Analytical Sociology: A Bungean Appreciation

Abstract

Analytical sociology, an intellectual project that has garnered considerable attention across a variety of disciplines in recent years, aims to explain complex social processes by dissecting them, accentuating their most important constituent parts, and constructing appropriate models to understand the emergence of what is observed. To achieve this goal, analytical sociologists demonstrate an unequivocal focus on the mechanism-based explanation grounded in action theory. In this article I attempt a critical appreciation of analytical sociology from the perspective of Mario Bunge’s philosophical system, which I characterize as emergentist systemism. I submit that while the principles of analytical sociology and those of Bunge’s approach share a lot in common, the latter brings to the fore the ontological status and explanatory importance of supra-individual actors (as concrete systems endowed with emergent causal powers) and macro-social mechanisms (as processes unfolding in and among social systems), and therefore it does not stipulate that every causal explanation of social facts has to include explicit references to individual-level actors and mechanisms. In this sense, Bunge’s approach provides a reasonable middle course between the Scylla of sociological reification and the Charybdis of ontological individualism, and thus serves as an antidote to the untenable “strong program of microfoundations” to which some analytical sociologists are committed.

1 Introduction

As Filippo Barbera (2004: 43) points out, despite the evident diversity of theoretical inclinations, methodological preferences, and the empirical phenomena under study in social science, it is nevertheless possible to discern a shared approach, which can be summarized in the expression “analytical sociology”. Indeed, this intellectual project has garnered considerable attention across a variety of disciplines in recent years, especially since Peter Hedström (2005) published his manifesto. While the principles of analytical sociology and their implications for practical social research have been vigorously discussed and debated in journals, books, seminars and conferences (for a concise overview see Manzo 2010; see also Noguera 2006; Noguera et al. 2009; Barbera 2006; Hedström 2009; Hedström and Bearman 2009; Hedström and Ylikoski 2010a, b; Kron and Grund 2010), even more worthy of note is the founding of the European Network of Analytical Sociologists in 2008, generally recognized as an important step towards the institutionalization of an “analytically” oriented style of doing social science research.

As explained by Hedström (2009: 232), analytical sociology aims to “explain complex social processes by carefully dissecting them, bringing into focus their most important constituent components, and then to construct appropriate models which help us to understand why we observe what we observe” (Hedström 2009: 332). With its commitment to the importance of “microfoundations” in the causal reconstruction of the processes that give rise to given collective phenomena, this intellectual project has made strenuous efforts to bring to the fore the mechanism-based explanation grounded in action theory (Handlungstheorie) and a realist approach to causality. Besides, analytical sociologists share an emphasis on precision in definitions and reduction of ambiguities. Or, in the words of the German sociologist Hartmut Esser (1999: xvii), they endeavor to “represent the sometimes rather complicated details as transparently and understandably as possible”.¹ One important advantage of clarity and exactness is the possibility to accumulate reliable knowledge by avoiding misunderstandings and fruitless controversies (Greshoff 2010: 7).

A moment’s thought about all these characteristics should be enough to reveal that Mario Bunge’s writings have a lot to contribute to this emerging social science paradigm. Three are three interrelated reasons. Firstly, Bunge has attempted to elucidate the nature of (social) action by way of a devastating critique of rational choice theories, including the “linear system of action” formulated by James Coleman himself, who played a crucial pioneering role in the development of analytical sociology,² and it has been made clear that his emergentist systemism³ is best construed as a version of action-systems theory. Secondly, while the majority of analytical sociologists, with their insistence on the need for microfoundations of sociological explanation, are avid practitioners of “structural individualism” (a weak form of methodological individualism), Bunge’s works have been characterized by vociferous criticisms of ontological, epistemological and methodological individualism (e.g., Bunge 2000, 2003: 196–212). However, some have contended that Bunge’s polemic risks setting up a straw man, to the effect that Bunge’s systemism, as exemplified in the Boudon-Coleman diagram, is essentially equivalent to methodological individualism.⁴ In any case, the exact relationship between Bunge’s emergentist systemism and structural or methodological individualism awaits further clarification. Thirdly, one of Mario Bunge’s major contributions to the field of philosophy of (social) science is his explicit focus on the mechanismic (“translucent box”) explanation in relation to a refined systems-based ontology, even though Hedström and Ylikoski (2010a: 57) have complained of its “extra philosophical baggage” that they wish to avoid.

This article will therefore attempt a critical appreciation of the core principles of analytical sociology from a Bungean perspective. While Bunge’s ideas lend support to a significant portion of analytical sociologists’ intellectual advancements, this article argues that the tendency among them to adhere to the “strong program of microfoundations” (Hoover 2001: 109) is inappropriate, but will be easily remedied if closer attention is given to Bunge’s emergentist version of action-systems theory.

2 The Indispensability of Action Theory

Sociological theory has long been characterized by the pronounced dichotomy between systems and action-theoretical approaches (see e.g., Schwinn 1998; cf. Stichweh 2000), which finds dramatic expression in the sometimes stimulating (but often sterile) debates over the structure-agency dilemma or the micro–macro link.⁵ To put it in a somewhat simplified fashion, in conceptualizing social reality, action-theoretical approaches tend to elevate actors to primacy, while systems theorists settle their gaze on self-regulating systems (Schwinn 1998: 91; Scharpf 1997: 12). Action theorists sometimes hail humans as active beings endowed with agential powers to the point of exaggeration; systems theorists’ exclusive emphasis on social systems at the expense of human agency⁶ has on the other hand led radical social theorists to conclude that systems theory “does not reveal anything about the nature of the conflict, the dynamics of change, or even the motives of the various social groups involved” (Fotopoulos 2000: 440). Indeed, the systems tradition in social science, as represented by the works of Niklas Luhmann (e.g., 1995), has vigorously defended the view that human beings (psychic systems) are the environment of social systems, and that social systems can only receive “perturbations” from the psychic systems. Social systems theory therefore amounts to the denial of the theoretical relevance of human agency (Scharpf 1997: 12; cf. Wan 2011: 85). But as pointed out elsewhere (e.g., Wan 2011), a systems approach does not have to be focused on the alleged autopoietic, self-evolving social systems that undergo, for example, functional differentiation ad infinitum that characterizes Luhmann’s writings. Instead, once it is acknowledged that living people engaging in social activities are the fundamental components (instead of the environment) of social systems, and that human beings are endowed with specific emergent causal powers (such as linguistic abilities and reflexivity) responsible for the mechanisms of social emergence, one is able to liberate systems thinking from the “functionalist ideology that excludes the potentials for social change by human agency” (Fuchs 2008: 23) without abandoning the use of systems concepts.

Likewise, Hans Joas, a leading contemporary proponent of action theory, maintains that the emphasis on the “real actions of real actors” (Joas 1996: 232) should not be taken as a “counter-position to an emphasis on structural forces and restrictions”. The point is, rather, that an adequate conceptualization of action is required for any satisfactory account of “sociality, types of social order, structure or system and dynamics of social change” (Joas 2004: 309; for general discussions of Joas’ action theory, see e.g., Kron 2010: 133–155). In other words, it is necessary to link a systems framework to an action-oriented theory, or to put it slightly differently, to connect action theories with structural models (Maurer and Schmid 2008: 2882; for two recent expositions of the relevance of action theory to social scientific explanation, see Kroneberg 2011 and Panebianco 2009: 15–96).

This is where Bunge’s systems approach and analytical sociology converge. Bunge’s approach is best construed as a sophisticated variant of action-systems theory (Handlungs-systemtheorie; théorie des action-systèmes). On the one hand, it fully recognizes that if an adequate action-theoretical microfoundation is missing, the systems approach will easily fall into the pit of obscurantist holism. To place the systems approach on a firm ontological footing, it should always be remembered that a social system is “ultimately the aggregate outcome of individual actions”, or more specifically, “the features of a social system depend upon the nature, strength, and variability of social relations, which in turn are reducible to social actions” (Bunge 1998: 163, 311, emphasis added), and Bunge has in several works attempted to elaborate on a systemist action theory (e.g., Bunge 1989a: 319–353, 1998: 306–353). For a Bungean sociologist, therefore, social actions are taken as the building blocks not simply for (the overarching term) “society”, but also for its key elements, such as institutions, customs and power relations (Balog 2008: 260). On the other hand, one of Bunge’s most important contributions to the study of systems is the CESM (composition-environment-structure-mechanism) model,⁷ which is grounded in his critical naturalist and dynamicist ontology.⁸ According to this ontology, “all concrete (material) entities are active in some way or other: there are no changeless and totally passive things in the world” (Bunge 1989a: 323). Social systems (understood as concrete/material entities, such as families, firms, schools, and political parties) are no exception, since the mechanisms that construct a social system or keep it going are material processes driven by human (inter)actions (Bunge 1999: 61).⁹

Analytical sociologists also share an uncompromising emphasis on theories of action. As such analytically oriented sociologists as James Coleman, Raymond Boudon, and Jon Elster have steadfastly insisted, changes in macro-social phenomena “must be explained by reference to the actions that brought them about” (Hedström and Ylikoski 2010b: 387). This insistence on the central role of action theory in sociological explanation¹⁰ is, in turn, grounded in the ontological view that it is the individual action and the products of individual action that are ultimately responsible for any persistence of or changes in collective or social properties (i.e., social explananda). Or, in the concise words of Daniel Little (2010: 401), “social outcomes and historical changes are ultimately the result of the actions of individuals within social relations and constraints”.¹¹

It has to be borne in mind, however, that analytical sociology is only a research program that provides “a ‘syntax’ of explanation” (Manzo 2010: 140) or “some general ideas about what good social science is all about” (Hedström and Ylikoski 2010a: 58), which means that its emphasis on theories of action does not “imply a commitment to any form of rational-choice theory and in its barest form it may not make any reference to mental or intentional states whatsoever” (Hedström and Bearman 2009: 8).¹² That is to say, one does not have to subscribe to (even a moderate version of) rational choice theory to appreciate the core principles of analytical social science.¹³

Furthermore, as Manzo (2010: 151–152) usefully reminds us, a number of leading analytical sociologists have critically distanced themselves from an orthodox version of rational choice theory either by exposing its “empirically inconsistent assumptions and hypotheses” (Kroneberg 2008: 238),¹⁴ or by foregrounding “multiple and bounded rationalities” (Selznick 1996: 274; see also Maccarini 2011: 110–112).¹⁵ Boudon, an outstanding example of the latter, argues for the case that “being rational” is best understood as “behaving, acting or believing on the basis of a set of reasons perceived as strong and well articulated with one another” (Boudon 2007: 208).¹⁶ This more sophisticated notion of rationality, or cognitive/ordinary rationality as Boudon calls it, posits that actors tend to ground their choices and decisions in “strong reasons”, which may involve “prudential considerations, empirical data, well-confirmed scientific hypotheses, or universal moral principles that have proved to promote individual welfare and social values” (Bunge 2001a: 198).¹⁷ In fact, a more socially contextualized understanding of rationality has been one of the most fruitful and remarkable developments in this debate (Manzo 2007: 41), as has been made clear by Boudon’s influential writings (for an interesting application of Boudon’s model to empirical research, see Brym and Hamlin 2009), even though I agree with Manzo (2010: 161) that an empirically more adequate theory of action has to integrate cognitive rationality with the dimensions of habitus and especially emotions that receive insufficient attention in Boudon’s model.¹⁸ It is therefore fallacious for critics to assume that the project of analytical sociology amounts to nothing but “rebranded rational choice sociology” (Hedström and Ylikoski 2010b: 394).¹⁹

3 The Quest for Microfoundations and the Mechanismic Explanation

Bunge’s integration of the action dimension into his systems approach is bound up with his critique of both holism and atomism. Holism either treats a system as an indecomposable totality and thus paves the way for philosophical intuitionism and irrationalism, or it prioritizes a system over its component parts by asserting that the former determines the latter (see Wan 2011: 85–86). In stark contrast, Bunge’s systemism serves as a justification for performing explanatory (micro)reduction. That is to say, it invites us to analyze every system into its composition, environment, structure, and mechanism, and seeks to explain the emergent properties of a system (at least in part) in terms of the complex interaction of its component parts.²⁰ In social research, these components are precisely human beings and their (inter)actions that form, maintain, transform, or dismantle a social system. On the other hand, atomism (ontological individualism) has to be rejected for it typically downplays social relations (i.e., the ties that hold people together) and macro-social facts and entities.²¹ The methodological prescription of systemism is therefore divide et coniuga, or analysis and synthesis (see e.g., Bunge 1998: 77, 2003: 126; Bunge and Ardila 1987: 97). Hedström’s account of the research strategy of analytical sociology—“dissecting” the social processes into basic constituents and building models to causally reconstruct the generative processes of collective phenomena—is obviously well in line with that of systemism (see also Hedström and Ylikoski 2010a: 60).

What is involved here are two key terms that recur in analytical sociologists’ works: microfoundation and mechanism. These two words are so intimately related that they are often used interchangeably or simultaneously. To put it more precisely, for analytical sociologists, what is of paramount importance in sociological explanation is to provide rigorous microfoundations, which amount to fleshed-out accounts of “the pathways by which macro-level social patterns come about” (Little 1998: 4), or of “how efforts on the ground…may prompt macro-level changes and responses” (Powell and Colyvas 2008: 295). These are in turn classified into different types of mechanisms in order to fully capture the “process character of the social” (Kron 2010: 220n). In the standard Boudon-Coleman diagram (also known as the Coleman boat), three basic mechanisms can be distinguished: macro–micro (situational), micro–micro (action-formation), and micro–macro (transformational) mechanisms (see e.g., Hedström and Swedberg 1996, 1998; Hedström and Ylikoski 2010a, b; Kroneberg 2008).

The situational mechanism, which is equivalent to Karl Popper’s concept of “situational logic” (Situationslogik) and what Esser (1994, 1999: Chapter 10) calls the “logic of the situation”, refers to a specific social situation that shapes the actors’ beliefs, opportunities, and orientations. It involves at least three aspects: (1) the available alternatives to the actors involved; (2) the restrictions that govern the choice of the alternatives; and (3) the evaluation of the possible consequences of the choices made (Esser 1994: 180).²² As Wikström (2008: 130) explains, the situational mechanisms “help to identify what aspects of society…and what aspects of an individual’s development and life-history…are relevant as ‘cause of the causes’ in the broader explanation”. It is the serious consideration of these structural elements that distinguishes analytical sociologists’ “structural individualism” from naïve versions of methodological individualism. Hedström and Bearman (2009): 4; see also Hedström 2005: 5n; Hedström and Ylikoski 2010a: 60, 2010b: 393) therefore write that “[s]tructural individualism differs from traditional methodological individualism in attributing substantial explanatory importance to the social structures in which individuals are embedded”. In other words, social institutions, relations, structures and orders appear not only in the explanandum but also in the explanans (Schwinn 2008: 310). Or, as Esser (1993: 404, quoted in Heinz 2004: 17) puts the matter, structural individualists acknowledge that society is “a force (Kraft) that is actually prior to the concrete individuals and strongly shapes their actions”.²³

The action-formation mechanism involves a multiplicity of psychological and social-psychological mechanisms (e.g., the well-known “framing effect” [i.e., the manner and order in which options are presented will influence people’s choices] studied by psychologists and behavioral economists, or preference adaptation [i.e., the nonconscious adjustment of wants to possibilities] studied by Amartya Sen and Jon Elster, among others) that operate at the micro level, which demonstrates “how a specific combination of individual desires, beliefs, and action opportunities generate[s] a specific action” (Hedström and Swedberg 1998: 23). What Esser calls the “logic of the selection”, that is, “the logic, or nomological rule, of the selection of action” (Esser 1994: 181) is intended to specify these aspects and processes. Most analytical sociologists maintain that the proposed action-formation mechanism should be based on findings of psychological and cognitive sciences (Hedström and Ylikoski 2010a: 60), and a Bungean sociologist would not hesitate to agree.²⁴ One of the main weaknesses of the hermeneutic tradition in social studies, from the Bungean perspective, lies largely in its frequent allusions to mental events without trying to grasp them as complex biological, psychological, and social processes (Moessinger 2008: 238).²⁵

Finally, the transformational mechanism, or what Esser calls the “logic of aggregation”, specifies or reconstructs the often complex processes whereby the actions and interactions of interconnected actors generate the intended or unintended macro-level outcomes, such as the intriguing (resolution of) the tragedy of the commons (e.g., Vollan and Ostrom 2010; Rustagi et al. 2010). (Since this process is rarely one of simple linear aggregation of individual actions, Esser’s term is partly misleading.) As self-proclaimed “structural individualists”, analytical sociologists also attach particular importance to this type of mechanism, because it reminds researchers of the fact that actions are not performed by mutually independent individuals, but take place in various “topologies of social networks” (Hedström and Ylikoski 2010a: 60).

Now, a number of analytically oriented sociologists strongly argue that only an analysis that studies the whole chain of these three “logics” or mechanisms passes for a satisfactory sociological explanation (Hedström and Ylikoski 2010a: 59; 2010b: 393), but this is also where analytical sociologists diverge. Indeed, for some of them, the emphasis on mechanisms is easily translated into a focus on individual persons and their (inter)actions, since it is generally believed by analytical sociologists that “theories of individual action” constitute “the core of the explanatory mechanisms” (Hedström and Ylikoski 2010b: 393, emphasis added; see also 2010a: 59). But there are others skeptical of this dogmatic insistence on the need to spell out individual-level mechanisms in all social research. For example, no sooner does Barbera (2006: 45, original emphasis) argue, along Hedström and Ylikoski’s line of reasoning, that “macro-phenomenon [is] to be explained by means of some generative mechanism at the level of the action system”, than he emphasizes that the concept of action invoked here is not restricted to individual agents, but includes super-individual entities (Barbera 2006: 45n). That is to say, while theories of action are deemed as necessary to provide the microfoundations of sociological explanation, analytical sociologists differ in their views on the necessity of individual-level action-formation mechanisms.

4 Supra-individual Actors qua Concrete Systems with Emergent Causal Powers

The nub of this problem, in my view, lies in the crucial question regarding the ontological status of macro-social entities. This question can in turn be subdivided into two questions: (1) Do supra-individual actors/agents exist, or, are actors/agents endowed with causal powers necessarily (socially situated) individuals? (2) Are macro-social mechanisms real processes, or are they merely collective concepts to which any ontological status should be denied? Let me address the first question in this section and consider the second in the next.

A mechanism-based explanation generally has to do with certain forms of part-whole relationship (e.g., prising opening the “black box” of a system), whether the object of research is the working of a concrete system, or an event/phenomenon that involves interacting concrete systems and their emergent causal powers. In this sense, to say that a mechanism is invoked in a causal explanation typically means that certain details at a lower level of organization are specified (and therefore certain sorts of microfoundations for the causal claims are offered), or that a multi-level (e.g., macro–micro–macro or micro–macro–micro) explanatory strategy is carried out. In other words, mechanismic explanations are “deep” insofar as they integrate levels of analysis by performing micro-reduction (Marchionni 2008: 326; see also Kuorikoski 2009: 148; Barbera 2006: 34; see also Wan 2011 on reduction and reductionism). However, as Kincaid (1997: 26, 2002: 304) argues, the description of mechanisms can be in more or less detail and at different levels. That is to say, the micro-foundations of a social explanation need not be built exclusively at the level of individual persons. To challenge or tone down the dogmatic insistence on going down to the individual level to explain collective phenomena in all cases, one tends to emphasize (1) the pragmatic reasons for not doing so (e.g., when suitable long-term data concerning individuals is unavailable), (2) the possibility of successful causal inference without referring to any individual-level mechanisms (e.g., with the help of sophisticated statistical techniques), and (3) the well-known “infinite regress” question that leaves individualist mechanism-based accounts untenable as well (see Wan 2011: 148–150, 153–154). But the equally important issue that interests us here is the ontological status of supra-individual actors of various kinds, which possess systemic or emergent properties absent from the individuals that constitute them.

In spite of Hedström and Swedberg’s claim that “the elementary ‘causal agents’ are always individual actors” (Hedström and Swedberg 1998: 11), a number of analytical sociologists do recognize the possibility of enlarging the set of “causal agents” to include not only individuals but also supra-individual entities. For example, as Boudon (1981: 37) argues, a supra-individual entity can be treated as a “many-headed” but unitary actor insofar as it has a mechanism for collective decision-making. Boudon refers to this kind of actors as “organized groups”, which should be kept distinct from non-organized (or “latent”) groups, namely, the collections of individuals with certain common interests, such as social classes (see e.g., Boudon and Bourricaud 1989: 18, 406; Boudon 1981: 37). As I will explain later, these “non-organized groups” are in fact concepts instead of concrete systems in Bunge’s sense.

Besides, the typology of supra-individual actors developed by actor-centered institutionalists (see especially Mayntz and Scharpf 1995; Scharpf 1997; Schimank 2004) is particularly useful for our purpose. For Scharpf (1997: 52), a supra-individual actor, or a composite actor as he calls it, is above all characterized by “a capacity for intentional action at a level above the individuals involved”,²⁶ and they typically have an institutional structure, within which individuals (and sometimes lower-level composite actors) interact to coordinate joint actions. Scharpf (1997: 54) further distinguishes between two kinds of composite actors:

• 1. Collective actors, which are “dependent on and guided by the preferences of their members”. Grouped under this category are coalitions, movements, clubs, and associations.

• 2. Corporate actors, which have “a high degree of autonomy from the ultimate beneficiaries of their action and whose activities are carried out by staff members whose own private preferences are supposed to be neutralized by employment contracts”. As analyzed by James Coleman (e.g., 1974, 1982, 1990),²⁷ these actors, typically composed of networks of roles and positions (instead of natural persons) and structured by formal decision rules, are greater in their power, size, complexity, flexibility, and longevity, and therefore they stand in “asymmetric” relations with natural persons in contemporary society.

A number of social theorists and philosophers have attempted to specify the conditions under which composite actors become “group agents” stricto sensu, which “have a distinct intentional profile from the profiles of their members” (Pettit and Schweikard 2006: 33). Whatever their disagreements, it seems true that all composite actors have the means of formulating, taking, and acting upon decisions (Sibeon 2004), and they are more or less endowed with what Margaret Archer (2000: 266; see also 1995: 259–260) calls corporate agency, that is, “capacities for articulating shared interests, organizing for collective action, generating social movements and exercising corporate influence in decision-making”.²⁸ However, the extent to which and the way in which this agency is exercised is of course an empirical matter that can only be decided on a case-to-case basis.

Now, what does Bunge’s approach have to say about the question of supra-individual actors? The aim of Bunge’s emergentist systemism is to provide a “completely general model of concrete systems of any kind, living or nonliving” (Bunge and Ardila 1987: 46). As such, it does not and is not intended to serve as a ready-made solution to all theoretical problems concerning the nature of supra-individual actors.²⁹ But it does constitute a general framework for understanding supra-individual actors as concrete systems with emergent properties and causal powers (e.g., the cohesiveness of a union, the power of a lobbying group to influence the decisions of legislators, the capacity of a firm for efficient production, etc.), definite structures (e.g., the relations between group members and the administrative staff in the case of corporate actors), and an essential mechanism or its modus operandi (e.g., the decentralized decision-making mechanism of a grassroots organization).

Another merit of Bunge’s approach is that it provides an ontologically sound basis upon which to assess other approaches to conceptualizing supra-individual actors and other seemingly elusive collective concepts. Let me illustrate this with some examples.

• 1. Mouzelis (2008: 112, 258) famously refers to “classes” as macro-actors, which he defines as “collective actors or very powerful individuals whose contribution to social construction is considerable”.

• 2. Similarly, Adloff (2006: 4) writes that “classes behave in ways not ascribable to individuals”.³⁰

• 3. Presumably most Marxists endorse the notion that “social classes are historical actors that emerge as collective agents” (Allen 2008: 32).

A social group, as Bunge (1998: 67, 1999: 23) construes it, is a collection of individuals that share certain (biological, economic, political, or cultural) features, such as “the poor”, “wage earners”, “city dwellers”, and “college student voters”. Importantly, social groups are not necessarily concrete (material) systems. Even when the members of a group may share certain common interests as in the case of socioeconomic classes or consumers (to which Ralf Dahrendorf [1959] refers as “latent groups”), it remains misleading to accord any ontological status (e.g., being supra-individual actors in possession of emergent causal powers) to these types of groups. In short, only cohesive social groups (i.e., what Boudon calls “organized groups”), the members of which are held together by connections, bonds, or forces of some sort, are concrete systems. In social research, only concrete supra-individual systems (e.g., clubs, firms, schools, hospitals, trade unions, and political parties) can enjoy the ontological status of supra-individual actors (and can be further divided into subtypes such as collective actors, corporate actors, group agents, and so on).

In the above three examples, social classes are treated as composite actors that are somehow capable of behaving as a unit. A Bungean social scientist will undoubtedly dispute this by emphasizing that a social class is not a concrete system, but a taxonomic collective, i.e., an analytical construct that represents commonalities among individuals (see also Harré 1981, 1993, 1997; Sayer 2010: 101, 119, 244). As Bunge (1998: 69, emphasis added) puts it, “classes are objective yet not concrete…because the members of a class, unlike those of an organization, are not necessarily bonded together by any social ties”. In other words, a class cannot be considered to be a supra-individual actor, which (1) is eo ipso a concrete system with a bonding structure and emergent causal powers, and (2) can be further analyzed by, for example, Bunge’s CESM model.

5 From Supra-Individual Actors to Macro-social Mechanisms: A Dangerous Move?

It has been demonstrated in the preceding section that (1) the Bungean approach joins the majority of analytical sociologists (e.g., Boudon and Coleman) in acknowledging the ontological status of supra-individual actors (cohesive social groups) as concrete systems with novel causal powers, a bonding structure, and specific mechanisms that make it behave as a unit in certain respects; (2) with its insistence on the distinction between concrete systems and taxonomic collectives such as social classes, the Bungean approach is highly cautious about ascribing ontological existence and causal powers to macro-social entities.

Now, a second argument against the alleged necessity of specifying individual-level mechanisms in all social research is that macro-social mechanisms do play a significant role in social scientific explanation, and they can be ascribed sui generis ontological status in a number of cases.³¹ Given that analytical sociologists have maintained that “there is no such thing as pure social causation from macro-state to macro-state” (Little 1998: 198; see also Hedström 2005: 74) or that “there are no social causal mechanisms that do not supervene upon the structured choices and behavior of individuals” (Little 1998: 203; see also Bennett 2008: 215; Schwinn 2008: 312), does it follow that macro-level mechanisms do not exist (e.g., Hedström and Swedberg 1996: 299; Abell et al. 2010)? Is it really the case that any reference to macro-social mechanisms in social explanation violates the principle of rigor and precision so cherished by analytical sociologists?

A more orthodox methodological individualist may insist that macro-social mechanisms are dubious “collective concepts”, and that when one lumps “composite actors” and “collective concepts” together, the fact will be neglected that collective concepts cannot be treated as causally generative or real.³² Or, in Max Weber’s well-known words, the task of sociologists is to “put an end to the mischievous enterprise which still operates with collectivist notions (Kollektivbegriffe)” (quoted in Udehn 2001: 98)³³ by reflecting upon their ontological nature: they are nothing but “a certain kind of development of actual or possible social actions of individual persons” (Weber 1978: 14; for general discussions see Udehn 2001: 96–99, 2011: 211; Donati 2010: 63–64).

To clarify the matter, I wish to make two points. Firstly, as argued above, the majority of analytical sociologists in fact support a weak (or “enlightened”, see Greshoff et al. 2011: 11) version of methodological individualism, which fully incorporates macro-social entities, properties, institutions, and relations as determinants of the social action of individuals (Udehn 2001: 295–309, 318–319, 2003: 157, 2011: 212), as the Boudon-Coleman diagram clearly exemplifies. Hedström and Ylikoski (2010b: 393; see also Hedström and Bearman 2009: 9) therefore attach significant explanatory importance to such standard collective concepts as “institutional or informal rules and norms”, while the more orthodox methodological individualists will probably accuse analytical sociologists of “smuggling” collective concepts like norms and values into their analysis (see Brym and Hamlin 2009: 93-4n; Hamlin 2002: 65). More specifically, these “macrosociological variables” (to borrow Boudon’s phrase) are causally efficacious in at least two important ways (see e.g. Bunge 1996: 250; 1998: 441; Udehn 2001: 318–319; Wan 2011: 134–137)³⁴:

• 1. They may operate “behind the back” of people by shaping individual and composite actors’ opportunity sets and enabling and/or constraining their actions and interactions to a significant extent;

• 2. They may exert influence on individuals’ preferences, beliefs, and intentions (that is to say, de gustibus est disputandum), and sometimes they may be what individuals intuitively or reflexively take into consideration when deciding what to do, when, and how (see Hedström and Bearman 2009: 11). For example, a political event is often the outcome of the concerted actions of people in response to such irreducibly macro-social issues as inflation, unemployment, political oppression, or war; a successful businessman tends to take into account various macro-social processes or facts before taking action, such as labor movement, state of the market, business opportunity, and so on.

Secondly, I believe that Bunge’s understanding of social mechanisms as the processes unfolding in and among social systems is an effective response to the claim that any reference to collectivities other than composite actors in sociological explanation runs the risk of reification or what Boudon (following Jean Piaget) calls “totalitarian realism” (see e.g., Boudon and Bourricaud 1989: 5–6). A Bungean social scientist will readily recognize that such collective concepts as class and race, which are analytical constructs that represent similarities among individuals, do not qualify as real/material/concrete systems, and therefore they lack genuine efficient causal powers to “push and pull”, i.e., to bring about (or prevent) change by acting on objects. In the words of critical realists (e.g., Rom Harré), they are not “powerful particulars”, i.e., causally active things or complex objects endowed with generative powers (see e.g., Harré and Madden 1975; Kurki 2008). However, unlike these analytical constructs, mechanisms and processes occur in concrete/real/material systems, and they are what make the (emergent) causal powers of these systems possible. To wit, they are perfectly real/material (Bunge 2006a, b: 11).³⁵ As Bunge (e.g., 1981, 2001a, 2006a, b, 2010a) puts the matter, to be real (material) is to become (i.e., undergo changes, events, or processes). Things, their properties and the processes thereof are all real/material from this perspective (Bunge 2006a, b: 28).

Now, since it may still be argued that the ontological status of macro-level mechanisms is dubious, more discussion on the terms “macro” and “mechanism” is called for:

• 1. The macro–micro distinction is always relative, and is bound up with Bunge’s overall understanding of systems and levels: If the macro-level denotes the collection of all the systems sharing certain peculiar properties, then the corresponding micro-level is the collection of the components of the systems in question (Bunge 2003: 133). Therefore, when a Bungean social scientist writes of macro-social levels, factors, events, processes, or mechanisms, the term “macro” (pertaining to social systems of all types and sizes, from a firm to a nation-state) is used only relative to the “micro” (components of a given social system, be they individual persons or composite actors), not in some absolute sense. It also follows that it is possible and often fruitful in social research to interpolate a meso-level between a micro- and a macro-level so that more same-level and inter-level processes can be taken into account (see also Greshoff et al. 2011: 19–22).

• 2. In Bunge’s usage, the ontological concepts of system and mechanism are defined in terms of each other. A mechanism means the collection of processes that occur within a system (and often among systems in the case of the social world), bring about (or block) its transformations, and alter (or maintain) its structure. To examine the mechanism(s) of a system in detail, therefore, it is necessary to perform a reduction to a lower level by discovering the system’s composition, structure(s), and connections with the environment, and particularly by showing what the system’s components do and how they do it (Bunge 1999: 24). This is why the mechanism-based explanation is multi-level in nature (see e.g., Mahner and Bunge 1997: 94; Bunge 2001a: 88, 2003: 116–117) and can be represented by the two-tier (or three-tier if necessary) Boudon-Coleman diagram, even if the mechanism invoked is “macro” relative to the actors involved.³⁶

• 3. If macro-social mechanisms are not “real”, it follows that chemical reactions in reactors, metabolism in cells, mental processes in brains, and the circulation of blood in the cardiovascular system are not either. Importantly, there is no doubt that these macro-social processes (but not such collective concepts as social classes, which are merely taxonomic groups) can in turn be dissected into their components (i.e., individual and composite actors), their features, socially situated actions, and structured interactions. That is to say, they can be further provided with an action-theoretical microfoundational basis. But from this does not follow that “every adequate description of a social mechanism must be phrased in individualist terms” (Steel 2005: 950). Indeed, as Hedström and Ylikoski (2010a: 52) themselves have acknowledged, “for a mechanism to be explanatory it is not required that the entities, properties, and activities that it appeals to are themselves explained”. The same holds for macro-social mechanisms (such as the processes of cooperation, competition and emulation that take place in and among social systems) that (provisionally) take as given the involved individual and composite actors, their emergent properties, and (inter)actions.³⁷

To conclude, Bunge’s approach is a praiseworthy antidote to what Hoover (2001: 109) calls the “strong program of microfoundations”, which “reject[s] realism with respect to macroeconomic [and macrosocial] entities”. In this sense, it arguably provides a middle course between the Scylla of “totalitarian realism” and the Charybdis of ontological individualism.

6 Concluding Remarks

In this article I have argued that while the principles of analytical sociology and those of Bunge’s approach share a lot in common (especially the importance of action-theoretical microfoundations), the latter brings to the fore the ontological status and explanatory role of supra-individual/composite actors (as concrete systems endowed with emergent causal powers) and macro-social mechanisms (as processes unfolding in and among social systems), and therefore it does not stipulate that every causal explanation of social facts has to include explicit references to individual-level actors and mechanisms.

As Bunge (1999: 62) contends, since “[a]ny number of intermediate levels may…be interpolated”, entities at the micro-level can be “persons and social subsystems”. Trade unions, political parties, social movement organizations, as well as their decisions and interactions, for example, are often all one has to refer to at the lower tier of the Boudon-Coleman diagram.³⁸ Sometimes it is indeed necessary or fruitful to construct a three-tier diagram to take individual-level actors and mechanisms into account (see e.g., Bunge 2001a: 127), but this depends on the nature of one’s research question (see particularly Pozzi 2007).

Bunge makes this point even clearer when he discusses his CESM model:

In practice we use the notions composition, environment, structure, and mechanism at a given level. …Except in particle physics, we never handle the ultimate components of anything. …[W]hen forming a model of a social system (or group) we usually take it to be composed of whole persons; consequently we limit the internal structure of the system to interpersonal relations. However, nothing prevents us from constructing a whole sheaf of models of the same society…. We do so when we take certain subsystems of the given social system — for instance, families or formal organizations — to be our units of analysis.

(Bunge 2003: 37, emphasis added)

To conclude, with its insistence on integrating top-down (analytic, microreductive) and bottom-up (synthetic, macroreductive) research strategies, Bunge’s systemism fully appreciates the merits of the microfoundations approach and structural individualism, serving both as a “healthy safeguard against overly general and indefinite ideas” (Selznick 1992: 51) and as an antidote to the ontological-individualist dismissal of all macro-social entities as causally irrelevant fictions. Furthermore, by clarifying such fundamental concepts as system, mechanism, and emergence, Bunge’s emergentism rovides a plausible way of locating and settling such troubling questions as whether “social wholes” exist in their own right or whether “society is more than the sum of individuals”. Freed from the burden of dogmatic insistence on making reference to the individual level as a conditio sine qua non for sociological explanations, it invites us to delve deeper into the “multiscaled” (DeLanda 2006: 34–40) or “heterarchical” (Kontopoulos 1993) social reality, in which a multiplicity of concrete social systems interact and various intertwined social mechanisms at different levels are at work (see also Brante 2001; Moessinger 2008: 44–52, 224–241; Johnson 2008; Crossley 2010).³⁹ This article therefore submits that the laudable intellectual project of analytical sociology may be further advanced if Bunge’s insights are integrated effectively.