1. RESEARCHMETHODS IN ENTREPRENEURSHIP By Mahdikazemi Presented to the Faculty of Entrepreneurship University of Tehran Under the Supervision of Professor A.Arabiun, PhD, Arabiun@ut.ac.ir 2010 Tehran, Iran
2. This slides is an abstract for chapter one (Pages 262-247) from: Handbook of Qualitative Research Methodsin EntrepreneurshipEdited by HelleNeergaard(Associate Professor of Entrepreneurship and Small Business Management, the Aarhus School of Business, Denmark) John ParmUlhøi(Professor in Organization and Management Theory, the Aarhus School of Business, Denmark)Published By Edward Elgar Publishing Limited
5. Unique or typical case characteristics Intensity sampling involves the same logic as extreme sampling, but with less focus on the extreme (Patton 1990).Whereas extreme cases may distort the manifestation of a particular phenomenon, intensity sampling seeks examples that are remarkable in some way.
6. Unique or typical case characteristics A critical case is identified by asking the question: ‘if it happens here, it can happen anywhere’ or oppositely ‘if it does not happen here, it is not going to happen anywhere’ (Patton 1990). A critical case is often used to challenge preconceived ideas or existing theory (Yin 1989). In entrepreneurship, a critical case could be used to test the applicability of, for example, personality trait theory.
7. Unique or typical case characteristics A variation of the critical case strategy involves the choice of politically sensitive or important cases (Patton 1990). The choice of a political case can be directed by the wish to attract political attention to a certain aspect of entrepreneurship, for example micro/peer lending programmes as described in Kibria et al. (2003). In this study the authors conduct in-depth interviews with 17 members of the nonprofit organization Working Capital.
8. Small or great variation across cases According to Patton (1990), the advantage of using a maximum variation strategy is that ‘any common patterns that emerge from great variation are of particular interest and value in capturing the core experiences and central, shared aspects’ (ibid.: 172).
9. Small or great variation across cases Stratification is a ‘subspecies’ of maximum variation sampling and is often used in conjunction with typical case sampling (Patton 1990). Stratified sampling aims at identifying cases that are, for example, above average, average and below average. Contrary to maximum variation, stratified sampling aims at catching major differences rather than commonalities, which means that the focus is on detecting the contingent premises. Each stratum constitutes a relatively homogeneous section of cases .
10. Small or great variation across cases Matched cases mean that a researcher compares pairs of cases, i.e. two and two, three and three etc. A matched pair would, for example, be found within the same industry and then differentiated on various analytical variables. However, it is not possible to construct a matched pair by choosing
11. Small or great variation across cases Homogeneous cases constitute the direct opposite to maximum variation (Patton 1990). This strategy is predominantly used to describe some particular subgroup and/or topic in depth
12. Reference-based selection It may be difficult to distinguish expert from key informant selection, because basically key informants may also be experts. However, as a rule of thumb experts are generally used in the beginning of a project case, which are useful for documenting uniqueness, as well as important shared patterns that derive their significance from having emerged out of heterogeneity (Patton 1990).
13. Reference-based selection Stratification is a ‘subspecies’ of maximum variation sampling and is often used in conjunction with typical case sampling (Patton 1990). Stratified sampling aims at identifying cases that are, for example, above average, average and below average. Contrary to maximum variation, stratified sampling aims at catching major differences rather than commonalities, which means that the focus is on detecting the contingent premises.
14. Reference-based selection Matched cases mean that a researcher compares pairs of cases, i.e. two and two, three and three etc. A matched pair would, for example, be found within the same industry and then differentiated on various analytical variables. However, it is not possible to construct a matched pair by choosing
15. Reference-based selection Homogeneous cases constitute the direct opposite to maximum variation (Patton 1990). This strategy is predominantly used to describe some particular subgroup and/or topic in depth,
16. Reference-based selection It may be difficult to distinguish expert from key informant selection, because basically key informants may also be experts. However, as a rule of thumb experts are generally used in the beginning of a project to help define the boundaries of an investigation or provide ideas about what cases may be chosen (starting off the snowball or chain selection) or at the end of the project to provide insight and information, and supplement the results obtained
17. Reference-based selection Key informants, on the other hand, are ‘ideal’ informants in a study. Most researchers have at some point in their career come across an informant who was very reticent and reserved. According to Tremblay (1957), key informants may be chosen according to criteria such as willingness and ability to communicate or cooperate, and impartiality;
18. Specific theoretical basis The logic of criterion sampling is to study cases based on a predefined criterion of importance (Patton 1990). Their ability to elucidate major systems or programme weaknesses based on which these may be improved
19. Specific theoretical basis Theory-based (selective) sampling is a more formal version of criterion sampling. The researcher identifies specific theoretical constructs and selects cases or informants, even time periods, on the basis of their potential to manifest or elucidate the chosen constructs. Such sampling is typical of theory-driven research (Johnson 1990; Sandelowski et al. 1992).
20. Event-based selection Using critical incidents or events can both be a sampling strategy and an interviewing technique (Flanagan 1954).When used as a sampling strategy it is important to theoretically define a priori what constitutes an event; for example a critical event in a study on how growth is influenced by the provision of external capital could be constituted by the number of capital injections received. When used as a technique either (a) the informants are asked to define what they perceive as critical incidents, or (b) the researcher identifies critical events
21. Sequential cases Snowball or chain sampling aims at identifying cases that are rich in information about a particular subject. This strategy is often combined with reference-based selection in which the first informants are experts in the particular topic under investigation, and who therefore have the knowledge to point the researcher in the direction of which case/s or informant/s may constitute exemplars of the subject of interest (Patton 1990).
22. Sequential cases Confirming/disconfirming case selection is a strategy often used in iterative theory-building, theory-testing designs or data-driven research. Confirming cases are examples which fit into the already emerging pattern; they enrich the study, give it greater depth, and increase the trustworthiness of the study. Disconfirming cases are examples that do not fit the emerging pattern. Hence they constitute a source of rival explanations of the patterns, whence they delimit the explanatory power of the emerging theory.
23. Sequential cases Contrary to most of the other types of sampling discussed here that depend on some prior knowledge of the setting to be investigated, ‘opportunistic sampling takes advantage of whatever unfolds as it unfolds’ (Patton 1990). Field research often involves making on-thespot decisions about sampling either new cases or new informants, and one of the primary advantages of qualitative research over quantitative research is the opportunity to follow unexpected new leads.
24. Sequential cases Linked cases may involve, for example, three generations of business owners, in which the process from establishment to current date needs to be investigated.
25. Random choice Convenience sampling entails that the researcher chooses cases in the proximity of the university or cases that it is easy to obtain access to because an organization member is related or other such reasons. This strategy, however, is neither purposeful nor strategic, and leads to bias! The only exceptions when convenience sampling is acceptable are (a) if the population is very homogeneous or (b) if informal, social networks constitute the only means for identifying and selecting cases or informants,
26. Random choice Patton (1990) does not actually reject random purpose sampling, arguing that for many audiences such sampling substantially increases the credibility of the results. However, as the results will neither be representative nor generalizable, it totally defeats the objective and the argument seems to fall apart.
27. Single-case versus multi-case study strategy Different kinds of sampling require different minimum sample sizes. The choice depends on the purpose of the study. A single, isolated case study often uses the critical case sampling strategy to test, challenge or extend existing theory.
28. Single-case versus multi-case study strategy the single case study may also be constituted by an extreme or unique case which reveal a rare phenomenon. The danger of an isolated case study strategy is that the case may turn out not to be quite as critical, extreme or unique as first assumed. Further, even a single case may also be ‘linked’.
30. isolated and multiple cases Whilst isolated cases help investigate a specific problem in depth, a multiple case design helps strengthen confidence in the precision, validity and stability of the result because repetition of the same procedure across a number of cases helps eliminate accidental similarities between theory and case.
31. single cases and multi-case Both single cases and multi-case studies may be embedded (Yin 1989) or nested (Miles and Huberman 1994). This type of case study is a bit like a babushka nesting doll. It contains several ‘units’ of analysis within the same case and requires several steps of selection. First the main case is selected. This is followed by a selection of the various units to be studied within the case, for example various management levels. Finally, the informants at each level need to be chosen
32. single cases and multi-case It is quite common not to use a single case sampling strategy in isolation. Often several strategies are combined in the same study in order to reach the best choice of cases; for example criterion sampling may be combined with maximum variation sampling in order to increase the robustness of the findings. Reference-based sampling is probably the strategy which may be used in conjunction with most other strategies.
33. Negotiating access When a sampling strategy has been decided upon, the researcher encounters the more practical problem of negotiating access to particular cases. Purposeful sampling means just that: the researcher chooses precisely those cases that will yield the most useful information for that particular inquiry given the research question(s).
34. Choosing the ‘right’ strategy It is, however, not always easy to determine which sampling strategy to use. Some researchers suggest that certain types of sampling and sample sizes are favoured in certain types of qualitative research (Morse 1994; Cresswell 1998), others that the purpose of the inquiry directs the choice of sampling strategy (Patton 1990; Maxwell 1996). This tends to be a very personal choice.
35. Choosing the ‘right’ strategy Open sampling takes place in the early stages of an inquiry and is defined as ‘sampling open to those persons, places and situations that will provide the greatest opportunity to gather the most relevant data about the phenomenon (ibid.: 181). In essence this is what Patton (1990) would call opportunistic sampling because the researcher is supposed to follow unexpected new leads in order to lead to theory generation
36. Choosing the ‘right’ strategy the latter is a strategy that should be avoided at all costs, because it does not necessarily lead to the best information; in other words it fails to fulfil the quality criteria of appropriateness. open sampling involves ‘going from one person or place to another on a list’ (Strauss and Corbin 1990: 184),which constitutes random purposeful sampling in Patton’s terminology.
37. Choosing the ‘right’ strategy the basic assumption behind qualitative research makes random sampling inappropriate and the worst choice. It indicates a wish to generalize from sample to population, which all the sources agree is neither possible nor desirable in qualitative research. Qualitative research does not aim to ensure representativeness, but rather the field under study yields substantive information that will contribute to elucidate the problem issue, and on this basis facilitate ideographic, holographic, naturalistic or analytical generalization
38. Choosing the ‘right’ strategy Relational and variational sampling take place in the process of looking for evidence of variation or differences within and across the data and test how well the emerging theory holds up in new settings (Strauss and Corbin 1990: 185ff.). These purposes can be fulfilled through the selection of, for example, confirming or disconfirming cases, using matched or maximum variation sampling.
39. Generalizability Whether few or many cases are included in a sample, this is an issue that cannot be avoided since qualitative studies are often rejected by reviewers as they disbelieve in the value of small purposeful samples simply because these cannot be generalized to ‘a larger universe’. It is an inherent feature of qualitative studies that they are context dependent and not representative of a larger universe, neither do they allow generalization across time and space.
40. Analogous generalization Analogous generalization is concerned with extrapolation of an insight from the situation researched to recognizing this insight in new and foreign contexts, or with identifying analogous situations. It means that the researcher thinks about the likely application of the findings to other situations under similar, but not identical, conditions (Patton 1990). Extrapolation is not exclusively made by the researcher, but may also be undertaken by the reader who recognizes the situations in question.
41. analytical/theoretical generalization analytical/theoretical generalization, which finds its application when the researcher operates within a theoretical framework to which findings can be generalized (as e.g. proposed in Yin 1989). This type of argumentation may also be described as abductive reasoning (Danermark et al. 2003), which aims at hypothesis or theory generation.
42. Conclusion As the above highlights, there is more to qualitative sampling than meets the eye. It is an elaborate process of making the ‘right’ choices. Careful sampling pays attention to what can be controlled in terms of characteristics of events, cases and informants as well as to what cannot be controlled. Purposeful sampling may not solve the problems of selection bias but it reveals the selection criteria which reduces the vulnerability to criticism for not being sufficiently rigorous .
43. Conclusion in a publication perspective, it is extremely important to argue each and every step in the selection process convincingly and in detail. The usefulness of qualitative research is often judged on the basis of the logic and the purposes that are associated with probability sampling. Therefore, as the review of mainstream, peerreviewed journals showed, it is decisive that the methodology is sufficiently succinct, well argued and accounted for.
44. Conclusion Looking at the long tradition in quantitative research for common rules and procedures and at the more or less standard format of accounting for sampling choices and analytical procedures, researchers using qualitative methods should not be surprised that in a publication perspective qualitative research is lagging behind. Qualitative scholars should provide a thorough account of their sampling strategy and analytical procedure, demonstrating that qualitative research can be as methodologically rigorous as its quantitative counterpart. If they do this, experience shows that it will result in successful publication, even in journals that are known to be inclined towards quantitative research .