Methodological Implications of Critical Realism for Mixed-Methods Research

ions are a crucial part of the retroductive analysis because they allow us to strip out and separate the constitutive structures and necessary (internal) properties of the objects we are studying from the contingent (external) ones that affect the events we observe (Sayer 1992; Tsoukas 1989). Within this process, mixed-methods played a key role in identifying generative mechanisms (revealing why things are as they appear) and forming robust meta-inferences that can guide our thinking in different research contexts. In our case, understanding the components and structure (as a set of internally related objects) of SWIFT (e.g., its technological capabilities and network infrastructure, message standards, community and governance, etc.) helped us to theorize about what causal powers and liabilities SWIFT might hold. For example, the ability of SWIFT to transmit payment messages quickly and securely, or the availability of knowledge and expertise from the SWIFT community, provides key information about its potential impact on organizations. On the other hand, concrete research helped us understand what happened when these powers and liabilities were exercised under certain contingent conditions. In our case, these were recovered by employing qualitative methods, including a historical study, alongside econometric analyses to understand the differences between adopters in the early and late years and the different characteristics of financial institutions and the banking sector in the 1970s and the 21 century. This use of quantitative and qualitative methods was vital in order to identify both internal (necessary) and external (contingent) relations. In the early stage of our research interviews, historical narratives, and quantitative analysis uncovered constituents of SWIFT such as the size of its network, its technological capabilities, governance structure, etc., while, in the later phases, further interviews, archival work, quantitative analysis, and case study enquiry provided information about the contextual factors such as the types of SWIFT adopters, and regional and temporal influences. The interplay between methods not only helped to ensure completeness and complementarity in our study but also advanced our research strategy by enabling us to build on the results from different methods and develop a systematic process of inquiry through retroduction. Finally, different methods were used to compensate for the limitations of the other methods and thus provide diversity by looking at different levels of abstraction as shown in Table 3. The identification of causal mechanisms through abstraction and concrete research has commonalities with metainferences in MM research as both concepts effectively use quantitative and qualitative methods (Tashakkori and Teddlie 2003; Venkatesh et al. 2013). However, there are some key differences. Meta-inferences are often described as a merge between the quantitative and qualitative inferences once a research enquiry has come to an end. However, our study highlights the identification of generative mechanisms as an iterative and nonlinear retroductive process where quantitative and qualitative methods work in conjunction and feed into each other until a robust mechanism that can explain the phenomena is recovered. Further, the main route of developing meta-inferences is principally an inductive one (Venkatesh et al. 2013), while the identification of causal mechanisms is driven by a creative retroductive process (Mingers 2004, 2005). Had we used a linear inductive approach to meta-inferences that combined findings from the two different types of research (extensive and intensive), we would have been confronted by conflicting views on the impact of SWIFT on organizations. For example, our initial interviews supported the argument that SWIFT is a neutral plumbing of the banking sector that did not add to the bottom line performance of banks, whereas statistical findings demonstrated a robust correlation between SWIFT adoption and measures of financial performance. The retroductive logic allowed us to go beyond the conventional synthesizing of the results and question the assumptions and conditions (i.e., degree of closure and contingent factors) under which MIS Quarterly Vol. 37 No. X/Forthcoming 2013 21 Zachariadis et al./Methodological Implications of CR for Mixed-Methods Research they were produced in order to uncover the generative mechanisms. Demi-Regularities and Validity Through Closure In our paper we also demonstrate how demi-regularities not only focus the research design but also facilitate the hypothesizing and generalizing about existing causal mechanisms. In addition, we show that they can also help to assess and explain the results in the analysis phase. By acknowledging that social systems are inherently open, and thus observable, events are a result of complex interactions between several causal mechanisms that may offset each other. We highlight that conditions of closure are imperative to establish these demi-regularities and produce valid generalizations and causal arguments. For that reason, we show how intrinsic and extrinsic quasi-closure was important in our study in establishing validity, and in particular providing some possible generalized statements about the real mechanisms in place by merging them with actual events (Tsoukas 1989). For example, intrinsic closure in our study was achieved when we restricted our research to reflect no qualitative variation in SWIFT, that is to say, when the necessary constituents of SWIFT remained largely unchanged during a specific period of time. In parallel, extrinsic closure, which required that the external mechanisms in the environment of SWIFT (e.g., the country or region of use, the timeframe, the type of adopter organizations, etc.) are constant, was achieved when we restricted our sample to a particular set of banks of specific type and size. Once these two conditions were applied we could argue that the demi-regularities observed in terms of event correlations could be a manifestation of a causal mechanism recovered during the interplay between the abstract and concrete research. In our example, we demonstrated how with the use of MMs we managed to narrow our study down to a specific subsample where the two closure conditions could be met. For example, through the interviews we conducted, we realized that certain intrinsic aspects of SWIFT changed over time, so we had to confine our quantitative study to particular periods where there wouldn’t be much qualitative variation of these aspects (intrinsic closure). As Runde and de Rond (2010) note, not only is it necessary to embrace that the questions asked and answers given by people depend on their interest and situation, but more importantly this kind of situatedness, given the impossibility of citing all the causes of an event, is in fact a resource that makes causal explanations feasible at all. The reason for this has to do with our why-questions always being from a point of view, thereby always focusing on one or other aspect of some event.... That people always come to things and ask about them from a particular (subjective or situated) point of view, rather than from what Nagel (1986) calls the “view from nowhere,” provides the necessary restriction on the cases we need to cite in any particular explanation (p. 436). The qualitative methods were boundary-making, helping us to narrow the range of issues in our inquiry, revise the scope of (quasi-)closure at work, as well as forming part of the elimination process by which we moved the study forward. Meanwhile, through our statistical analysis, we uncovered a correlation between smaller retail banks and SWIFT adoption. As a result, we restricted our qualitative study to that domain of small retails banks and we were able to isolate and uncover the mechanisms that lead to increased bank performance without the influence of other objects and mechanisms (i.e., extrinsic closure). Meeting these conditions allowed us to generalize our findings and argue that within that specific domain (that of a small, late-adopter retail bank) the application of SWIFT led to increased performance. This generalization in the form of internal (establishing a causal link between events) and external (generalizing the same mechanism between different banks) validity was achieved through closure and the recovery of robust demi-regularities in the process of our retroductive approach to mixed methods. Theory generation in CR lies on the development of abstractions that hypothesize about and identify generative mechanisms (rather than regularities) that are observed in different contexts and are expressed by means of empirical categories—empirical generalizations (Danermark et al. 2002). In that respect, generalizing means to identify universal structures or mechanisms that are theorized to exist in different contexts but not necessarily cause the same events (Wynn and Williams 2012). Therefore, in our study, the process of generalizing the findings was not focused around replication of the correlation in a different context but rather, through the use of MM, it examined the links between generative mechanisms in different contexts, for example, in other small retail banks. Our study also identifies different levels of generalization. For example, going from a specific case study to a small sample (Seddon and Scheepers 2012), and from there to the wider population, so as to provide explanations in other contexts (Lee and Baskerville 2003; Lee and Hubona 2009; Seddon and Scheepers 2012; Venkatesh et al. 2013). 22 MIS Quarterly Vol. 37 No. X/Forthcoming 2013 Zachariadis et al./Methodological Implications of CR for Mixed-Methods Research