Deductive, Inductive, Abductive: Inference strategies and argument construction in three management research traditions
Its been a while since I've written here. Its been a summer of ethnographic fieldwork and much of my "writing energy" has been consumed in memo-ing. However, I happened to find an old essay I wrote for my comprehensive exams around about a year ago, and wonder of wonders - I don't hate it. So here it is.
First, the question in response to which this essay was written: What reasoning strategies exist for advancing from empirical generalizations to theoretical arguments?
When researchers advance from empirical generalizations to theoretical arguments, they are in effect making two moves: making inferences or advancing from empirical generalizations to theoretical claims, and constructing theoretical arguments or rhetorically persuading an audience that said theoretical claims are valid (Ketokivi & Mantere, 2010; 2021). These moves are not strictly sequential- arguments are not necessarily constructed before or after the inferences have been made, although in some research traditions, certain components of the argument may need to be constructed before making any inferences from the data, or even before collecting the data. Research traditions in organizational scholarship, like theory testing research, inductive case research, and interpretive research are at times conflated with forms of reasoning that are employed to make inferences that link empirical data and theoretical claims, namely deductive, inductive, and abductive reasoning (Mantere & Ketokivi, 2013). While the nomenclature may suggest that deductive, inductive, and abductive reasoning are employed respectively in theory testing, inductive case, and interpretive research, multiple types of reasoning can and do feature in each of these research traditions, and can be used strategically and intentionally by researchers at various points in the process of constructing theoretical arguments from empirical generalizations. Here, I will discuss how, in general, the research tradition within which a researcher operates informs the principal (though often not the only) form of reasoning employed when making inferences that link empirical generalizations and theoretical claims, and how the forms of reasoning employed in constructing theoretical arguments apply more broadly across all traditions of organizational research.
Theory testing research involves generating predicted observations or empirical generalizations from pre-existing theory, and then testing these predictions against empirical data to confirm or disconfirm the theory (e.g., Deephouse, 1996). A principal form of reasoning in this tradition of research is “deduction” or inference to a case, where a particular case or observation is inferred (or predicted) based on some rules and explanations derived from pre-existing theory. For example, from the pre-existing theoretical claim that institutionalized organizations become isomorphic relative to other organizations in the field in response to challenges to their legitimacy (ibid.), and the explanatory prior that academic journals may be institutionalized organizations, I inferred or predicted that an academic journal facing a legitimacy challenge will display some form of isomorphic response. This tradition of research and reasoning in organizational scholarship derives from the physical sciences and inherits from the latter the claim that the inferred cases follow logically from the premises of the rules and explanations from theory, which suggests that the same inferences would be made from the same data regardless of the identity of the researcher. However, even in the case of theory testing research, researchers have to interpret theories and theoretical concepts, and translate these theoretical concepts (e.g.; isomorphism in organizational form in institutionalized fields) to empirical concepts (e.g.; some relevant measure of isomorphism in the case of academic journals). The form of reasoning underlying this translating process is abductive, where the “best” explanation (or a mapping of an empirical and theoretical concept) is inferred based on the researcher’s recognition of structural similarities between an observation (or empirical generalization) and one of several possible rules (or theories).
Inductive case research involves the construction of new theoretical claims from empirical data (e.g. Eisenhardt, 1989), often using some form of grounded theory (Glaser & Strauss, 1967). Inductive case research and theory testing research are related in that theoretical claims (often in the form of propositions about relationships between theoretical concepts) generated by inductive case research are often taken as the starting point for theory testing research. A principal form of reasoning in this tradition of research is “induction” or inference to a rule, where some rule (or theoretical claim) is inferred from multiple empirical observations often in conjunction with an explanatory prior. For example, I inferred from multiple observations of hospitals being compelled to take in patients from and share resources with other hospitals during the COVID-19 pandemic and the explanatory prior that healthcare systems are loose-coupled, that erstwhile loose-coupled systems become tight-coupled (where one organization has to address the needs of other organizations in the system) during system-wide crises. Within the inductive case research tradition in organizational scholarship, an eliminative form of induction is often used, where propositions of increasing generality are inferred in a step-wise fashion from the similarities and differences between observations (Ketokivi & Mantere, 2010). A common conceit of inductive case research is that the theoretical propositions generated by such methods are arrived upon “computationally,” independent of the identity of the researcher (ibid.), which constitutes a further similarity between inductive case and theory testing research. However, this computational or researcher-invariant conceit of inductive case research implies that both the similarities and differences between observations, and the theoretical claims inferred from these similarities and differences are somehow “self-evident” and not contingent on the researcher’s understanding of a phenomenon. For near-universally known and understood phenomena, such inferences may be justified as self-evident. In most cases, however, even the process of eliminative induction at the heart of inductive case research often will involve abductive reasoning both to map empirical and theoretical concepts, as well as in the choice of the central theoretical concepts emerging from the inductive process, concepts that will eventually feature in the theoretical claims stemming from the research (Corbin & Strauss, 1990; Suddaby, 2006; Richardson & Kramer, 2006).
In the interpretive research tradition, researchers engage in a dialogical process between theory and empirical grounds. While the inductive case and interpretive traditions both involve the generation of new theory from empirical grounds, they differ in their claimed stance on the researcher-invariance of inferences, as well as their perspective on what theory is. In the interpretive tradition, theory often takes the form of a narrative that reflects the researcher’s cognition, choices, and the tendencies and traditions of the scholarly community to which she belongs (e.g.; Whiteman & Cooper, 2000). A principal form of reasoning within this tradition is “abduction” or inference to an explanation, where some “best” explanation is inferred if it accounts for an empirical observation in light of a rule or theory. For example, I inferred the explanation that journal editors-in-chief and manuscript authors constitute temporary teams in knowledge producing organizations because it best (in my estimation) accounted for the empirical observations that the research output of journals varied less within the tenure of journal editors-in-chief than across tenures, in light of the rule that leaders of knowledge producing organizations can greatly influence the form and content of knowledge produced by their organizations even when they are not directly implicated in the knowledge-creating process (Von Krogh, Nonaka & Rechsteiner, 2012). Interpretation and subjectivity pervade this research tradition. Researchers in this tradition interpret and construct “libraries” of theories and theoretical concepts within which they search for an interpreted “fit” to the empirical observation, a process often triggered by a “surprising” observation, which ipso facto is also interpretive and subjective, given that what is surprising for some may be ordinary for others. However, interpretive research also often involves the use of analogies (Mantere & Ketokivi, 2013; Ketokivi, Mantere & Cornelissen, 2017). This process features inductive reasoning, though not the eliminative form of induction that characterizes inductive case research. Here, the induction is in the form of analogical reasoning, where apparent similarities between the source and target of an analogy may be the basis for inferring further similarities between source and target.
Empirical grounds are necessary but often not sufficient for making theoretical claims. Across all these research traditions, in order construct theoretical arguments in support of the knowledge claims made by the research, researchers need to deploy warrants or licenses that justify these claims (Ketokivi & Mantere, 2021). These warrants may be procedural, where methodological formalism is deployed to justify some data collection or analysis method. These warrants may also be inferential, where causality is inferred from statistical correlation based on established statistical theory and precedent. Theoretical warrants may be deployed to justify the aforementioned mapping between empirical and theoretical concepts. In many of these cases, the warrants deployed in the argument are arrived at strategically and consciously, involving practical choices made by the researcher in selecting one warrant over the other, depending often on the intended audience (e.g. scholarly community or journal) for their arguments. The choice of these warrants too, may be thought of as the result of the process of abductive reasoning.
Stephen Toulmin's Structure of Arguments: From Ketokivi and Mantere, 2021
Ultimately, studies and the arguments they feature need to be presented to an audience. Often, this takes the form of manuscripts and conference talks. Across all traditions in organizational research, certain features of these presentations have come to be taken-for-granted as necessary, like introductions that motivate the study, theoretical foundations on which the study rests, a summary of results or findings, and some discussion of the findings and theoretical implications. Deductive reasoning may be deployed to motivate the research. Often, this is in the form of inference to a case (e.g. the existence of a gap) from rules (e.g. established theory) and explanatory priors (e.g. the particular phenomena or concepts already covered by the theory). Abductive reasoning may be used to select and identify the “best” theoretical foundation for the study based both on the empirical grounds and the intended scholarly audience. Deductive reasoning may also be deployed in the presentation of findings, often in the form of deductive chains with some general principle at one end, and a specific case or piece of empirical evidence at the other, linked by some connective explanation. Finally, in the discussion, abductive and inductive reasoning may be used to infer theoretical claims based on the empirical generalizations covered in the results and findings.
In summary, researchers may deploy a plurality of reasoning strategies in their efforts to build theoretical arguments from empirical generalizations. These strategies reflect the tradition of research within with the researcher operates, as well as the components of the theoretical arguments and research presentations they are constructing.
Corbin, J. M., & Strauss, A. (1990). Grounded theory research: Procedures, canons, and evaluative criteria. Qualitative sociology, 13(1), 3-21.
Deephouse, D. L. (1996). Does isomorphism legitimate?. Academy of management journal, 39(4), 1024-1039.
Eisenhardt, K. M. (1989). Making fast strategic decisions in high-velocity environments. Academy of Management journal, 32(3), 543-576.
Glaser, B. G., & Strauss, A. L. 1967. The discovery of grounded theory: Strategies for qualitative research. Hawthorne, NY: Aldine de Gruyter
Ketokivi, M., & Mantere, S. (2010). Two strategies for inductive reasoning in organizational research. Academy of management review, 35(2), 315-333.
Ketokivi, M., & Mantere, S. (2021). What warrants our claims? A methodological evaluation of argument structure. Journal of Operations Management, 67(6), 755-776.
Ketokivi, M., Mantere, S., & Cornelissen, J. (2017). Reasoning by analogy and the progress of theory. Academy of Management Review, 42(4).
Mantere, S., & Ketokivi, M. (2013). Reasoning in organization science. Academy of management review, 38(1), 70-89.
Richardson, R., & Kramer, E. H. (2006). Abduction as the type of inference that characterizes the development of a grounded theory. Qualitative Research, 6(4), 497-513.
Suddaby, R. (2006). From the editors: What grounded theory is not. Academy of Management Journal, 49(4), 633-642.
Whiteman, G., & Cooper, W. H. (2000). Ecological embeddedness. Academy of Management Journal, 43(6), 1265-1282.