7. Quantitative section research design: studying venture emergence in Technology
7.1. Method options and selection
There has been an open call to change the focus in entrepreneurship research, from the focus on the entrepreneur characteristics, studying traits and factors that would make an individual an entrepreneur, to study entrepreneurship as a process (Moroz & Hindle 2012). Such change in study focus has also introduced a change in the methods used for entrepreneurship research, therefore we have seen the start of a shift from static tests to longitudinal studies, experiments and models that would capture additional information on the “process” nature of the phenomenon (Eckhardt & Shane 2003). The nature of the objectives of this research are aligned with the necessity to capture data on the dynamic nature of this process and its outcomes.
This research proposes to follow a hypothetic-deductive approach, building from the literature on marketing and technology innovation we have built a set of hypotheses to be tested with observed data. Therefore, the top-down deductive approximation implies that we are building on existing theories in other fields that can be helpful to explain a new phenomenon (Zahra 2007). The adoption of this approach implies that we aim to explain the phenomenon and its characteristics through the lenses of theoretical developments that have been raised and developed in this and other research fields. Thus, special attention has been given to the
justification of the hypotheses and the description of the theoretical framework, as it is recommended in this type of research approximations (Zahra 2007).
Therefore we combine an interest in studying the process that the entrepreneur follows in the early stages of the NTBF development, with a quantitative approach to respond to the hypothetic-deductive research questions we are proposing to contribute in the theory development of NTBF’s venture emergence. As previously described, the entrepreneurship process has a complex nature (Moroz & Hindle 2012); in fact, this complexity is probably part of the essential differences of entrepreneurship when we compare it to other phenomena in the management area. As suggested by Delmar & Johnson (2015), there are five dimensions that influence on the requirements of an adequate research design to study our phenomenon:
(1) specific different characteristics and traits of the entrepreneur, (2) individual dynamics of engagement and disengagement in the entrepreneurial project or new venture, (3) influence of the context where the process is happening, (4) nexus between the individual and the opportunity, including the changes in perceptions and co-evolutionary dynamics, (5) the skewness and kurtosis in the distribution of outcomes, making outliers an important part of the study of the process.
As a result, we consider a longitudinal study research design to observe entrepreneurial behavior and closely follow the within-subject changes (for example changes in their technology resources), and the between-subject differences (as individual or firm different characteristics). Overall this research design provides the necessary tools to advance in our research objectives.
7.1.1. Longitudinal data to study entrepreneurship
The development of a longitudinal research design is often associated with lengthy and expensive research projects. The possibility of observing the early activities in the development of a new venture is a rather challenging task for a researcher: it requires to be able to monitor the variables under study and be able to keep track of changes or modifications occurring during the time of observation. In addition, the outcomes of some of the processes and activities might be delayed in time, making it very difficult to extend findings beyond the case or cases being studied. The subtle and often informal nature of many of the early organizing activities leaves them out of the official statistics (Tornikoski 2007), thus some common databases used for socio-economic analyses on established companies do not provide the needed data to track growth and evolution of new venture emergence.
In the field of entrepreneurship research there has been a breakthrough contribution on longitudinal research with the development of major initiatives to build panel data surveys.
The Panel Studies of Entrepreneurial Dynamics (PSED) research initiatives (Reynolds 2006) and its international evolutions (for example the Norwegian and Swedish versions), as well as the follow-up - PSED II - in the USA (see Table 13) have offered a new territory to test hypothesis and advance entrepreneurship research (Gartner & Shaver 2012).
Prior to the development of panel data for entrepreneurship research, we would have to rely on data from known entrepreneurs, requesting to remember their initial experiences and activities to build a process view of how their venture emerged. This approach suffered from different bias sources as described by Gartner & Shaver (2012): “survival” bias as more often than not nascent entrepreneurs fail to create an organization that gets in the public records ;
“hindsight” bias or the “tendency to distort the initial probability of an event when the outcome becomes known” (Gartner & Shaver 2012, p.2) affecting the estimation of probabilities as the entrepreneur looks back in time; finally the “social desirability” and “social influence” biases mixed with memory decay that produce adjustments to the representation of the actual events to fit with the perceived expectations of the interviewer and social stereotypes (Gartner & Shaver 2012).
Therefore the selection of panel studies data fits the goals of this research as it offers the possibility to have a research design where we observe both the changes in independent and dependent variables with a time separation; this permits the development and test of causality hypotheses (Davidsson & Gordon 2011). This type of research designs also guarantees a closer view on the dynamics of the NTBF’s venture emergence as we have repeated measures of data across time, something that we would not be able to have with a cross-sectional analysis (Delmar & Johnson 2015) where we would be limited to specific data intakes in different moments in time. Unless there is a systematic data collection, we cannot observe the dynamics of the firm in relation to the time dimension.
7.1.2. Panel Data Surveys in Entrepreneurship
The Panel Studies of Entrepreneurial Dynamics (PSED) has long been the reference source for longitudinal data in new business formation (Reynolds & Curtin 2008), and has firm-level data related to theories used in entrepreneurship. The PSED data set had the cases screening in 1999-2000 and the second PSED data set had the cases screening in 2005-2006, this second data set collected data in six waves, closing the file book in 2011.
The interest in further data collection at an international level, but with a similar method and data structure, resulted in the effective internationalization of the PSED across different countries. The different initiatives to build longitudinal datasets, also shared the theoretical understanding of the entrepreneurship process (Reynolds & Curtin 2011; Reynolds & Curtin 2008); consequently, the type of questions and overall design of the surveys is done with this activities and sequences in mind (see Figure 8).
Figure 8. Theoretical perspective of the PSED research design (Reynolds & Curtin (2008))
In addition to the PSED, the Kauffman Firm Survey (KFS) was also launched in 2004 (Robb &
Reynolds 2009) to capture additional data on financing and innovation activities, and introduced a different sampling strategy, taking as a reference the new ventures created in 2004 and listed by Dun & Bradstreet (D&B) as reference population. More recently, new entrepreneurship data panels have started to be collected in Australia under the code name CAUSEE project, or the new China PSED dataset currently under development (see Table 13).
Although most of the longitudinal datasets (including PSED, the different PSED in different regions, and the KFS), shared similarities in the research design and also had common motivations and interests in better understanding entrepreneurship. The results were not always in line with the expectations, the difficulties on sustaining the data collection effort, the data revision and response quality assurance, as well as the needed means to make it open to the academic research community, have favored that most of the research work has focused in the data captured by a few of all the datasets available (Davidsson & Gordon 2011).