 -0,12734 -0,01964 -0,07172 -0,01235 -0,04406 -0,02294 -0,0815 -0,0476 e 0,038837 0,041737 -0,03511 -0,06247 -0,02792 -0,06814 -0,05199 -0,04298 -0,04831 -0,02843 av_lgo -0,0265 -0,1069 -0,0209 -0,035 0,061588 0,056043 0,057027 0,040614 0,049222 0,023582 0,011992 -0,08797 -0,08834 -0,03864 -0,03558 -0,02777 -0,05221 av_pao -0,04 0,048981 0,085943 0,096558 0,098796 0,103183 0,096048 0,050032 0,125346 0,078655 -0,15528 -0,15342 -0,12216 -0,12074 -0,00804 av_pgo av_perfor 0,012943 0,076811 0,099649 0,092458 0,046899 0,033059 0,059955 0,097908 0,059426 -0,13705 -0,13693 -0,08626 -0,08381 -0,00707 ient 0,038085 0,042231 0,037471 -0,10863 -0,09513 -0,09404 -0,03929 -0,03928 -0,04169 -0,03677 -0,07183 -0,12156 av_rich -0,1159 -0,0297 av_caree 0,013467 0,006799 0,048831 -0,10841 -0,10923 -0,09151 -0,08918 -0,02879 -0,00281 -0,02401 -0,06915 -0,06818 -0,04061 -0,07789 r av_solvei 0,137823 0,102005 0,099411 0,004422 0,042911 -0,13788 -0,10946 -0,07384 -0,08017 -0,07207 -0,12086 -0,02504 -0,08913 0,14019 dent av_shape 0,166571 0,165368 0,142299 0,139001 0,012559 0,101812 0,002886 -0,05274 -0,02269 -0,02836 -0,03735 -0,06417 -0,02036 -0,0265 ident av_social 0,035703 0,053061 0,050376 0,006354 0,044012 0,070882 -0,01739 -0,05267 -0,13461 -0,03302 -0,04898 -0,01919 0,04636 prob -0,008 av_world 0,023219 0,022313 0,049368 0,030688 0,037628 -0,05004 -0,11818 -0,08987 -0,05977 -0,12421 -0,00217 -0,04193 0,04649 -0,0481 av_robus 0,028746 -0,05131 -0,05915 -0,05806 -0,05759 -0,02472 -0,10369 -0,10796 -0,03861 -0,04616 -0,08083 -0,02828 -0,03063 -0,0115 t av_deepl -0,11498 -0,12282 -0,08481 -0,05182 -0,10116 -0,07873 -0,11457 -0,06584 -0,10296 -0,13947 -0,03517 -0,10528 -0,0819 -0,0647 yan av_supply 0,101581 0,100908 0,054205 0,050808 0,006068 0,049494 -0,13334 -0,13466 -0,05897 -0,05047 -0,03941 -0,10344 -0,03221 -0,0298 ident av_agree 0,119618 0,114083 0,017872 0,065639 0,037256 -0,07386 -0,14944 -0,01908 -0,03368 -0,00413 -0,08434 -0,03167 0,11749 0,11746 av_wrldct 0,032335 0,003599 0,001796 0,045336 0,012698 -0,01139 -0,08064 -0,04542 -0,06995 -0,02206 0,02985 0,01319 -0,0843 0,0389 zn av_wrldpl 0,027471 0,028912 0,017436 0,022973 -0,01178 -0,11074 -0,04428 -0,04827 -0,01738 -0,10425 -0,01476 -0,00315 0,01859 0,01233 c av_comp 0,026938 0,016708 0,025866 0,032897 0,014165 -0,00234 -0,00556 -0,01947 -0,07083 -0,02951 0,01902 -0,0199 -0,0086 -0,0674 et av_advco 0,061691 0,059893 0,045229 0,046343 0,005138 0,015298 -0,01594 -0,03025 -0,11716 -0,01715 -0,00156 -0,04705 -0,04797 -0,03459 mpet av_focusi 0,145512 0,152505 0,109166 0,106649 0,040523 0,053846 0,080487 0,002193 -0,01319 -0,05042 -0,03403 -0,03888 0,00467 0,04191 dent 0,145108 0,152535 0,096417 0,092974 av_supid 0,032846 0,025194 0,069954 0,023963 -0,04379 -0,00136 -0,05162 -0,07739 -0,02609 -0,0475 ent av_social 0,017035 0,021306 0,073779 0,000666 -0,05061 -0,12218 -0,04388 -0,06923 -0,06473 -0,15609 -0,03463 -0,00597 0,00256 -0,0001 av_persu 0,015987 0,029705 0,007173 0,015538 -0,02028 -0,01559 -0,04193 -0,03604 -0,00163 -0,06101 -0,12965 -0,06492 -0,00553 0,01572 ade 0,190224 0,184885 0,156465 0,157705 av_team 0,148241 0,154669 0,101615 0,145608 0,200046 0,068607 0,138817 0,175433 0,10167 0,17712 build

Appendix B: extract from correlation matrix – part III

96

3.b 2.b 1.b 5.b 3.a 2.a 1.a 5.a 4 gen + ac_lvl+ gen yescontrol yescontrol yescontrol 3° model - model 3° control no - model 2° control no - model 1° control no 3° model - model 3° - model 2° - model 1° 4° model 4° gen effectuat effectuat effectuat effectuat effectuat effectuat effectuat effectuat effectuat scient scient scient scient scient scient scient scient scient number number of obs of 298 298 298 298 302 302 305 305 305 305 305 305 305 305 298 298 298 298 F(20,277) F(20,277) F(20,260) F(20,260) F(19,268) F(19,268) F(17,287) F(17,287) F(11,276) F(11,276) F(10,294) F(10,294) F(41,237) F(41,237) F(34,244) F(34,244) F(8,276) F(8,276) Fisher 7,52 4,94 5,77 5,82 7,06 6,76 \ \ \ \ \ \ \ \ \ \ \ \ Prob > F Prob 0 0 0 0 0 0 \ \ \ \ \ \ \ \ \ \ \ \ R-squared 0,3207 0,2725 0,3609 0,3122 0,1279 0,1760 0,3730 0,3209 0,2112 0,2067 0,3523 0,2990 0,1739 0,1551 0,5433 0,4296 0,5182 0,4110 Root MSE Root 0,71729 0,98191 0,71813 0,98544 0,82599 0,71124 0,97645 0,77091 0,71235 0,97758 0,77945 0,63583 0,93998 0,64365 0,94134 1,0546 1,0198 1,0398 suind_agric suind_oth suind_commu - n ^ suind_animcar ------e *** *** *** *** *** *** * * suind_publishi + + + + + + + + + ng *** *** *** *** *** *** *** * ^ suind_electro - - - - - ** ^ ^ ^ * suind_energ + ------*** *** *** ** ** ** ** ^ * suind_finance suind_food - - - - ^ ^ ^ ^ suind_educ + + + + + + ** ** ^ * ^ * suind_hardw + + + + + + + + + + + *** *** *** *** *** *** *** ** ** * ^ suind_entert + ^ suind_fash suind_health suind_induser - - - - v ** ** ^ * suind_homese rv suind_sw + - - - - - *** *** * ^ ^ * suind_transpl - - - og *** ** * av_age Center North South - - - * ^ * fem_maj fem_unan fem_oneplus

Appendix C: recap of all the regression models – part I

Legend: ^ p<15% (alert, no relevant statistical significance) ; * p<10% ; ** p<5% ; *** p<1% 97

su_numer + + + + + + + + + + + + *** *** *** *** *** *** *** *** *** *** ^ * av_weekly_hr sxmember + + + + + + + + + + + + *** *** *** *** *** *** *** *** *** *** ** ** su_phase su_ph1 su_ph2 + + + + *** *** *** *** su_ph3 + + *** *** su_ph4 + - - * ^ * su_ph5 + + + *** *** * prod_offer serv_offer + + + + + *** *** *** ** ^ bothcomb_off + + + er ** ^ * av_weekly_hr sonsu av_aclvlyrs_bi ased maj_currstem maj_curreco + * maj_curroth atleast1_stem + + eco * ^ rep_currstem rep_curreco rep_curroth av_aclvlyrs_rig + + + ht ^ * ^ av_workexpyr s av_workexpyr s_suind av_numindust ries_exp + + + + + + + + + + + + *** *** *** *** *** *** ** ** ** * ^ ^ av_execyrs alr_establ + + *** *** av_numestabl + + + + - - *** *** ** ^ * * exp_bp + + + + + + ** ** ^ ^ ^ ^ ecomgt_crs + ^ entrep_crs av_vert1 av_horiz1 av_vert2 av_horiz2 + + ^ ^ intern_netw + + + + + + + + + + + + + + + + *** *** *** *** ** ** ** ** * ^ * ^ * * * * Appendix C: recap of all the regression models – part II

Legend: ^ p<15% (alert, no relevant statistical significance) ; * p<10% ; ** p<5% ; *** p<1%

98

diff_quality diff_design diff_usab diff_prco diff_oth av_rlvncprco av_rlvncqualit y av_rlvncusab av_rlvncdesig n av_rlvncoth av_selfesteem av_riskappet av_novelty av_planner av_firstline av_lgo av_pao av_pgo - - - - ** ** ** * av_perforient - - * * av_rich av_career av_solveident av_shapeident av_socialprob av_world av_robust av_deeplyan av_supplyiden t av_agree av_wrldctzn

Appendix C: recap of all the regression models – part III

Legend: ^ p<15% (alert, no relevant statistical significance) ; * p<10% ; ** p<5% ; *** p<1%

99

av_wrldplc av_compet av_advcompe t av_focusident av_supident av_social av_persuade av_teambuild av_scient av_effectuat bu_gender bu_age + + + + + + + + + + *** ** ** ** ** ** ^ * ^ ^ bu_region bu_maxaclvl bu_maxaclvl_ nophd bu_lvlstu bu_othcom bu_curstude + + + + ^ ^ * * bu_curacbg bu_expbp bu_alrestabl + ^ bsc_maj - ^ msc_maj master_maj - - * ^ nothing_maj phd_maj - - - ^ * * bsc_ing_maj msc_ing_maj + + + + + + *** *** *** *** *** * master_ing_m - aj ^ postlaur_ing_ maj

Appendix C: recap of all the regression models – part IV

Legend: ^ p<15% (alert, no relevant statistical significance) ; * p<10% ; ** p<5% ; *** p<1%

100

Appendix D: crossed scatter plots referred to the sub-variables of interest

av_scientth

5

av_scienthp

0 5

av_scienttest

0 4

2 av_scientvalid

0 5

av_scientdecthresh

0 5

av_effectuatbird

0 5

av_effectuatloss

0 4

2 av_effectuatquilt

0 5

av_effectuatlemon

0 5

av_effectuatpilot

0 0 5 0 5 0 5 0 2 4 0 5 0 5 0 5 0 2 4 0 5

101

102

References Adler, P., Kwon, S.W. (2002) Social Capital: Prospects for a New Concept. The Academy of Management Review, 27(1), 17-40. http://www.jstor.org/stable/4134367

Adner, R., Levinthal, D. (2004) What is not a real option: Considering boundaries for the application of real options to business strategy. Academy Management Review. 29(1): 74-85

Aldrich, H., Kim, P. (2007) Small worlds, infinite possibilities? How social networks affect entrepreneurial team formation and search. Strat. Entrepreneurship J., 1: 147–165. DOI: 10.1002/sej.8

Alessandri, G., Consiglio, C., Luthans, F. and Borgogni, L. (2018) Testing a dynamic model of the impact of psychological capital on work engagement and job performance. Career Develop. International, Vol. 23 No. 1, pp. 33-47. https://doi.org/10.1108/CDI-11-2016-0210

Amit, R., MacCrimmon, K.R., Zietsma, C., & Oesch, J. M. (2001) Does money matter? Wealth attainment as the motive for initiating growth-oriented technology ventures. Journal of Business Venturing, 16: 119-143

Anna, A.L., Chandler, G.N., Jansen, E., Mero, N.P. (2000) Women business owners in traditional and non-traditional industries. Journal of Business Venturing, Volume 15, Issue 3, Pages 279-303. https://doi.org/10.1016/S0883-9026(98)00012-3

Apesteguia, J., Azmat, G., Iriberri, N. (2012) The Impact of Gender Composition on Team Performance and Decision Making: Evidence from the Field. Management science, vol. 58, no. 1, https://doi.org/10.1287/mnsc.1110.1348

Arif, Tasleem (2015) The Mathematics of Social Network Analysis: Metrics for Academic Social Networks. International Journal of Computer Applications Technology and Research 4(12):889-893. DOI: 10.7753/IJCATR0412.1003

Aven, B., Hillmann, H. (2018) Structural Role Complementarity in Entrepreneurial Teams. Management science, vol. 64, n. 12, https://doi.org/10.1287/mnsc.2017.2874 Bacco, F., et al. (2020) Different Approaches to Entrepreneurial Decision-Making: A Field Experiment in Italy. AEA RCT Registry. September 07. https://doi.org/10.1257/rct.6263- 1.1

Barron, E., Amoros, J. (2019) When science meets the market: a multidisciplinary approach of entrepreneurship education. IEEE Integrated STEM Education Conference Princeton, DOI: 10.1109/ISECon.2019.8882028 Block, K., Gonzalez, A.M., Schmader, T., Baron, A.S. (2018) Early Gender Differences in Core Values Predict Anticipated Family Versus Career Orientation. Psychol Sci. 29(9):1540-1547. doi: 10.1177/0956797618776942 Bloem, Jeffrey R. (2018) How Much Does the Cardinal Treatment of Ordinal Variables Matter? An Empirical Investigation. 23 Feb, University of Minnesota

103

Butler, Garg, Stephens (2019) Social Networks, Funding, and Regional Advantages in Technology Entrepreneurship: An Empirical Analysis. Information systems research, vol. 31, no. 1, https://doi.org/10.1287/isre.2019.0881 Camuffo A., Cordova A., Gambardella A. (2017) A Scientific Approach to Entrepreneurial Decision-Making: Evidence from a Randomized Control Trial Carr, P.B., Steele, C.M. (2010) Stereotype threat affects financial decision making. Psychol Sci, 21(10):1411-6. doi: 10.1177/0956797610384146

Chatterji, A., Delecourt, S., Hasan, S., Koning, R. (2019) When does advice impact startup performance? Strategic Management Journal, https://doi.org/10.1002/smj.2987

Clark, T., & Wiesenfeld, D. (2017) 3 things are holding back your analytics, and technology is not one of them. Harvard Bus. Rev.

Cooper, A.C., Woo, C.Y., Dunkelberg, W.C. (1988) Entrepreneurs’ perceived chances for success. Journal of Business Venturing, 3(2), 97-108

De Visser, L., Van der Knaap, L.J., Van de Loo, A., Van der Weerd, C., Ohl, F., Van den Bos, R. (2010) Trait anxiety affects decision-making differently in healthy men and women: towards gender-specific endophenotypes of anxiety. Neuropsychologia, 48(6):1598-606. DOI: 10.1016/j.neuropsychologia.2010.01.027

Eisenmann T., Ries E., Dillard S. (2013) Hypothesis-driven entrepreneurship: The Lean Startup. Harvard Business School, Note 90-812-095

Fairlie, RW, Miranda, J (2017) Taking the leap: The determinants of entrepreneurs hiring their first employees. Journal of Economics and Management Strategy, 26 (1), 3-34

Fern, M. J., Cardinal, L. B., & O’Neill, H. M. (2012) The genesis of strategy in new ventures: Escaping the constraints of founder and team knowledge. Strategic Management Journal, 33: 427-447

Fitzgerald, D.R., Mohammed, S., Kremer, G.O. (2017) Differences in the way we decide: The effect of decision style diversity on process conflict in design teams. Personality and Individual Differences, Volume 104, Pages 339-344

Forlani, D., Mullins, J.W. (2000) Perceived risks and choices in entrepreneurs' new venture decisions [J]. Journal of Business Venturing, vol. 15, no. 4, pp. 305-322

Franceschini, F., Galetto, M., Maisano, D., Mastrogiacomo, L. (2019) “Quality engineering”. Fourth edition, Clut ed. Freedman, M.B., & Jaggi, B. (1992) An Examination of the Impact of Pollution Performance on Economic and Market Performance: Pulp and Paper Firms. Journal of Business Finance & Accounting, Vol. 19, No. 5, pp. 697-713 Frisch, R. (1926) On a problem in pure economics. Norsk Matematisk Forenings Skrifter, Oslo. 1 (16): 1–40

104

Frost, R.O., Marten, P., Lahart, C. et al. The dimensions of perfectionism. Cogn Ther Res 14, 449–468 (1990). https://doi.org/10.1007/BF01172967 Gans, J., Stern, S., Wu, J. (2017) Foundations of entrepreneurial strategy. http://dx.doi.org/10.2139/ssrn. 2844843

Gatewood, E. J., Shaver, K. G., & Gartner, W. B. (1995). A longitudinal study of cognitive factors influencing start-up behaviors and success at venture creation. Journal of Business Venturing, 10(5), 371–391. https://doi.org/10.1016/0883- 9026(95)00035-7

Geremias, R.L., Lopes, M.P., Soares, A.E. (2020) Enhancing Internal Learning in Teams: The Role of Network Centrality and Psychological Capital of Undergraduate Students. HISPS Lisbon, Front Psychol, 10;11:2197. doi: 10.3389/fpsyg.2020.02197

Gilbert, D.T. (1991) How mental systems believe. American psychologist, 46(2), 107

Gioia, D., Corley, K., Hamilton, A. (2013) Seeking Qualitative Rigor in Inductive Research. Organizational Research Methods, 16(1):15-31, Sage Journals. https://doi.org/10.1177/1094428112452151

Greenwald, A., Banaji, M. (1995) Implicit Social Cognition: Attitudes, Self-Esteem, and Stereotypes, American Psychological Association, Vol. 102, No. 1,4-27

Herfindahl, O. C. (1950) Concentration in the U.S. Steel Industry. Unpublished doctoral dissertation, Columbia University

Hoogendoorn, S., Parker, S. C., & Van Praag, M. (2017). Smart or Diverse Start-up Teams? Evidence from a Field Experiment. Organization Science, 28(6), 1010-1028. https://doi.org/10.1287/orsc.2017.1158

Jehn, K., Chadwick, C., Thatcher, S. (1997) To agree or not to agree: the effects of value congruence, individual demographic dissimilarity, and conflict on workgroup outcomes. International Journal of Conflict Management, Vol. 8 No. 4, pp. 287-305. https://doi.org/10.1108/eb022799

Jehn, K., Northcraft, G., Neale, M. (1999) Why Differences Make a Difference: A Field Study of Diversity, Conflict, and Performance in Workgroups. Administrative Science Quarterly, Vol. 44, No. 4, pp. 741-763, https://doi.org/10.2307/2667054

Jurkevičienė, J., & Butkeviciene, E. (2018) Social Capital in Social Media Networks. Filosofija Sociologija 29(2). International Social Survey Programme. DOI: 10.6001/fil-soc.v29i2.3705

Kahneman, D. (2011) Fast and slow thinking. Allen Lane and Penguin Books, New York Kahneman, D., Lovallo, D. (1993) Timid choices and bold forecasts: a cognitive perspective on risk taking. Management Science, 39(1), 17-31

Kahneman, D., Slovic, S.P., Slovic, P., Tversky, A. (1982) Judgement under uncertainty: Heuristics and biases. Cambridge University Press

105

Kaplan, M. F., & Miller, C. E. (1987). Group decision making and normative versus informational influence: Effects of type of issue and assigned decision rule. Journal of Personality and Social Psychology, 53(2), 306–313. https://doi.org/10.1037/0022- 3514.53.2.306

Kerr, W.R., Nanda, R., Kropf, M. (2014) Entrepreneurship as experimentation. The Journal of Economic Perspectives, Vol. 28, No. 3, pp. 25-48 Klein, G. (2007) Performing a project premortem. Harvard Business Review. 85(9): 18-19

Klotz, A. C., Hmieleski, K. M., Bradley, B. H., & Busenitz, L. W. (2014). New Venture Teams: A Review of the Literature and Roadmap for Future Research. Journal of Management, 40(1), 226–255. https://doi.org/10.1177/0149206313493325

Kluen, L., Agorastos, A., Wiedemann, K., Schwabe, L. (2017) Cortisol boosts risky decision- making behavior in men but not in women. Psychoneuroendocrinology, 84:181-189. doi: 10.1016/j.psyneuen.2017.07.240

Knight, F. (1921) Risk, Uncertainty, and Profit. Houghton Mifflin Co., Boston, MA

Koekemoer, DJ, Kachieng’a, MO (2003) Technological entrepreneurship: financing new technology-based enterprises in South Africa. IEEE International Engineering Management Conference Cambridge, DOI: 10.1109/IEMC.2002.1038472

Koellinger, P., Minniti, M., Schade, C. (2007) “I think I can, I think I can”: Overconfidence and entrepreneurial behavior. Journal of economic psychology, 28(4), 502-527

Kristinsson, K., Candi, M., Saemundsson, R. (2015) The Relationship between Founder Team Diversity and Innovation Performance: The Moderating Role of Causation Logic. Long Range Planning. Diversity and innovation project. DOI: 10.1016/j.lrp.2015.12.013

Lee, K., Ashton, M.C. (2020) Sex differences in HEXACO personality characteristics across countries and ethnicities. Journal of personality, 88(6):1075-1090. doi: 10.1111/jopy.12551

Likert, R., & Murphy, G. (1932) A technique for the measurement of attitudes. Archives of Psychology, Vol. 22, N. 140, 55

Masson, A.M., Cadot, M., Ansseau, M. (2003) Failure effects and gender differences in perfectionism. Encephale, 29(2):125-35

McGrath RG, MacMillan IC (1995) Discovery-driven planning . Harvard Business Review 73(4): 44-54 Miozzo, M., & DiVito, L. (2016). Growing fast or slow?: Understanding the variety of paths and the speed of early growth of entrepreneurial science-based firms. Research Policy, 45(5), 964–986. https://doi.org/10.1016/j.respol.2016.01.011

Mitchell, J. R., & Shepherd, D. A. (2010). To thine own self be true: Images of self, images of opportunity, and entrepreneurial action. Journal of Business venturing, 25(1), 138-154

Nonaka, I., and Takeuchi, H. (1995) The knowledge-creating company. Vol 3/V3N1A2

106

Orsini, C., Willis, M., Gilbert, R.J., Bizon, J., Setlow, B. (2016) Sex differences in a rat model of risky decision making. Behav Neurosci, 2016 Feb;130(1):50-61. doi: 10.1037/bne0000111 Osterwalder, A., Pigneur, Y., & Tucci, C. (2005). Clarifying Business Models: Origins, Present, and Future of the Concept. Communications of the Association for Information Systems, 16, pp-pp. https://doi.org/10.17705/1CAIS.01601 Parker, S.C. (2006). Learning about the unknown: How fast do entrepreneurs adjust their beliefs?. Journal of business venturing, 21(1), 1-26

Phillips, K.W., O’Reilly, C.A. (1998) Demography and Diversity in Organizations: A Review of 40 Years of Research. Research in Organizational Behavior 20:77-140. JAI Press

Ping, Z., He Gui-bing (2004) The role of personal influence in social combination process of group decision making: an extension of SDS theory[J]. Acta Psychologica Sinica, vol. 36, no. 1, pp. 37-43

Pugliese, R., Bortoluzzi, G., & Zupic, I. (2016). Putting process on track: empirical research on start-ups’ growth drivers. Management Decision, 54(7), 1633–1648. https://doi.org/10.1108/MD-10-2015-0444)

Reagans, R., Zuckerman, E., & McEvily, B. (2004). How to Make the Team: Social Networks vs. Demography as Criteria for Designing Effective Teams. Administrative Science Quarterly, 49, 101 - 133. DOI:10.2307/4131457

Resnick, L., Levine, B., John, M., Teasley S. (1991) Perspectives on Socially Shared Cognition. American Psychological Association

Ries E. (2011) The Lean Startup: How today’s entrepreneurs use Continuous Innovation to create radically successful businesses (Crown Business, New York)

Sarasvathy S.D. (2003) Entrepreneurship as a science of the artificial. Journal of Economic Psychology. 24 2003, 203-220. University of Maryland and RH Smith School of Business

Sarasvathy S.D. (2001) Causation and Effectuation: Toward a theoretical shift from economic inevitability to entrepreneurial contingency. Academic Management Review, 26(2): 243-263

Schleimer, S.C., & Shulman, A.D. (2011). A comparison of new service versus new product development: configuration of collaborative intensity as predictors of performance. Journal of Product Innovation Management, Vol. 28, pp. 521–535

Shah, S. K., & Tripsas, M. (2007) The accidental entrepreneur: The emergent and collective process of user entrepreneurship. Strategic Entrepreneurship Journal, 1: 123-140 Shah, S. K., & Tripsas, M. (2012) When do user innovators start firms? A theory of user entrepreneurship. Working paper no. 12-078, Harvard Business School, Boston, MA Shepherd DA, Haynie JM, McMullen JS (2012) Confirmatory search as a useful heuristic: Testing the veracity of entrepreneurial conjectures. Journal of Business Venturing, 27(6): 637- 651 107

St-Jean, E., Tremblay, M. (2020) Mentoring for entrepreneurs: A boost or a crutch? Long- term effect of mentoring on self-efficacy. International Small Business Journal. Researching entrepreneurship, 1-25. DOI: 10.1177/0266242619901058 Stanton, S., Liening, S., Schultheiss, O. (2010) Testosterone is positively associated with risk taking in the Iowa Gambling Task. Horm Behav, 59(2):252-6. doi: 10.1016/j.yhbeh.2010.12.003 Stevens, S.S. (1946) On the Theory of Scales of Measurement. Science. 103 (2684): 677–680. Bibcode:1946Sci...103..677S. doi:10.1126/science.103.2684.677. PMID17750512. S2CID4667599 Stock, J.H., Watson, M.W. (2012) Introduction to Econometrics. Pearson Education, third edition

Stoet, G., O’Connor, D.B., Conner, M. et al. (2013) Are women better than men at multi- tasking? BMC Psychology 1, 18. https://doi.org/10.1186/2050-7283-1-18

Su-li, Z., Ke-fan, X. (2011) Research on entrepreneurial team members' personality traits influence on group risk decision-making. School of Management, Wuhan University of Technology, China

Tinkler, J.E., Whittington, K.B., Ku, M., Davies, A.R. (2016) Gender and venture capital decision-making: the effects of technical background and social capital on entrepreneurial evaluations. Soc Sci Res. 51:1-16. doi: 10.1016/j.ssresearch.2014.12.008

Urbano, D., Aparicio, S. (2016) Entrepreneurship capital types and economic growth: International evidence. Technological Forecasting and Social Change, 102, pp. 33 – 44

Valliere, D., Peterson, R., (2009) Entrepreneurship and economic growth: Evidence from emerging and developed countries. Entrepreneurship and Regional Development, 21 (5-6), pp. 459 – 480

Van de Ven, Andrew H., Roger Hudson, & Dean M. Schroeder. (1984) "Designing new business startups: Entrepreneurial, organizational, and ecological considerations." Journal of management 10.1: 87-108.

Van den Bos, R., Homberg, J., De Visser, L. (2012) A critical review of sex differences in decision-making tasks: focus on the Iowa Gambling Task. Behav Brain Res, 1;238:95-108. doi: 10.1016/j.bbr.2012.10.002

Van Knippenberg, D., & Schippers, M.C. (2007) Work Group Diversity. Annual Review of Psychology, 58, 515-541. http://dx.doi.org/10.1146/annurev.psych.58.110405.085546

Van Knippenberg, M. (2016) Past, present, and potential future of team diversity research: From compositional diversity to emergent diversity. Organizational Behavior and Human Decision Processes, 136:135-145. DOI: 10.1016/j.obhdp.2016.05.007

108

Veretennikova, N., Vaskiv, R. (2018) Application of the Lean Startup Methodology in Project Management at Launching New Innovative Products. 2018 IEEE 13th International Scientific and Technical Conference on Computer Science and IT, Lviv, 169-172 Vining, G. (2013) Technical Advice: Scientific Method and Approaches for Collecting Data, Quality Engineering, 25:2, 194-201, DOI: 10.1080/08982112.2013.764228 Vliamos, S. J., & Tzeremes, N. G. (2012). Factors Influencing Entrepreneurial Process and Firm Start-Ups: Evidence from Central Greece. Journal of the Knowledge Economy, 3(3), 250–264. https://doi.org/10.1007/s13132-011-0043-x

Wang, Y., Tsai, C., Lin, D., Enkhbuyant, O., Cai, J. (2019) Effects of Human, Relational, and Psychological Capitals on New Venture Performance. Front Psychol. 18;10:1071. doi: 10.3389/fpsyg.2019.01071

Wu, Y., Hall, A., Siehl, S., Grafman, J., Krueger, F. (2020) Neural Signatures of Gender Differences in Interpersonal Trust. Front Hum Neurosci, 14:225. doi: 10.3389/fnhum.2020.00225

York, J. L., & Danes, J. E. (2014). Customer development, innovation, and decision- making biases int he lean startup. Journal of small Business strategy, 24(2), 21-40

Zenger, T.R. (2016) Beyond competitive advantage: how to solve the puzzle of sustaining growth while creating value. Harvard Business School Press, Boston, MA

Zhang, Y., & Zhang, W. (2019) "How does the team expertise heterogeneity improve entrepreneurial performance? Analysis from the perspective of academic social network", Library Hi Tech, Vol. 38 No. 2, pp. 434-445. https://doi.org/10.1108/LHT-11-2018-0180

Zhao, N., Zhang, J. (2016) Gender differences in trusting strangers: Role of the target's gender. The Institute of Psychology, Chinese Academy of Sciences and John Wiley & Sons Australia, Psych J, 5(2):83-91. doi: 10.1002/pchj.120

109