Why should we bother? What are configurations for?

Configurations for capturing patterns among organizational elements

Configurations are about patterns, arrangements of parts, or elements. Miller (2017: 2) in describing the historic approach of configurational theorists says, “The core theses were that it is important to distinguish among different types of organizations in predicting outcomes, that many elements describing an organization cluster in thematic fashion, and that a few thematic types or ‘configurations’ may encompass many organizations” (Meyer et al., 1993; Miller and Mintzberg, 1983). Similarly, Short et al. (2008) emphasize the existence of internally consistent elements that produce differentiated sets of organizations. Meyer et al. (1993) are particularly concerned with distinguishing configurational approaches from contingency theory; the great advantage of such approaches is that they are holistic in nature and emphasize nonlinear, reciprocal relationships between organizational elements.

Indeed, classification has been at the heart of organizational analysis at least since the work of Max Weber (1924, 1947; Hinings and Meyer, 2018). The argument is that there is diversity among organizations but that this diversity is systematic not random, and limited. “Scholars need to be careful not to atomize the essential interconnectedness of organizations” (Meyer et al., 1993: 1181). Typologies such as that of Weber have long been at the heart of organizational classification and theorizing. Often, such typologies are not initially empirically based but are a way of summarizing the key elements of a particular organizational form. As Weber (1963) put it, an ideal type “(…] is not a description of reality but aims to give unambiguous means of expression to such a description […) it is no hypothesis but rather it offers guidance to the construction of a hypothesis” (p. 396). As Hinings and Meyer (2018) put it, “Weber was not expecting to find his ideal types empirically in their pure form nor was he suggesting that all three authority systems could not be found in any particular era or society—an important point that got somewhat lost in later discussions” (p. 23).

In a similar vein, with a similar result, the well-known Burns and Stalker (1961) typology of organic and mechanistic organizations was never meant to be a complete statement. Indeed, as they emphasize, “… the two forms of system represent a polarity, not a dichotomy; there are, as we have tried to show, intermediate stages between the extremities empirically known to us. Also, the relation of one form to the other is elastic, so that a concern with oscillating between relative stability and relative change may also oscillate between the two forms. An (organization) may (and frequently does) operate with a management system which includes both types” (Burns and Stalker, 1961: 122; see also, Hinings et al., 2018). Thus, the ideal type thinking is to enable us to theorize and hypothesize about characteristic configurations without making assumptions about their existence.

However, other organizational typologies are presented as summations of empirical findings, accepting configurational analysis as a parsimonious way of thinking about organizations. Perhaps the best known one is Mintzberg’s (1979) distinctions of Simple Structure, Machine Bureaucracy, Professional Bureaucracy, Divisionalized Form, and Adhocracy. While they retain “the essential elegance and simplicity that is the hallmark of typologies” (Meyer et al., 1993: 1182), they are actually meant to be soundly based summaries of empirical research. Miles and Snow’s (1978) typology of defenders, analyzers, prospectors, and reactors are similarly presented as configurations of the essential elements of empirical findings.

Thus, it would seem to be legitimate to take typologies such as those of Mintzberg and Miles and Snow and subject them to taxonomic methods, asking the question of whether they actually occur in the form suggested (Doty et al., 1993). But that would be less the case for Weberian ideal type theorizing, where establishing that bureaucracies or organic and mechanistic organizations do not exist in the ways specified misses the point of such typologizing. The essential nature of configurational analysis in the debate over methods such as QCA, principal components analysis, and cluster analysis is taxonomic. As Meyer et al. (1993) put it, “the logic of taxonomy… lies in empirical classification based on multivariate analysis of multiple dimensions that may cover structures, processes, strategies, and contexts” (p. 1182). Of course this does not mean “empirical” devoid of theory. An early example of a taxonomic approach is the work of the Aston school (Pugh et al., 1969), where the theoretical starting point was a multidimensional analysis of bureaucracy, not to show that Weber was “wrong,” but to elaborate the basic idea. Ketchen et al. (1993) argue that inductive, atheoretical approaches do not generate consistent configurations.

Essentially, the QCA argument is that previous methods used to produce taxonomies are problematic and that QCA has better prospects. Clearly, sound and appropriate methods are extremely important. But, the whole idea of configurations, and the particular configurations being examined and established have to be based on robust theoretical foundations. While QCA takes us in this direction, there is perhaps, a more general methodological problem within the study of organizations, that the primary methods are multivariate within what Donaldson (2001) calls a Cartesian approach. Donaldson (2001) states that the evidence for stable organizational configurations is poor (e.g. Doty et al., 1993; Gresov and Drazin, 1997). He argues that “real organizations are widely distributed along both the structural and the contingency dimensions. There are many more fits than the few configurations postulated” (Donaldson, 2001: 152). His Cartesian approach is that contingency theory is a better approach than configurations because “contingencies and structures form a multidimensional framework within which fits are continuous” (Donaldson, 2001: 152). Interestingly, however, Donaldson does see the worth of ideal types as mental models that have an orienting role for the development of theory which is a somewhat Weberian view.

Van de Ven et al. (2013) on the other hand, argue that a configurational perspective on organizational design and strategy is an important addition to theorizing about organizations. In particular, as Fiss (2007) and Miller (2017) also suggest, the configuration perspective is important in dealing with equifinality in organizational design, context, and performance. Research and theorizing on equifinality is still at an embryonic stage (Van de Ven et al., 2013), but configurational analysis is particularly promising for this purpose.

I have suggested that one of the reasons why most configurational research has been about organizational forms and strategies is that from the early days of organizational analysis, there were well-articulated theories of a configurational kind (Miles and Snow, 1978; Weber, 1949). The resulting work in the 1960s and 1970s inspired by Weber’s ideas on bureaucracy was centered on both multivariate and taxonomic approaches to uncovering empirically based types of organizations (Hinings and Meyer, 2018). As a result, there has been considerable agreement about the elements of structure, strategy, and context that comprise an organizational or strategic configuration. But it has recently been argued that there has been a decline in interest in the organization per se and a rise in interest in a myriad of topics inside and outside organizations (Hinings et al., 2014; Hinings and Meyer, 2018). Perhaps it is the expansion in theories and topics within organization theory, strategy, and organizational behavior that has led to the idea that configurational analysis as a method and a tool can be extended to many, many more theoretical areas than simply organizations and strategy.

Configurations as a generic classification tool

There is a further argument about configurations as generic building blocks for theory. Essentially, this is based on the theoretical position that classification is essential to all analysis (McKelvey, 1982). McKelvey (1982) and Ulrich and McKelvey (1990) essentially argue from a biological, taxonomic perspective that establishing populations of organizations is central to a successful organization science. It is, they argue, a sounder approach to move from findings on one set of configurationally established organizations to another to establish generalization. The starting point for generalizing is systematically generated organizational difference. McKelvey (1982) argued that homogeneous groups of organizations had to be discovered, their characteristics identified, and then comparison could be made with other different, homogeneous groups. There is a need to simultaneously understand similarity and differences and, at any level, this is best done through configurational theories and methods.

The thrust of the generic argument is thus that in examining anything to do with organizations, for example, culture, logics, performance, and communication, configurational approaches are at the heart of comparison. While Greckhamer et al. (2018) primarily refer to configurational approaches for organizations and strategy, they suggest that the approach is scale-free. Indeed, it is clear from the discussion of QCA that they see it as applicable to a wide range of phenomena. Täuscher (2017) more directly suggests that QCA has the potential to advance research on business models and identifies three different interpretations of business models as (1) attributes of firms, (2) as cognitive and linguistic schemas, and (3) as formal representations of how an organization functions. Clearly, this represents the idea and use of configurational analysis (and QCA) as generic. Short et al. (2008) argue directly for an expansion of configurational approaches to human resources management, organizational behavior, ethics and corporate social performance, entrepreneurship, and international business.

Meyer et al. (1993) made a number of suggestions for applying configurational approaches at the individual level. They pointed out that there was a considerable amount of typological thinking in areas such as personality, personality-organizational fit (Chatman, 1989, 1991) motivational theories (McClelland, 1985), and leadership (Kipnis and Schmidt, 1988). The latter, for example, used cluster analysis to establish four general influence styles. Meyer et al. (1993) noted, “we believe that addressing such configurational questions might push research beyond univariate and bivariate analyses toward multivariate analyses of reciprocal and nonlinear relationships” (p. 1189).

Meyer et al. (1993) also argued for the application of configurational analysis at the group level. Using the work of socio-technical systems (Emery and Trist, 1969), they suggested that configurations of task, social, authority, and reward systems could be fruitfully studied. They also suggested that “traces of configurational thinking can also be found in the multiple constituency model of work unit effectiveness (Tsui, 1990; Tsui and Milkovich, 1987)” (Meyer et al., 1993: 1190). They ended their article with a call that represents what all that those who see configurational analysis as a generic approach centered on patterns in social phenomena believe in:

Our goal has been to convince AMJ readers that the configurational approach represents an exciting intellectual direction with the potential for revitalizing theory and research about individuals, groups, and organizations. Our optimism is fueled by the conviction that although the possible combinations of individual and group attributes may be infinite, only a finite number of coherent configurations are prevalent in the social world. (Meyer et al., 1993: 1192)

At the heart of these extensions is the general idea of patterning in phenomena of all kinds, together with the idea that such patternings are important in explaining a wide range of phenomena. Indeed it is this idea that is the lynchpin of configurational approaches. First is the idea that there is coherence between phenomena of all kinds. It is that idea of coherence that can produce configurations. Second is the theoretical position that such coherent configurations are important because classification is central to theorizing for all organizational, sociological, and psychological phenomena. But it is probably necessary for those rooting for configurational analysis to be clearer about this and to argue more strongly about patterning of elements because the dominant models are multilinear rather than systemic or configurational in both theorizing and methods.

The call for configurational analysis to be expanded to a wider range of theoretical areas (Fiss, 2011; Short et al., 2008) suggests a highly specific set of elements for each configuration and the use of the term “configurational analysis” as a general, guiding principle rather than a particular theoretical approach. And as such, methods become very important as it is left to each theorist to decide what elements to include in the analysis. This is actually quite different from the original approaches in the study of organizational forms and strategic approaches which aimed to deal with very basic classifications and what were seen as stable defining elements.

One theoretical area where one does not see discussion or use of configurational analysis is institutional theory. Yet theorizing about patterns that are ongoing and stable is central. One attempt has been made theoretically through the idea of archetypes (Greenwood et al., 1990; Greenwood and Hinings, 1988, 1993; Purdy and Grey, 2009). Donaldson writes of this approach approvingly (!) and Short et al. (2008) use these ideas as a part of their framing. The conceptual “innovation” of the concept is in taking the known elements of organizational design, summarized as structure and systems, and adding interpretive schemes (Ranson et al., 1980) as the element which gives meaning to the patterning of structures and systems. The concept of interpretive scheme was a forerunner of the move to institutional logics (Thornton et al., 2012). Institutional logics are patterns of values and beliefs that are strongly related to structural and systemic elements in organizations and fields. Institutional theory uses both the ideal type approach through the idea of logics as being institutionally legitimated but not necessarily present in a complete empirical sense (Thornton et al., 2012), and the taxonomic approach of measuring the extent of the presence of a particular logic in an organizational field (Greenwood et al., 2011). Institutional theory is ripe for a configurational approach at the level of method because the theoretical structure already speaks strongly of the existence of stable patterns. I will return to this later.

Configuration for understanding organizational coherence in specific cases

The ideas of patterning and coherence which are so central to configurational analysis can also be developed at the level of the individual organization. In particular, the argument that organizational coherence produces better performance (Van de Ven et al., 2013) can be applied without necessarily placing an organization in a wider configurational category. Thus, configuration can be thought of as a quality of specific organizations, expressing the degree to which the specified elements of an organization are coherent. Miller (1996), having dealt with typologies and taxonomies as the basic approaches to configuration, suggested a third way: configuration as a quality. “Configuration, in this sense, can be defined as the degree to which an organization’s elements are orchestrated and connected by a single theme. Such themes can be found within or across categories” (Miller, 1996: 509). Thus, within a single organization, coherence or configuration is found because organizational elements are tied together by a primary goal, an all-encompassing culture, a primary function. Through this, Miller argues, competitive advantage is gained because of the way in which all of the thematic elements work together. It becomes a task of the researcher to examine how far individual organizations operate through a configuration, without necessarily making any assumptions about that particular configuration being general and thus present in other organizations.

This approach to configurations as being about coherence or fit between organizational elements in a single case is found in the work of Siggelkow (2001, 2002), in his studies of Liz Claiborne and the Vanguard group. The concept of fit has long been central to contingency theory (Donaldson, 2001; Van de Ven et al., 2013) with the idea of congruence (or consonance, Pfeffer, 1997) between elements. In contingency theory, it is primarily congruence between structural elements of the organization and contingent or environmental elements. However, implicit in this theorizing is the idea of fit between organizational elements themselves. Indeed, Siggelkow (2001) distinguishes between internal and external fit; Liz Claiborne had high internal fit between the elements of its design, marketing, production, and processes, that is, it was a configuration. These aligned strategic and operational choices were also in fit with the environment, such as customer demand, competitors’ strengths and weaknesses, and technology and retailers’ requirements. And the result was strong performance.

Siggelkow’s (2002) case study of the Vanguard Group is also about fit, but more directly utilizes the concept of configuration. Accepting that configuration is associated with higher performance, Siggelkow asks the question of how an organization evolves to an internal fit or configuration. He adds two ideas to the nature of configurations; first, that effective configurations are made up of core elements rather than all elements; second, that configurations are systems. He suggests, “A possible definition of configuration is thus an organizational system with no inconsistent core elements and a number of reinforcing core elements” (Siggelkow, 2002: 128).

Something similar can be seen in approaches to business models (Arend, 2013; Coff et al., 2013), where one approach is to analyze them as typologies and taxonomies (Zott and Amit, 2013) whereas others examine business models as configurations within individual organizations (Baden-Fuller and Mangematin, 2013). At the level of the individual organization the aim is to understand the way in which various elements are patterned, as a result of which positive outcomes are produced. The whole notion of a business model is of a configuration of goals, strategies, structures, and systems.

Täuscher (2017) is, of course, concerned with the application of QCA and Systems Dynamics to business models. The approach through systems dynamics is an individual organizational application. While systems theory and systems dynamics are expressed at a very general level, their application is primarily at the level of the individual entity. While systems theorists might argue that there are generic systems, the way they cohere can be different from organization to organization. The point of the analysis is not primarily to produce an empirically based generalizable set of categories, but to show how system elements work together dynamically in a specific setting. Thus, as with all of the approaches that are reviewed under this third approach, it is related to understanding organizational uniqueness and specificity, which is particularly important in strategy. The issue is one of understanding the nature of fit in specific cases.

The future of configurational thinking

Clearly, Miller, Täuscher, and Greckhamer et al., regardless of their particular stance on method, all believe that there is a strong future for configurational analysis. While the thrust is to put forward a method (or methods in the case of Täuscher), those methods are only valid if configurations are worth pursuing (see also Meyer et al., 1993; Short et al., 2008). As someone who has made a small contribution to configurational thinking, I believe that it certainly does have a future! However, that future has to be considered in the light of two issues. First, which, if any, of the three uses of configurational thinking that I have suggested may be more fruitful? Second, what changes have been and are taking place that can affect approaches to configurational thinking?

Responding to the first issue of which approach to configurations will bear fruit, depends strongly on one’s theoretical stance with regard to the purposes of configurations. In particular, whether one believes that the goal is to find the limited number of configurations that occur (Meyer et al., 1993). This theoretical starting point can be applied either at the level of organizations or to building a classification system for any and all social phenomena. Miller (2017) and Greckhamer et al. (2018) emphasize configurations of strategy and structure but Täuscher (2017) extends the ideas to business models. The basic idea is that patterns exist but the theoretical question is whether our expectation is that such patterns occur in limited and stable numbers and thus over long periods of time. If we take the idea of bureaucracy as an example, studies have shown three things (Hinings and Meyer, 2018). First, the patterning of elements that analysts call bureaucracy has been around for at least 200 years; it has some stability. Second, deeper analysis of individual bureaucracies demonstrates that the dynamics of operation are highly dependent on non-bureaucratic elements and actions. Third, the prevalence of bureaucracy has changed over time and now we have much more discussion of “new organizational forms” (Davis, 2016). In part, it is the configurational analysis of the 1960s and 1970s in particular that has enabled us to establish these changes. The future of configurational thinking may well lie less in theorizing a number of limited, stable patterns, and more in utilizing configurations as ways of understanding organizational change and development.

In part, Davis’ (2016) discussion of new organizational forms relates to my second point, that in thinking about configurational analysis it is necessary to take into account of changes in both theoretical approaches and empirical circumstances. Theoretically, configurational analysis has primarily been about strategy and structure. QCA, and indeed all methods to do with patterning, such as cluster analysis, principal components analysis, or correspondence analysis, are agnostic about theory. Indeed, they can properly claim as, for example, Short et al. (2008) do with regard to QCA, that they can deal with any appropriate data set. I have already referred to institutional theory and I would like to take this argument a little further.

It has been claimed that institutional theory is the dominant approach in organizational theory (Greenwood et al., 2017). If that is the case, or even close to it, then what configurational thinking is to be found there? I have already suggested that the concept of archetypes is configurational. But we can also see an approach centered on patterning in discussions of institutional logics. Thornton and Ocasio (1999) define an institutional logic as “the socially constructed, historical patterns of material practices, assumptions, values, beliefs, and rules by which individuals produce and reproduce their material subsistence, organize time and space, and provide meaning to their social reality” (p. 804). Ocasio et al. (2017) describe this definition as “foundational imagery” (p. 510).

There are two aspects to this definition that are important for configurational thinking. First, the definition is about “patterns” and discovering patterns is the heart of configuration and methods such as QCA. Second, there is a listing of the elements of logics, namely, material practices, assumptions, values, beliefs, and rules. Thus, it is clear what the search for configuration is to be based upon. Raynard (2016) uses the term configurations in taking forward the debate around institutional complexity (Greenwood et al., 2011). Using the three elements of logic incompatibility, unsettled field-level prioritization, and jurisdictional overlap, she produces four configurations. While the aim is to draw together empirical work on logics, Raynard (2016) is really producing a typology. But her approach is ripe for the use of configurational methods. Similarly, Ocasio and Radoynovska (2016), examine institutional pluralism and complexity and their effects on strategic choices and business models. Again, the language lends itself to configuration analysis with terms such as homogeneity and heterogeneity, that is, differing patterns. My point here is to suggest that institutional theory and institutional logics in particular is a ripe area for configurational analysis.

The second issue for the future of configurational thinking relates to actual changes occurring that, in particular, affect the nature of organizations and strategy given the centrality of these topics. In particular, there is the impact of digital innovation (Davis, 2016; Dougherty and Dunne, 2012; Hinings et al., 2018; Leonardi and Vaast, 2015). The arguments here are that digital innovation is producing new organizations that are platform-based, introducing new institutional infrastructure such as blockchain, and affecting existing organizations in major ways. Indeed, the argument is that these changes go beyond innovation to digital transformation. Thus, the decline of existing organizational configurations and their replacement with new configurations is an important topic for research. This relates to configurational approaches being used as building blocks of theories of organizational change as in the work on archetypes (Greenwood and Hinings, 1991). For example, if organizations such as banks or hospitals are subject to digital innovation (as indeed they are), then what “non-digital” configurations are they moving from and how does digital innovation play out as part of a new configuration. In other words, configurational approaches need to be sensitive to the dynamic, changing elements in organizations and society.