Empirical investigation

I now move from state of the art towards deductive investigation (Decoo, 1996). Quanti-tative research depends on deductive reasoning (Sekaran and Bougie, 2016) and employs a selection of quantitative analysis techniques (Saunders, Lewis and Thornhill, 2015) which explore underlying relationships through numerical data (Bryman and Bell, 2015).

Previous research has showed that there are a substantial number of elements to be con-sidered under one framework umbrella which all influence the EA of new venture devel-opment, as indicated in Publications I and II. Hence, it became imperative to consider conducting empirical investigations to observe these relationships. The quantitative ap-proach would be fitting for the purpose of observing and accounting for all these elements in a two-country setting; moreover, it would make it possible to adopt a more holistic perspective to describe and explain these underlying relations in the EE context of a larger sample size.

Data collection

From interactions with the Estonian authority and founders, I understood that I could not implement the full range of items and scales developed by previous research in my sur-vey; instead, it was necessary to adopt another viewpoint and focus on basic measures to comply with the participants’ suggestions. In the EE research field, there were not many measures explicitly available at that time (Audretsch and Belitski, 2017), and every team was trying to develop their own (Corrente et al., 2019; Liguori et al., 2019). Therefore, to discover the EE and BG start-up interconnections and dynamics, I focused on two dimen-sions relevant to detecting EE performance in the perceptions of founders. I wanted to understand how critical these elements were for new venture development during the dis-covery and validation stages, as well as how these elements performed in terms of their availability and access in the local context. Thus, I extended this reasoning to all elements and established a scale from 0, indicating not critical/not available or accessible to 100, highly critical/readily available and accessible. Even though ‘availability’ and ‘accessi-bility’ refer to different features, in the context of EE, they represent the extent to which an element is present in the system. I did some preliminary testing, and it was commonly understood by the entrepreneurs. The rest of the survey questions addressed the back-ground of the founders and their new ventures and was delivered in the English language.

3.3 Research methods 57

Afterwards, I proceeded to compile the appropriate list of candidates from both countries as potential respondents to the survey. I compiled the list from internet databases, such as Start-Up Estonia, Garage48, FunderBeam, Crunchbase and ArcticStartup. The assump-tion was that all these start-ups found in the databases have been certified by their respec-tive start-up community and confirmed as new ventures that possess the potential of be-coming a BG of the next generation. I then employed a stratified sampling method (Ney-man, 1934) to target only firms which have experience in the discovery (stand-up) and validation (start-up) stages by employing filtering options of these databases based on self-reporting of their current stage; therefore, the search results returned only the most recent knowledge and experience in this regard. Subsequently, I gathered their contact details and other information from the ORBIS database. The list provided 347 potential BG start-ups from Finland and 143 from Estonia. When the survey was converted into an online format, pretested and the contact list compiled, I started to call start-up founders. I also emphasised that only founders could respond to the survey and had a verification question set up. The survey method allowed me to gather data from the grassroots to accurately comprehend the perceptions and opinions of the local new ventures.

Data quality

I received an acceptable response rate from 51 (14%) founding entrepreneurs from Fin-land and 33 (24%) from Estonia. I verified that most of them were already BG-qualified by checking their speed, scale, and scope (Kuivalainen, Saarenketo and Puumalainen, 2012). However, I was less stringent about the scope in terms of continents because digital affordances (Autio et al., 2018) enable new ventures to be global on the spot. Previous studies have also been lenient in this regard (Nummela et al., 2014), as it is mostly rele-vant in the context of larger countries and less of a concern for smaller ones (e.g. Finland, Estonia) (Choquette et al., 2017). Further expected quality issues were that some new ventures would not have revenue yet, for example, due to their young age (less than three years), and some would not be internationalised, even though they responded that they were. As I targeted only peer-reviewed ventures with BG potential, I argue that these start-ups could therefore have employed inward internationalisation (Fletcher, 2001) and could have internationalised via other value-chain activities (e.g. global team allocation, entering platform ecosystems).

Additionally, due to the utilisation of the survey method, it was vital to control for the presence of common method bias (Podsakoff et al., 2003). When reviewing EE literature, it was critical to introduce sub-elements (e.g. finance, talent) to study the ecosystem con-ditions. Hence, I argue that using this division helped to clarify distinctions between dis-similar types of elements and sub-elements to ensure that the respondents would assess these in an objective manner by relying on their own experience. Thus, this direct over-view of their differences should have mitigated the risk of common method bias. Addi-tionally, I conducted Harman’s single-factor test post hoc and did not detect any single factor affecting the analyses.

Data analysis

Analyses were conducted to conform to the research questions and the overall agenda of the empirical investigation. Firstly, all mean averages were calculated for each element in their respective stages for Estonia (Publication II, RQ 3.1), for Finland (Publication III, RQ 3.1), and for both countries combined (Publication IV, RQ 3.1); subsequently, they were ranked in descending order. I then calculated the aggregate averages as cut-off points to indicate that elements above the line were critical to BG start-up development, while I considered those below somewhat inessential for these stages. This allowed me to clearly demonstrate which elements are critical to the preliminary stages of the BG life cycle, depending on the context. Similarly, movements in criticality rankings between the stages enabled me to detect the stage-wise changes and argue for the existence of dynamic interactions in between respective elements and stages.

Secondly, I added the presence (availability/access) dimension to introduce an index ratio system exclusively for these critical elements and derived a set of performance ratios by juxtaposing the averages of perceived presence with the averages of perceived criticality for each element. These ratios demonstrated inconsistencies in performance by showing how well each element behaved in comparison to its influence on new venture develop-ment. The lowest ratios indicated hindering propensities because these elements would likely constrain new venture progress. This aided me in grouping these into strengths and weaknesses of the Estonian (Publication II, RQ 3.2) and Finnish (Publication III, RQ 3.2) ecosystems and in calculating the overall performance of each (Publication III, RQ 3.3), yielding similar results as found in previous research (Acs, Autio and Szerb, 2014). The rest of the elements were less relevant. Additionally, by utilising STATA software, I ap-plied a paired samples t-test to all elements to confirm that the previous grouping was solid and without any significant statistical differences amongst connected pairs. How-ever, the inessential elements in Estonia indicated that one was even perceived as irrele-vant to the discovery and validation stage.

Lastly, I collected data on multiple features of new ventures regarding their organisational (i.e. age, team size, service-product orientation and revenue) and international character-istics (i.e. team allocation, origin of the owners, speed of internationalisation and market scope) and, based on the data, divided the firms into binary groups. These characteristics elaborate how ecosystem elements influence specifics of BG start-up development. I then again utilised STATA by applying the analysis of variance (ANOVA) to compare the statistical significance of the founder’s perceptions to uncover relational inconsistencies between their new venture and transnational environment (Publication IV, RQ 3.4). I found nine differences in the discovery stage and seven in the validation stage. This con-cluded my empirical investigation.

Regarding the sophistication level of the quantitative investigation, I have to relate that as I was constrained and somewhat afraid of the founders being too reluctant to respond to my questionnaire if it were full of items on each element. Therefore, I felt I had no better choice than to gather data that did not allow me to apply more complex statistical

3.3 Research methods 59

methods. However, the results stand for themselves, and there is value in demonstrating that other methods would lead to similar results (Morgan, 2014, p. 11).

In conclusion, I have now presented my research design and methodology by creating a concise overview of my doctoral dissertation process. I have discussed the horizontal lay-ers of my research by elaborating its theoretical focus, publication course, empirical out-line, and my iterative thinking process. Additionally, I have presented each vertical phase of this dissertation process through a brief storyline and covered the methodological con-siderations to demonstrate my personal line of argumentation. It has been a rewarding endeavour initiated from a practical need to comprehend how to create successful start-ups, beginning with a need for founder’s roadmap and culminating in writing up my work while reflecting on my academic work and becoming a scholar. In the next chapters, I will go through the findings regarding each publication and discuss the research question in more detail, finally concluding the dissertation by presenting implications, limitations and future research avenues.

61

4 Publications and Findings

In this chapter, I discuss the state of the art (Publication I) and empirical investigation (Publications II–IV) of my inquiries and provide a concise overview regarding their ob-jectives, findings and contributions and role in the dissertation (see Table 5).

4.1