Critical realism in economics and open-systems ontology: a critique.

Abstract This paper examines the treatment of ontology offered by critical realism. It addresses much of the material elaborated upon in two editions of this journal. Three main groups of criticisms are made here of the critical realist treatment of open systems. It is argued that critical realism, particularly in the project in economics emanating from Cambridge, UK, tends to define systems in terms of events. This definition is shown to be problematic. The exemplar of a closed system provided by critical realism of the solar system is shown to be flawed in that it is not closed according to the closure conditions identified by critical realism. Second, the negativity of the definitions adopted is problematic for heterodox traditions attempting to build positive programmes. Furthermore, the dualism of the definitions is also inconsistent with Dow's approach, which has ramifications for the coherence of post Keynesianism. Third, the definitions tend to polarize open and closed systems and ignore the degrees of openness evident in reality. The polarization of systems leads to polarized methodology and unsustainable arguments to reject so-called "closed-systems methods."

Keywords: open systems, closed systems, critical realism, post-Keynesianism, dualism

INTRODUCTION

Increasingly, the term "open systems" is being used to describe the complex and unpredictable environment faced by economists and economic agents (Setterfield 2003; Lawson passim). "Open systems" has even been advanced as a potential basis for heterodox economics (Hodgson 1988; Dunn 2001; see also Downward 1999), and has arguably become a tacit assumption therein. However, Hodgson (2000) and Mearman (2002a) have argued that the concept is somewhat undeveloped and that, therefore, it would seem inappropriate for heterodox economics currently to be based squarely on it (cf. Dunn; Dow 2000). Even if an "open-systems methodology" remains only one of the pillars of heterodoxy (as in Downward 1999; Lee 2002), it still requires development. This paper considers particularly the ontology of open systems.

This paper aims to develop the concept of "open systems" by offering a constructive critique of the increasingly influential critical-realist view. Critical realism (CR) is, of course, a variegated literature. This variety is not a weakness per se. Indeed, uniformity of meaning can reduce flexibility and prevent change and therefore make an approach analytically poorer. However, while the critical-realist literature is somewhat diverse, a key presupposition of this paper is that there exists a "Cambridge school" view developed by Tony Lawson (1997, 2003) and others, mainly those (mostly at Cambridge University) closely influenced by Lawson. (1) This view has figured prominently in this journal (Winter 1996; Fall 1998). Again, there is variety within this group; however, there is sufficient coherence on key issues to justify the description of the group as a distinct school. Arguably, "open systems" is, along with ontological depth, one of the two most important concepts in this Cambridge school view. Crucially, moreover, this paper holds that CR in economics is dominated by the Cambridge view.

This paper argues that there are three problems in this Cambridge school's treatment of open systems: 1) it is dominated by event-level definitions--which also reflects an underdeveloped concept of "system;" 2) it emphasizes negative definitions; and 3) it tends towards polarizing definitions. These problems are shown to be problematic in many ways: It weakens the ability of CR to engage in constructive work, it raises questions about the possibility of coherence for post Keynesianism, and it leads to polarized methodological accounts on various issues. The paper proceeds in the order of these criticisms.

Before embarking on this critique, one other presupposition should be stated: In criticizing the existing treatment of open systems in CR in economics, an alternative conception of open system is held in mind. It is far beyond the scope of this paper to elucidate fully this alternative conception (see Mearman 2002a, 2004). Moreover, this alternative is not the subject of the paper; this paper does not stand or fall on that alternative. Furthermore, the alternative is as yet merely a sketch. Nonetheless, several key elements of this sketch can be noted. The alternative conception of an open system implicit in this paper (hereafter referred to as the "Open System Ontology," or OSO) is one that exhibits ontological depth, structures and causal mechanisms, multiple and interacting mechanisms, internal relations and emergent properties; it thus owes much to the ontology outlined by CR. However, reflecting the broader intellectual history of the term "open system", rooted, for example, in the General Systems Theory (GST) of von Bertalanffy (1968), the OSO also stresses a system boundary, which is fuzzy and permeable, so that mechanisms can affect the other mechanisms in the system. In such an open system, it is unlikely that strict regularities of events (of the kind "if event X occurs, then event Y will occur") will result; in this sense, the OSO shares the key defining characteristic of an open system held by CR. However, the definition of the OSO is not exhausted by that lack of event regularity; rather, in this alternative conception, a "closure" refers to situations in which, to some extent, the sources of openness, for example, emergence, are absent, are inoperative, have disappeared or have been removed. Openness, therefore, can occur, for example, in the nature of the object, its constituent mechanisms, relations between mechanisms, or in the nature of the system boundary.

DEFINITIONS OF SYSTEMS IN TERMS OF EVENT REGULARITIES

Definitions

There is a range of critical-realist definitions of closed systems. Closed systems are variously defined as being "cut off" from external influences (Collier 1994: 128; Bhaskar, 1978: 69); "isolated" (C. Lawson, 1996); where outside factors are "neutralise[d]" (Collier 1994: 33); and in which all disturbances are anticipated and "held at bay" (Lawson 1997: 203). The net result, according to Collier (1994: 33), is that one mechanism alone operates, unaffected by other mechanisms (Lawson 1997: 203). (2) The obvious example of such a scenario is the experiment. Thus, Collier (249-50) and Lawson (1999a: 216) effectively equate closed-systems methods as experimental (and open-systems as non-experimental). (3) Strictly, the focus on isolation is incorrect, since experimentation imposes requirements inside the isolated area. Archer (1998: 190) notes that even in isolated environments, the problem-solving and perhaps capricious nature of humans means that the "closure" of experimentation cannot be achieved (see Mearman 2004). These problems are such that Bhaskar (1986: 101) claims that closed systems are "impossible" in social science.

Additional (partial) definitions of a closed system in CR are that relations within a system are stable (Beed and Beed 1996: 1099); and that conditions are imposed on the individuals in the system (Bhaskar 1978: 69). A fully closed system is where all individual and system criteria for closure are satisfied (Bhaskar 1978: 104), which suggests both a regularity of behaviour (Bhaskar 1978: 253, n. 1) and a homogeneous--unchanging and uniform--environment (Lawson 1997: 218); consequently, transformative action is impossible in a fully closed system (Bhaskar 1986: 31). According to CR, closure is achieved when specific closure conditions hold. The most significant of these are the Intrinsic Condition of Closure (ICC) and the Extrinsic Condition of Closure (ECC). The ICC requires that the object in question has a constancy or constancy of change, such that elements "inside" the system are stable enough to be identified. The ECC entails that no outside forces impinge on the particular object or system, or that any external effect is constant. It is significant that a strict distinction is usually made between "closures" and the ICC and ECC, the "conditions for closure". Both the distinction between closure and the conditions for it, and the closure conditions themselves, are prevalent in the "Cambridge" treatments of closed systems.

This paper argues that despite the apparent variety of definitions--which, to repeat, is in many ways to be valued--the Cambridge school definition of open and closed systems has been, and is being increasingly, restricted to one. That is, closed (and hence open) systems are defined in terms of events and their regularity.

Consider the following definitions of closed systems: Closed systems can be identified when the symmetry (of explanation and prediction) thesis holds (Sayer 1992: 130), or where there is a warrant for education (inference to particular instances) (Bhaskar 1986: 30). A closed system also means that there is a one-to-one relationship (isomorphism) between mechanisms and events (Lawson 1994a: 517), which implies the main definition of a closed system: a unique relationship between antecedent and consequent (Bhaskar 1978: 53); a stability of empirical relationships (Collier 1994); or a constant conjunction of events. This definition flows from the Humean argument that only regular successions between events--not underlying mechanisms (should they exist)--can be identified. Hence, causality is conceived as merely correlation, which in turn calls for the identification of event regularities between isolated atomistic states (Rotheim 1999: 73).

Lawson identifies closed systems as being where the formula "if event of type X occurs, then event of type Y will occur" (where X and Y can be scalars, vectors or matrices--Lawson 1994a: 507, n. 9). More recently, Lawson has modified this definition further to take into account the common practice of completely specifying the conditions under which closure holds. Thus, closed systems conform to the formula, "if X, then Y, under conditions E" (Lawson 1989a: 63, 1995a: 15). Such event regularities could be either deterministic or probabilistic (Lawson 1999b: 273), which in the latter case means that events will be in regular succession within some well-behaved probability distribution (Lewis and Runde 1999: 38; Lawson 1997: 76). (4) In this case, the closure is stochastic (Lawson 1997: 153-154). Lawson repeats this event-level definition in his most recent work (Lawson 2003: 5, 15, 23, 41, 103, 105, 143, 222, 306). (5)

Moreover, instances of this event-level definition of closed systems in terms of constant conjunctions of events have appeared in this journal: Lawson (1996: 407, 1998a: 359, 369), Pratten (1996: 439), C. Lawson (1996: 451, 459), Lewis (1996: 487) and Rotheim (1998: 326, 329-331) all use the definition. Of course, all of those authors affiliate in some way to Cambridge University. Clearly the event-level definition is not the only one offered by CR. However, it is argued that this is beginning to be the dominant definition, particularly in economics, and even more so for those (mainly at Cambridge) influenced by Lawson. This is shown most clearly in the definition of an open system.

Bhaskar (1978) defines open systems as the lack of "regular" (p. 33) or "invariable" (1978: 73) succession; no unique relationship between variables (1978: 53); or a non-invariance of empirical relationships (1978: 132). For Sayer (1992: 122), openness entails short-lived or non-existent regularities. Essentially, an open system is identified as, "Not 'if X then Y'", or as where there are no constant conjunctions of events (Bhaskar, 1989: 16). This definition has been adopted by the recent literature on CR in economics. Therein, open systems are mainly defined as where there are no event regularities (Pratten 1996: 423, Rotheim 1999; Lawson 2003: 79, 82, 119, 223-224) (6) or as systems lacking sharp (i.e. precise) stable event regularities (Lewis and Runde 1999: 38).

Critique

It was argued above that the Cambridge school of CR in economics mainly defines open/closed systems in terms of event regularities. Immediately, to pre-empt any critique, it could be argued that if this is how CR defines open systems, that is the end of the matter. However, this paper holds that such an …