that the transcendental realist perspective throws up. The essential point here is that while the traditional constant conjunction view of science entails the goal of control along with the amelioration of events and states of affairs, the transcendental realist perspective offers instead the goal of human emancipation. For while on the former account the point is to fix certain event(s) x in order to determine/control other event(s) y, on the transcendental realist understanding the aim must be to transform structures in order to enhance the scope for realizing human potential and to broaden opportunities generally. In the latter case, in other words, the aim must (or can) be to replace structures that are unwanted, unneeded, and restrictive by those that are wanted, needed, and empowering. Of course, the emphasis on policy as concerned with structural transformation is hardly Critics of traditional economic theory have always denied the closed character of economic systems. They have stressed instead the open character of economic processes and have challenged above all the belief in their self-regulatory tendencies.
- K. William Kapp, 1976
The branch of contemporary economics that its orthodox or mainstream proponents refer to as "pure theory," "economic theory," or just plain "theory" appears currently to be in a state of some disarray. In recent years, in particular, various problems have come to light with the project that has led even its proponents to question its capacity for accounting for real-world events or for assisting in policy formulation. Needless to say, in response to the diagnoses made, various solutions have been put forward, and certain transformations in aspects of the project have been proposed. I want to argue, however, that the remedies suggested, if in the context sometimes remarkable, are not going to make things sufficiently better. And the reason for this is simply that the problems identified do not represent the real nature of the relevant disorder. The aim here, then, is to attempt to identify the project's essential problem and to point the way to its transcendence.
It should perhaps be emphasized at the outset that although the assessment that follows rests upon critical analysis - one that is essentially an elaboration of the quotation by Kapp reproduced above - the perspective that is argued for is found to be not only not debilitating for, but actually empowering of, an economic theory contribution to science and policy analysis.
The Changing Features of Orthodox "Pure Theory"
If it is commonplace to find references to something called "economic theory" or "pure theory," etc., in modern economic literature, there appears to be some disagreement about what precisely it is or entails. In the assessments of critics (and some proponents too), the project has tended to be characterized according to various nominal features that reoccur in prominent contributions. Specifically, "economic theory" is widely interpreted as a body of substantive thought that, among other things, focuses upon individuals rather than collectivities; upon exchange activities rather than production or distribution; upon optimizing (maximizing or minimizing) behavior rather than satisficing or habit following; upon conditions of perfect competition rather than oligopoly or monopoly; upon structures facilitating constant (or decreasing) returns to scale rather than increasing returns; upon presumptions of perfect knowledge and foresight or "rational expectations" rather than uncertainty or ignorance; upon end-states, fixed points, or equilibria rather than processes in time; upon functions (utility, cost, preference, profit, etc.,) that are well behaved (where appropriate: convex, differentiable, fixed, well ordered over all the arguments, etc.,) rather than otherwise, and so on.
Now if these are the sorts of nominal characteristics that have often been associated with "economic theory," it is equally noteworthy that, at least among the apparently more reflective "theorists," most of these features, though acknowledged as prominent, have not been considered to be in any way essential or fundamental. The various meta-theoretical contributions of Frank Hahn, at least those made prior to the last few years or so, provide good examples of this interpretation. In both his Jevons memorial lecture titled In Praise of Economic Theory and the introduction to his collection of essays titled Macroeconomics and Equilibrium, for example, a quite different account can be found. Specifically, here we find only a restricted set of features of the "neoclassical" economic theory project explicitly elaborated as essential. Three that are enumerated are (1) an individualistic perspective, a requirement that explanations be couched solely in terms of individuals; (2) an acceptance of some rationality axiom; and (3) a commitment to the study of equilibrium states.(1) In addition to these features, Hahn also expresses a preference for highly general, simple theories over particular ones and a conviction of the need to draw an explicit distinction between axioms and assumptions. On the latter issue, both axioms and assumptions are to serve as premises of theoretical analysis. But whereas axioms (such as the claim that people always act rationally) are thought, on grounds of empirical adequacy, to be beyond question, assumptions are recognized as less secure, as premises accepted for the time being in order to facilitate particular analyses getting off the ground.
From this perspective, which was still maintained by Hahn (and indeed by other economic theorists) at least until relatively recently, much about the contemporary state of affairs clearly is rendered intelligible. For example, it is because many of the nominal features noted by critics are regarded merely as assumptions (as opposed to axioms) that they can be interpreted as inessential to the theory project and ultimately treated as discardable. The metaphor of scaffolding is often invoked in this regard to give the impression that eventually most of these assumptions can straightforwardly be replaced or dismantled leaving the essential structure of the building intact. In turn, the particular focus upon (rational) individuals explains why so many of the prominent nominal features (hypotheses regarding beliefs, preferences, resources, etc.) relate to, i.e., are characteristics of, an individual's situation. And this focus also explains the concentration on exchange activity rather than on production or distribution. For it is only the former that can be interpreted at all plausibly as the momentary interaction of independent individuals, and so on. In short, although not every traditionally reoccurring aspect of the orthodox project is rendered intelligible by the sort of core claims asserted by Hahn (see below),(2) it is clear that a significant number of them are.
In the light of this traditional orthodox view, it would appear from the nature of recent "internal" developments within the project that something particularly unusual is up. For the project's high priests appear now to be changing significantly their sermons, or its philosophers their interpretations (or some of the most cherished ones at least). Specifically those who, like Hahn, have hitherto emphasized the virtue, indeed the necessity, of focusing more or less exclusively upon the decisions of rational individuals, upon equilibrium states, and perhaps most significantly upon simple, highly general theories appear now to be advocating, or at least allowing, other things. In place of rational individuals, we find, on the one side, stochastic, fixed (i.e., context-independent) rule following, automata, or "particles" [Canning 1990] and, on the other, groups or societies somehow coordinated in order to act as, or on average as, a single rational individual [Kirman 1989]. In place of equilibrium states, any of a variety of system solution concepts seems in order. And, most notable of all, in place of simple, general premises and the simple "elegant" theorems that they facilitate, the advice that is being given to potential "theorists" is to focus more upon complex, specific premises and to embrace the methods of computer simulation. In this, indeed, it is being predicted that "economic theory" will even need to address itself to matters of biology, psychology, sociology, and history. As Frank Hahn [1991, 50] himself has recently written:
Not only will our successors have to be far less concerned with the general (leave alone the "generic") than we have been, they will have to bring to the particular problems they will study particular histories and methods capable of dealing with the complexity of the particular, such as computer simulation. Not for them the grand unifying theory of particle physics which seems to beckon physicists. Not for them, or at least less frequently for them, the pleasures of theorems and proof. Instead the uncertain embrace of history and sociology and biology.
What, then, is going on? What explains this apparently dramatic change in the stance or perspective countenanced? The cause of it all, it would appear, is a final, if reluctant, acknowledgement that the highly generalized models of "economic theory" are too general to entail anything of interest. The result that appears to have tipped the balance derives from orthodox theory's central focus on both general equilibrium theory and game theory in particular, wherein apparently insuperable problems are acknowledged in the continuous search for possible situations of equilibrium, i.e., for situations in which all agents' plans are in some sense mutually compatible. Specifically, it has been widely observed in the highly generalized frameworks such as advocated previously by Hahn that (1) if an acceptable specification of an "economy" allows of one equilibrium it typically allows of many; and (2) even if any such equilibrium solution is found to be unique, there is no obvious mechanism in the standard orthodox specification of an "economy" to ensure that the economy could get there.(3) Clearly, if this is the case then, to the extent that orthodox "theorists" aim to predict or explain actual events, or to influence policy formulation, such results are not encouraging.
From this perspective, the change of orthodoxy's direction is largely comprehensible. In essence, the sorts of moves noted are designed to render the premises of theorizing more concrete or less general or "arbitrary." Several strands of this activity are discernable. One move is to replace the highly generalized, arbitrarily specified "assumptions" by more specific, albeit often increasingly implausible ones. In particular, the specifications of preferences and beliefs are made "stronger." Thus, we find assumptions of interpersonal utility comparisons (e.g., ideal utility functions) in welfare economics and cost benefit analysis; restrictions on the possible distributions of income in general equilibrium theory [Hildenbrand 1983]; assumptions of "similar" preference functions in "social choice theory" [Caplin and Nalebuff 1989]; conceptions of representative consumers in macroeconomics, and so on.
A second move or response is to endogenize more of the basic, previously given conditions or "assumptions" of the system. An obvious recent example in general equilibrium and game theory is that implemented under the heading of "rational expectations" of endogenizing beliefs and/or expectations, allowing the latter to be determined as an equilibrium condition of the model.(4)
A further line of response is the noted replacement of the axiom of rationality by the claim that agents follow fixed, context-independent rules. This move appears to have been accepted for an environment of random shocks and motivated by the observation that, in physical systems, the presence of randomness in the movements of atomic particles often prevents any such system from settling down in unstable equilibria or limit cycles. Indeed, a finding of statistical mechanics is that, in a system with a very large number of particles, the average position and velocity of these particles (i.e., when averaged over the entire population) are "well behaved." The point then is to seek analogous results for economic theory by considering human agents as particles, as simple rule followers, or as some such, where beliefs and preferences can be described by points in some high dimensional space. When two or more agents interact, their beliefs and perhaps preferences change - i.e., are endogenously amended - according to the rules that they follow, so that the agents move to a new point. If the system, i.e., the number of agents, is large enough, the behavior of any one agent, or so it is assumed, will be essentially random. In consequence, it is supposed that, with beliefs and actions averaged over the entire population, average outcomes will be "well behaved" and hence predictable [see Canning 1990].
A final move to consider is one taken by those orthodox theorists who appear most concerned that the project should be rendered as far as possible realistic. This is the recourse of specifying the assumptions of analysis, not arbitrarily, nor by endogenizing them as part of the system solution concept, but by incorporating, where possible, the findings of empirically based studies. Solow [1986, 24-25], for example, argues that economic theory must be modified to accommodate the fact that, in reality, individual decisions are conditioned by context-specific values and habits, etc. It is this strand of thought that gives rise to the conclusion that economists must pay more attention to matters of history, biology, and sociology. And it is the complexity of any such models as would result that encourages the prediction of Hahn and others that theorists must eventually be concerned with such procedures as computer simulation.
So what is the outcome or prospect? Is this turn to historical and biological matters, etc., a basis for expecting greater empirical success, for anticipating models of greater relevance to policy analysis? Are there now grounds for supposing that the often-noted fictitious nature of "economic theory" is about to be transcended? Do these proposed transformations or adjustments even entail that the "economic theory" project is, or will be, no more; that, in effect, it will be, or has been, transcended, as many of its critics appear to hope and applaud? I think that all of these questions must be answered in the negative. Such changes as have been occurring represent merely a latest set of maneuvers aimed at prolonging the life of an essentially misguided project. The nature of the proposed changes, I argue, are such that they necessarily leave the project still incapable of dealing with its manifest problems. They do not facilitate a theory capable of explaining real-world events or of assisting policy formulation. In elaborating now why this must be so various other issues can also be addressed. Why is it the case, after all, that in transforming the "theory" project, certain sorts of paths have been followed and not others? What, in any case, explains the previous persistent emphasis upon individualist explanation and rationality, etc.? Hahn does not elaborate. And why now condone a conception along the lines of agents as naive fixed (i.e., context-independent) rule followers? What, moreover, explains the continuing focus upon equilibrium states or similar system solution? These sorts of issues can be rendered intelligible, and the futility of the orthodox project as a whole rendered apparent. But in order to do so, it is necessary first to provide a more adequate account of the real or essential nature of the orthodox "pure theory" project than has so far been elaborated.
Economic Theory as Method
The point that has to be made is that there is one essential feature of the orthodox "pure theory" project that I have not so far emphasized, but which is at least as fundamental as any other. It is a feature, moreover, that does not turn on the choice of substantive premises. I refer to the mode of theorizing or specifically a form of explanation that is fundamental to it, one that can be referred to here as deductivism. Now to explain something is to provide an account (the explanans) whereby that something (the explanandum) is rendered intelligible. According to deductivism, as I am using the term, to be able to explain an actual event or state of affairs is to deduce a statement of it from a set of initial or boundary conditions plus universal "laws," i.e., plus universal regularities, or constant conjunctions, of the form "whenever event (type) x then event (type) y"; the explanation of such laws, as well as theories and sciences, likewise proceeds by deductive subsumption.
Now the fact that deductivism does not usually figure in the characterizations of the essential features of orthodoxy that are provided by its proponents does not mean that its centrality is ever denied. While the deducibility requirement, that the explanadum be deducible from the explanans, is more or less always transparent, the covering-law aspect, i.e., the inclusion of at least one universal law (of the form "whenever event (type) x then event (type) y") is usually met by the axioms.(5) Certainly, axioms take the "whenever this then that" structure and typically serve to relate hypothetical events or states of affairs, i.e., the sort of phenomena which, if the sort of world implied by such models were ever to hold, would, under such conditions, be objects of actual or possible experience. Many "theorists," such as Hahn, even take axioms to depend upon our actual experiences, i.e., of the social world in which we do live: "It is not that [axioms] . . . are divorced from experience or observation but rather that they mark the stage beyond which one does not seek to explain" [1984, 6]. Elsewhere, and more generally, Hahn writes:
For I do not wish to deny that there are empirical regular[ties of economic behaviour awaiting discovery. But I claim that these will be, as it were, much deeper down, more elementary and closer to the form in which axioms are postulated than are the complex, institution and history dependent "facts" of the econometrician [1984, 332].
Axioms are not plucked out of the air and far from distancing the theorists from what somewhat mysteriously is called the "real" world, they constitute claims about this world so widely agreed as to make further argument unnecessary [1985, 5].
In short, if a reliance upon the deductivist mode of explanation is not always explicit in orthodox accounts, it is not denied. Rather, its centrality and indeed universality are essentially taken for granted as obvious givens. Those critics who venture to suggest otherwise tend to be summarily dismissed. As Hahn [1985, 9] further writes:
Opponents of [economic] theory often argue that it is tautologous because it consists of logical deductions from axioms and assumptions. If one is kind to such critics one interprets them as signalling that they do not care for these axioms and these assumptions. In any case all theory in all subjects proceeds in this manner.
The orthodox attachment(6) to deductivism, in other words, is regarded as being so fundamental that no explicit attempts to defend or justify it are considered warranted. Its universality and legitimacy are accepted as being so obvious that no other form or mode of reasoning is really even contemplated. In short, its relevance, and thus any philosophical reasoning upon which it ultimately must rest, are regarded as being beyond question.
But are such matters beyond question? Is deductivism as a universally legitimate mode of reasoning secure? Are its philosophical underpinnings sound? The rarely questioned, largely unexamined, orthodox belief that this must be so leaves me unable to resist quoting from Whitehead [1926, 61]:
When you are criticizing the philosophy of an epoch, do not chiefly direct your attention to those intellectual positions which its exponents feel it necessary explicitly to defend. There will be some fundamental assumptions which adherents of all the variant systems within the epoch unconsciously presuppose. Such assumptions appear so obvious that people do not know what they are assuming because no other way of putting things has ever occurred to them. With these assumptions a certain limited number of types of philosophic systems are possible, and this group of systems constitutes the philosophy of the epoch.
Needless to say, despite the confidence with which "economic theorists" hold to the deductivist mode of explanation, it is precisely this, or the presupposition of its more or less universal legitimacy, that I intend to challenge. Specifically, it was noted above that the deductivist mode of explanation turns upon identifying regularities of the form "whenever event (type) x then event (type) y." Implicitly it is such relationships connecting actual or hypothetical events or states of affairs that, on the deductivist conception, constitute the goal or image of science. It is this general conception of science - and thus those activities and modes of reasoning that presuppose it, including the deductivist mode of explanation - that I particularly aim to criticize.(7)
Note that the foregoing deductivist characterization of orthodox theory, and in particular the observation that the enterprise depends upon the principle of empirical invariance - i.e., upon the premise that laws or significant claims in economics are, or depend upon, event regularities - is in no way undermined by the recognition that the axioms of "economic theory" have tended to be postulated at a very general level. It is true that this tendency has frequently encouraged the assertion that, in "economic theorizing," a range of outcomes or events is often possible. But it is equally the case that any such range of possibilities is not considered to be a feature of the reality being expressed; it is not ontological, but merely a result of the current state of limitations of contemporary economic theory. In consequence, the basic deductivist characterization of orthodox theory as set out above is not in this way compromised.
Let me elaborate this last claim a bit further. A significant point to emphasize here is that although the axioms of orthodox theory are formulated at a relatively high level of generality, these general claims are nevertheless usually interpreted as, or as depending upon, event regularities of our real social world - and ones that can be regarded as quite secure (I am not at this point concerned with the validity of this latter claim that axioms are empirically justified, which I dispute; I merely note its existence). Thus, an axiom of the form "firms maximize some function of profits" is, despite its apparent generality, seen as a putative real-world event regularity. Hahn, for example, writes:
The axiom that firms maximise some function of profits is stated as such because the theorist is not proposing to answer the questions why firms should do so. But it is not plucked out of the air or from dreams. It encapsulates an empirical phenomenon which many practical people and economists believe to be the nature of the capitalist. It does so at a more general level than, say, the pay-back theory but it is every bit as empirically motivated [1984, 6].
Now not only is it the case that the highly general claims of the sort in question are considered by orthodox economists to be empirically adequate, but so far it is only at this level of generality that hypotheses have been elaborated that are held to be empirically secure. That said, of course, it is not the case that orthodox analysis cannot and/or does not proceed at a lower level of generality - indeed if "economic theory" is to entail anything of relevance it has to. It is just that, because such lower level specifications, or "special theories," as abound - for example, specific profit, production, or preference functions - are considered to be more questionable in the light of empirical evidence, any deductions that are derived when employing such formulations are regarded as less secure. For this reason indeed, and as noted above, the more specific or lower level postulated event regularities utilized in orthodox theorizing are often referred to not as axioms at all, but as assumptions. Thus, Hahn asserts:
That managers have preferences is an axiom; that they are a particular form, for instance that they are linear in expected profit, is an assumption . . . [1985, 11].
That people have preferences and try to satisfy them we treat as an axiom while universal perfect competition, for instance, must count as an assumption [p. 10]
If you want more out you have to put more in, that is you must supplement the axioms with assumptions, e.g. that production functions are Cobb Douglas [p. 13].
The point, then, is that the indeterminacy that is supposedly sometimes allowed for in, and often even applauded as a virtue of, contemporary orthodox theorizing arises not because the event regularity conception of science is in anyway rejected. Hahn, as we have seen, is explicit that he does "not want to deny that there are empirical regularities of human economic behaviour awaiting discovery." Rather, "theorists" suppose that such indeterminacy as arises in orthodox models does so only because secure or empirically adequate knowledge is not yet available concerning event regularities that hold at the lower levels of abstraction required, i.e., at the level that Hahn refers to as "special theory." In consequence, if and when "theorists" praise various of their theories for not placing "single-valued restrictions" on the world, i.e., for not predicting specific economic outcomes or events, this claimed virtue must be seen for what it really is - a mere rhetorical device to mask what, on orthodoxy's own terms, is merely a state of ignorance. Hahn, indeed, more or less explicitly states as much himself:
Lastly, I want to argue that it is one of the virtues of theories derived from axioms more "fundamental" than those used in special theory, that they usually do not yield single valued restrictions on the world . . . This is a virtue because the economist is thereby restricted from claiming more than he has reasons for claiming. The axioms have summed up what one regards as pretty secure empirical knowledge. The set of outcomes which are possible is simply the reflection of our lack of knowledge. A special theory can usefully narrow them down. But our confidence in the special hypothesis is smaller than in the axioms. A claim of only one outcome should always include the proviso that given our state of knowledge there are also other possibilities [1984, 7].
In sum, the relatively high level of generality of much "economic theorizing" is quite consistent with the deductivist conception of explanation as elaborated above, including the event regularity view of laws or science on which this explanatory form depends. It is this latter conception of science, this necessary component of deductivist explanation, that I now want critically to question.
Developing an Alternative to the Deductivist Mode of Explanation
In order to pursue such matters, however, it is first useful to introduce some terminology that facilitates the discussion that follows. Specifically, I want immediately, if briefly, to elaborate notions of realism and ontology - conceptions that will figure significantly in the arguments developed. In a very general sense, any position can be designated a (philosophical) realism that asserts the existence of some disputed kind of entity. Some people are, for example, committed to the existence of quarks, black holes, class relations, utility functions, profit functions, states of economic equilibrium, etc. Clearly, on this conception we are all realists of a kind, and there are very many conceivable realisms. The question of interest in any context, then, is what kind of realism is at issue? This conception of realism is thus closely bound up with that of ontology, i.e., with enquiry into the nature of being, of existence, including the nature of the objects of study. Indeed, it is a forthright concern with ontology, a concern in particular to elaborate the broad nature of, or features of, natural and social reality that explains, in what follows, the usage of the term "realism" in labelling the perspectives distinguished.
Now the argument that I want to make here is that the conception of science elaborated above - i.e., as, or as depending upon, event regularities of the noted kind - is rooted (historically and analytically) in a philosophical perspective, one that will be referred to here as positivism. Let me emphasize, though, that I do not want to claim that contemporary orthodox economists necessarily explicitly acknowledge, or accept all or only that which is entailed by, this philosophical perspective, or even that they necessarily have any idea concerning what positivism is at all. Indeed, the main mechanism of influence may be resonance [see Lawson 1994c]. However that may be, the point to be elaborated below is that once or if the positivist perspective and its associated ontology of empirical realism (defined below) are accepted, then the "whenever this then that" conception of science, a conception that is fundamental to deductivism, necessarily follows. In consequence, clearly, if I am to argue for an alternative conception of science - one in which event regularities of the noted sort are not only insufficient, but also unnecessary for science - it is clearly beholden upon me to outline and defend an alternative philosophical perspective to positivism, one that gives rise to an alternative ontology to empirical realism that at least allows different possibilities for science to the event regularity view being contested.
First of all, what do I mean here by the perspective of positivism? I refer to a specific version of positivism that is rooted in the writings of Hume - or at least in Hume's arguments as they are most commonly interpreted. Now the version of positivism in question is first of all a theory of knowledge, its nature, and limits. Specifically, it is a claim that human knowledge takes the form of sense-experience or impression.(8) Hume encouraged this perspective with his attempted critique of any philosophical account of being, of ontology, with his denying the possibility of establishing the independent existence of things, and specifically of the operation of natural necessity. But of course, and as with Hume himself, it is never possible to dissolve completely any conception of being, of the object of study. For any theory of knowledge must assume, even if only implicitly, that the knowable world is such that it could be the object of knowledge of the required or specified sort. The Humean theory clearly presupposes an ontology of atomistic events, comprising the objects of actual or possible experience. Indeed, because reality is essentially defined as that which is given in experience then, following the practice of referring to the domain of experience as the empirical, I will, for obvious reasons, follow Bhaskar and others in referring to the perspective that arises as empirical realism.
Once it is supposed that knowable reality consists only of atomistic events of (clearly incorrigible) actual or possible experience, it follows that the only task open to science is to seek out event regularities of the sort noted above. For these are now the only sort of generalizations conceivable. In short, the "whenever this then that" formulation follows necessarily from acceptance of empirical realism, conditional only upon accepting science, or general knowledge, as a real, i.e., attainable, possibility. Indeed, the ex posteriori acceptance that the successes of science are widespread in turn entails, on this conception, that the Humean constant event conjunctions are ubiquitous.
To repeat the point emphasized above, it is, of course, not clear that the majority of economists would wish to go along with this Humean line of reasoning, even if the conclusions of such reasoning have been widely accepted. But this is not the primary issue to pursue here. The point of relevance is that once the empirical realist ontology is accepted, then for successful science to be possible (and, as noted, most people accept that it is - it exists), the "whenever event x then event y" formulation necessarily follows. In consequence, before criticizing this formulation as a scientific ideal, it is clearly necessary that I first elaborate a theory of ontology that at least allows for alternative conceptions of science and explanation, conceptions that do not reduce to the popular seeking of constant event conjunctions.
The alternative perspective that I want to maintain here can, following the contributions of Bhaskar,(9) be gathered under the general heading of transcendental realism. According to this perspective, the world investigated by science, in contrast to the claims of empirical realism, is constituted by not only events, and our experience or perception of those events, but also (irreducible) structures, mechanisms, powers, and tendencies that, although perhaps not directly observable, nevertheless underlie actual events and govern or facilitate them. Thus, not only does the autumn leaf pass to the ground, and not only do we experience it as falling, but, according to this perspective, underlying such movement are real mechanisms such as gravity. Similarly, the world is composed not only of such "surface phenomena" as skin spots, puppies turning into dogs, and relatively slow productivity growth in the United Kingdom, but also of underlying and governing structures or mechanisms such as, respectively, viruses, genetic codes, and the British system of industrial relations. In short, three domains of reality are, on this perspective, distinguished, namely the empirical (experience and impression), the non-empirical actual (events and states of affairs), and the non-actual or, metaphorically, the "deep" (structures, mechanisms, powers, and tendencies).(10)
Not only is it the case that these three domains are ontologically irreducible, according to transcendental realism, but, and crucially, they are unsynchronized, or out of phase as it were, with one another. Thus, while experiences are out of phase with events, allowing the possibility of contrasting, as well as revisions to, experiences of a given event, so events are typically unsynchronized with the mechanisms that govern them. On the latter structure/event non-correspondence, for example, autumn leaves are not in phase with the action of gravity for the reason that they are also subject to aerodynamic, thermodynamic, and other tendencies. In consequence, transcendental realism supports alternative conceptions of science and explanation. Science is the illumination and elaboration of the structures and mechanisms that govern the events of experience. And because the underlying structures are not straightforwardly manifest or "actualized" in events (or "appearances"), this goal of science can be recognized as necessary, possible, and non-trivial. Explanation, in its pure or most basic form at least, is then the providing of an account of those structures and mechanisms that are jointly responsible for producing or conditioning some identified phenomenon of interest.
To recap, according to transcendental realism, the world is composed, in part, of objects that are structured and (to use Bhaskar's term) intransitive - structured in the sense of being irreducible to events (which we may or may not experience),(11) intransitive in the sense of existing and acting independently of their identification. According to empirical realism, in contrast, the world consists only of experience and the events of (and which, in effect, are constituted by) experience. From these alternative theories of ontology flow different conceptions of science and explanation. Specifically, from the perspective of empirical realism, science and explanation are necessarily restricted to or rooted in the seeking out of constant conjunctions of events (including probabilistically formulated accounts of them), whereas from the perspective of transcendental realism, the primary scientific and explanatory concern is not with events (or their hoped for constant conjunctions) at all, but with identifying the structures and mechanisms that underlie and govern them.
Note, in particular, how these conceptions entail contrasting modes of inference. Empirical realism, with its focus upon generalizations about conjunctions of events, has been plagued with endless worries concerning the relative advantages and limitations of methods of induction versus those of deduction. Indeed, some version of this opposition appears to be the extent of most methodological discussions in economics [see, e.g., Pheby 1988]. It is important to recognize, then, that the essential mode of inference presupposed by the transcendental realist perspective is neither induction nor deduction but one that can be styled retroduction or "as if" reasoning or, as in the tradition of institutionalist writings, abduction [e.g., Peirce 1867].(12) It consists in the movement (on the basis of analogy and metaphor, among other things) from a conception of some phenomenon of interest to a conception of some totally different type of thing, mechanism, structure, or condition that is responsible for the given phenomenon. If deduction is, for example, to move from the general claim that "all polar bears are white" to the particular inference that the next one seen will be white, and induction is to move from the particular observation of numerous white polar bears to the general claim that "all polar bears are white," retroductive or abductive reasoning is to move from the observation of numerous white polar bears to a theory of a mechanism intrinsic (and perhaps also extrinsic) to polar bears that disposes them to being white. It is the movement, paradigmatically, from a "surface phenomenon" to some "deeper" causal thing.
So far, two ontological theories, transcendental and empirical realism, have been elaborated and contrasted, and attention has been drawn to their differing implications regarding the nature of science and explanation. As yet, I have not argued (as I will) for accepting transcendental realism as the more satisfactory theory, nor have I given reason to suppose that even if transcendental realism is accepted, constant conjunctions of events will not be a pervasive eventuality. These two issues can now be dealt with together.
It is essential to distinguish first the conditions in which experience is significant for science. Specifically under what conditions, if any, are the sought after constant conjunctions of Humean science actually found to occur? Consider first the situation in the natural sciences. Of fundamental significance here are two observations whose relevance has been well brought out by Bhaskar. The first of these is that, outside astronomy at least, most of the constant conjunctions that are held as significant to science only in fact occur under the restricted conditions of experimental control - i.e., they are not typically spontaneous occurrences but are a product of human intentional practice. The second observation is that the results or "laws" supported in controlled experimental activity are nevertheless frequently successfully applied outside of the experimental situation.
These observations raise certain problems for post-Humean accounts that tie laws and ultimately explanations to constant conjunctions of events. For, if scientific laws, or significant results, only occur in such restricted conditions as experimental setups, then this bears the (rather inhibiting) implication that science and its results, far from being universal, are effectively fenced off from most of the goings on in the world. In other words, most of the accepted results of science are not of the form "whenever event x then event y always follows" after all, but are of the form "whenever event x then event y always follows, as long as conditions e hold," where conditions e typically amount to a specification of the experimental situation. This also bears the rather counterintuitive implication that any actual regularity of events that a law of nature supposedly denotes does not, in fact, generally occur independently of human intervention. But, in addition to such problems and at least as seriously, the constant conjunction view of laws leaves the question of what governs events outside of experimental situations not only unanswered, but completely unaddressed. And, in doing so, it also leaves the observation that experimentally obtained results are successfully applied outside experimental situations without any valid explanation.
In order to render Bhaskar's observations intelligible, it is necessary to abandon the view that the generalizations of nature consist of event regularities and to accept instead the transcendental realist account of the objects of the world, including of science, as intransitive and structured. That is, experimental activity and results, and the applications of experimentally determined knowledge outside of experimental situations, can be accommodated only through invoking the transcendental realist ontology of generative structures, powers, mechanisms, and necessary relations, etc., that lie behind and govern the flux of events in an essentially open world. The fall of an autumn leaf, for example, and as already noted, does not typically conform to an empirical regularity precisely because it is governed, in complex ways, by the actions of different juxtaposed and counteracting mechanisms. Not only is the path of the leaf governed by gravitational pull, but also by aerodynamic, thermodynamic, inertial, and other tendencies. On this transcendental realist view, then, experimental activity can be explained as an attempt to intervene in order to insulate a particular mechanism of interest by holding off all other potentially counteracting mechanisms. For it is only in such a situation that a one-to-one correspondence can be obtained between the way a mechanism acts and the events that eventually ensue. In other words, on this view, experimental activity can be rendered intelligible not as creating the rare situation in which an empirical law is put into effect, but as intervening in order to bring about, to engineer, those special circumstance under which a non-empirical law, a power, a tendency, or a way of acting of some mechanism, etc., can be empirically identified. The law in question is always operative - if the triggering conditions hold, the mechanism is activated whatever else is going on. On this transcendental realist view, for example, a leaf is subject to the gravitational tendency even as I hold it in the palm of my hand. Through this sort of reasoning, then, transcendental realism can render intelligible the application of scientific knowledge outside of experimental situations. The context or milieu under which any mechanism will be operative is irrelevant to the law's specification. Once activated, the mechanism is operative whatever empirical pattern ensues.
In short, although the traditional conception of science upon which deductivist explanation rests necessitates results formulated as constant conjunctions of events, in practice such event regularities that have been reported have been restricted in the main to situations of experimental control. Transcendental realism, unlike empirical realism, can render this situation intelligible. And it follows from the transcendental realist perspective that the traditional Humean conception rests upon an inadequate analysis, and illegitimate generalization, of what emerges as a special case - a closed system situation wherein a single and stable (set of) aspect(s) or mechanism(s) is physically isolated and thereby empirically identified.
It will not have gone unnoticed that the discussion on this point has referred to the situation in the natural sciences. However, it is not difficult to see that the transcendental realist argument carries over to the social realm. A first point to note here is the observation that few if any interesting or significant event regularities of the sought after kind have yet to be turned up in the social realm. A second point is that many economists appear to share the intuition that human agents possess the capacity of real choice even if these same economists are unable to reconcile this insight with their understanding of scientific explanation. Now, if choice is real, any agent could always have acted differently than he or she in fact did. And a necessary condition for this, clearly, is that the world, social as well as natural, is open in the sense that events really could have been different. Put differently, if under conditions x an agent chose in fact to do y, it is the case that this same agent could really instead have not done y. Choice to repeat presupposes that the world is open and actual events need not have been.
But the possibility of choice not only presupposes that events could have been different, it also entails that agents have some conception of what they are doing and wanting to achieve in their activity. That is, if choice is real, then human actions must be intentional under some description. Now all agency (whether human and intentional or otherwise) is inherently transformational. Its exercise thus necessitates that there are real material causes that proceed it. Just as acid cannot corrode without materials to break down, nor a bird construct a nest without materials so to fashion, so human activity cannot take place without means, media, and resources of some kind - conditions that, through human agency, come to be transformed. The intentionality aspect of human agency is, in turn, bound up with knowledgeability. For agents must have some knowledge at least of the conditions that render their intended acts, when they are, as feasible. In turn again, knowledge presupposes sufficient endurability in the objects of knowledge to facilitate their coming to be known. Now if event regularities, or at least significant ones, do not, as widely reported, generally occur (or at least are yet to be uncovered) in the social realm, then the enduring objects of knowledge that condition actual human practices must then lie at a different level - at that of structures which govern, but which are irreducible to, events, including human activities, of experience.
But are there any real material causes or facilitating structures that can be said to be clearly social? If we accept as criterial for the social the property of depending upon human agency, then it is easy to see that the answer is yes. Items such as (societal) rules, relations, and positions clearly both depend on human agency as well as condition our every day (physical) activities. The human (intentional) activities of speaking, writing, driving on public roadways, cashing checks, playing games, giving lectures, etc., would be impossible without such social material conditions as rules of grammar, the highway code, banking systems, rules of play, teacher-student relationships, etc. If it is the dependency of such structures upon human agency that marks them out as being social, it is their ability, in turn, to make a difference in (to enable as well as to constrain) physical states, or actions, that (as with non-perceivable objects of the natural realm such as gravitational and magnetic fields) establishes that they are real.
At a very general level, then, the transcendental realist ontology carries over to the social realm. Now it was observed that in the natural realm constant event conjunctions appear mainly to be occurring (at least outside astronomy) only in conditions of experimental control. Of course, the feasibility of experimental control in the social realm is rather limited. Are there, then, any grounds for expecting significant constant event conjunctions to be a feature of social realm? If natural and social realms are similar in that both are characterized by structures underlying the events of experience, they are dissimilar in that social, unlike natural, structures depend for their existence on human agency. Human agency and social structure then presuppose each other. Neither can be reduced to, identified with, or explained completely in terms of, for each requires, the other. The point of relevance here, of course, is that because social structure is human agent-dependent it is only ever manifest in human activity. Thus, given the open nature of human action, the fact that any agent could always have acted otherwise, it follows that social structure can only ever be present in an open system. Thus, any economic laws must be interpreted as tendencies that are manifested usually only in partial fashion and very rarely - usually in cases where they are consciously brought about (e.g., the occurrence of annual holidays) - as strict event regularities, so that the Humean project in its economic guise must, as a general approach, be recognized as misguided.
In this section, I have argued that there are empirical and theoretical grounds for questioning the traditional (Humean/positivist) ideal for science as the elaboration of universal constant conjunctions of events. It is the misguided adherence to this conception of science, or the unreflective and undifferentiated wielding of techniques and methods that presuppose it, including accepting the universal applicability of the deductivist form of explanation, that constitutes the fundamental problem in the economic scientific project. In the light of this "diagnosis," the argued for cure, or solution, is that provided by the transcendental realist perspective whereby the primary aim of science and explanation is interpreted not as the production of constant event conjunctions, but as the identification and illumination of the structures and mechanisms that underlie the phenomena of experience and govern and produce them. Clearly, this transcendental realist "solution" requires some further elaboration, and this is briefly provided in the final section. First, however, some loose ends need tidying up. Specifically, from the perspective elaborated above, the so-far unaccounted for premises and orientations traditionally adopted by orthodox theorists, as well as the sorts of moves recently countenanced, can now be easily explained.
Explaining the Unaccounted for Features of Contemporary Economic Theory
We have seen that, although orthodox economists appear not to doubt the universal legitimacy of the deductivist mode of explanation, the conditions of the latter's justified application are in fact highly circumscribed. This is true even within the natural realm; in the social realm, the closed systems on which the method depends appear not to occur at all in a manner that is significant to science.
What, then, are the prospects for those typically nonorthodox economists who fail to appreciate (or refuse to accept) this situation? Specifically, what options exist for those who continue to act on the supposition that deductivism constitutes a universally applicable mode of explanation? As we will now see, it is through addressing this question that those features of "economic theory" that have so far been left unaccounted for can finally and straightforwardly be rendered intelligible.
The presumed universal applicability of the deductivist mode of explanation ultimately entails an adherence to a metaphysical thesis that can be referred to here as regularity determinism. According to this thesis, for every economic event or state of affairs y, there exists a set of events or conditions x1,x2...[x.sub.n], such that y and x1, x2,...[x.sub.n] are regularly conjoined under some (set of) formulation(s).
The interesting question to pose is what sort of constraints does such an orientation place upon, or at least encourage in, theory construction? All economic theorizing is couched in terms of economic individuals of some sort - whether consumers, employers, households, firms, banks, trade unions, or whatever. One obvious condition to seek to satisfy then is that any individual (or set of individuals) of analysis is so intrinsically constituted or organized that under any repeated, completely specified, or isolated set of states or conditions of action x1, x2...[x.sub.n], the same outcome y is guaranteed to follow. This can be referred to as the intrinsic condition for a closure.
Now any theoretical specification that is orientated toward satisfying the intrinsic condition so understood can be seen to entail two essential aspects. Prima facie all economic individuals are characterized by internal structure and complexity and thus may behave differently in identical extrinsic conditions by virtue of different internal states. For example, how any given person may perform some specified act depends upon whether he or she is alert or sleepy, optimistic or pessimistic, tense or relaxed, etc. A first obvious restriction to impose, or to seek to guarantee in the course of economic reasoning, then, is that of intrinsic constancy: that the internal, or intrinsic, state or structure of any individual of analysis be constant. Notice that if the individual or unit of analysis is not a person (although in orthodox theory it usually is), but an entity like a workplace wherein numerous people are situated, then the relevant condition to seek to satisfy in any theoretical specification is the constancy of the intrinsic structure of not only the workplace (or whatever), but also of each situated human individual.
The second obvious requirement of any theoretical construction that is to satisfy the intrinsic condition for a closure is reducibility: that the overall outcome event be reducible to the system conditions or states obtained. The point here is that even if the intrinsic structure of each and every relevant individual of analysis is so specified as to be constant, this need not yet guarantee that, given identical conditions of action or operation, the one same outcome will materialize - indeterminacy of outcomes remains a possibility. In other words, in order for the intrinsic closure condition to be satisfied, it is necessary that the intrinsic organization of each individual of analysis be so specified that, for any given intrinsic structure (or at least for the one assumed to be constant throughout the time-space extension of the analysis), only one outcome in any given set of specified conditions is possible. In short, the intrinsic organization of any individual must be such that each outcome is deducible from, or reducible to (in the sense of being expressible in terms or as a constant function of), the conditions or given states of the system.
Clearly, both of these aspects, i.e., intrinsic constancy and reducibility, are automatically satisfied if any and every relevant individual is characterized atomistically - i.e., as in effect lacking intrinsic structure. Then each reaction is only and always but a passive response to external, impinging forces or stimuli. As we will see below, such advantages or attractions of atomistic conceptions do indeed render intelligible fundamental aspects of substantive developments within the mainstream project.
Before considering such matters, however, it must be registered that a satisfaction of the intrinsic condition does not in itself guarantee a closure. A satisfaction of the intrinsic condition ensures that if the elaborated set of conditions or states x1 ... [x.sub.n] entails a complete list of potential "influences," then the same actuality y must always follow. A potential problem for deductivist attempts to explain any actual outcome arises, however, with the possibility of other conditions, including possibly novel factors that are not included in the specified set of conditions x1, x2 ... [x.sub.n], but which, nevertheless, are capable of influencing the actual outcome in question y. For deductivist explanation of actual events to be guaranteed, then, such influences must be either internalized within the system or shown to be constant in their action.(13) To complement the intrinsic condition for a closure, then, we now have a second extrinsic closure condition for mainstream economists to build into their models: that only the explicitly elaborated conditions x1, x2 ... [x.sub.n] have a systematic, nonconstant influence on the outcome event y in question.
The most straightforward form of specification that guarantees that this condition is met is one wherein the relevant individuals of analysis are, in their actions, physically isolated from all the (nonconstant) conditions not explicitly elaborated. In other words, while the desire to satisfy the intrinsic closure condition encourages conceptions of individuals as atomistic in character, a similar need to satisfy the extrinsic condition encourages a conception of such individuals as acting in isolation. In such a scenario, of course, the overall outcome of a system composed of numerous such individuals can then be determined merely by adding their separate responses together.
It is important to be clear what it is exactly that I am arguing here. Of course, even if regularity determinism is an untenable metaphysical thesis, a restricted event regularity may by chance come about. That is, even accepting that all individuals are internally structured and complex and act in an open, structured, and changing world, it is still theoretically possible that an apparently stable relationship between conditions and some outcome event is observed, at least over some restricted region of space-time. Such a possibility, if on the face of it highly unlikely in any significant context, is not theoretically refutable. Those who emphasize the theoretical nature of the orthodox enterprise, however, require much more than this. Their desire to "explain," i.e., to render intelligible, events in some theoretical manner, combined with a belief in the universal relevence of the deductivist structure of explanation, sets them off looking for or deriving theoretical constructions that, among other things, have conditions to guarantee the "whenever this event then that event" formulation built into them. While there will not be a unique set of sufficient conditions of this sort, the intrinsic and extrinsic conditions just outlined, by focusing directly upon both the nature of individuals and also their conditions of action, do seem the most straightforward, most obvious, contenders. And the most obvious conceptions to pursue in the endeavor to satisfy these two conditions, I have suggested, are those of atomistic individuals acting in conditions of relative isolation. The central point I want to argue in the light of all this is that the various substantive moves of orthodox theorists that have so far been left unaccounted for above are easily interpretable from this perspective, i.e., just as specific attempts to provide accounts of atomistic individuals acting in relatively isolated conditions in order to ensure that the identified closure conditions are met. Of course, economic theorists do not explicitly elaborate their project in these terms, and so their strategies must be reconstructed. But certainly their traditional premises and recently revised outlooks are easily rendered intelligible from this perspective now set out.
Let us see more closely how this works. Consider first the intrinsic closure condition. How, if at all, is this condition typically addressed in contemporary economics? What moves are made, if any, in orthodox theory to ensure that any individual of analysis is of a kind to always act in the same way in identical conditions? It must by now be obvious that this has usually been achieved in contemporary economics by (1) making human individuals more or less the only unit of analysis; (2) characterizing human nature in such a way that such individuals are only and always rational, i.e., economic optimizers (thus achieving intrinsic constancy), where (3) the perceived objective (utility, preference, profit, etc.) function to be optimized is fixed and permits of a (locally at least) unique optimum and so single course of action (ensuring outcome reducibility). It is clearly the implicitly or tacitly recognized need to provide some basis for supposing the satisfaction of identified intrinsic closure condition that epistemological individualism and the rationality "axiom" (and associated types of assumptions) have been so stubbornly maintained.
However, from the perspective now opened up, if the latter axiom at least is recognized as an obvious one to retain, it is seen to be not after all necessitated. Assumptions such as made by Canning  that agents always follow simple, fixed, context-independent rules will do equally well. All that is required is that human agents rebound in atom-like fashion to external influences. Indeed, in this light it is not surprising that Canning should attempt to model human agents by drawing directly upon the physics of atomistic particles. Conditional upon deductivism (as opposed to the substantive or material forms of the deductively structured orthodox formulations) constituting all that is essential to the orthodox project, any such move as Canning's is quite intelligible. For at the end of the day, all that is required on this front is that in mainstream theoretical specifications any agent of analysis is constrained to act in identical ways in identical conditions.
The extrinsic condition for a closure is clearly addressed by treating each agent as acting within an isolated and typically highly limited fixed set of conditions. Thus, the focus is unusually upon buyers and sellers interacting in a quite impersonal way in their attempts to satisfy their preferences in situations more or less completely specified by given endowments and prices. Production, too, tends to be reduced to the activities of relatively isolated "entrepreneurs" taking the form of exchanging inputs for outputs on the basis of some fixed technique of production or some such [see Hahn 1984, 80]. Significantly, when a collection of agents and their interactions is the focus, the setup is usually conceived of or presented as an economy or game as a whole, i.e., in its entirety, as some totally described and physically isolated system. As Kapp [1976, 210] observes in his summary assessment of neoclassical economics: "Concrete economic systems or processes are . . . believed to be adequately represented as isolated, self-contained and self-sustaining, closed mechanical processes with definable boundaries."
If it is traditionally the case that, in orthodox theorizing, individual agents have been specified as acting in a highly restricted set of supposedly isolated conditions, one obvious response to the failure of the project is to explore alternative, more complex formulations. In other words, in the place of overly simple formulations aimed at satisfying the extrinsic condition, we now see the latter being addressed through certain history and context-specific features of each agent's (nevertheless still isolated) situation being introduced into the formal specification. Perhaps the most significant point here is that until recently the possibility of deductivist analysis has rested upon the sought-after event regularities consisting in or resting upon a not too large set of conditioning factors in order to reduce the complexity of the analysis. In recent years, however, the accelerated rate of availability of computer resources has entailed that more complex formulations can be addressed. This, in part, explains why at this point in time, the possibility of more complicated, historical, biological, and sociological, etc., specifications or conditions is at least being openly and positively contemplated.
What can be said, though, about that one feature of orthodox theorizing not yet adequately addressed: the insistence that some sort of equilibrium notion be considered as a fundamental component of analysis? How can this orientation be rendered intelligible from the perspective so far elaborated? The answer, of course, turns on the observation that, given the dependence of the deductivist project upon a set of postulated behavioral regularities or formal equations, a question that arises naturally is whether there exists a set of parameter values or conditions that renders such equations or formulations mutually consistent? The notion of equilibrium serves precisely to capture any such condition. Thus, in its most common and weakest form, an equilibrium is interpreted just as a state in which all actions and/or plans and/or expectations, etc., (depending upon the setup) are mutually compatible, i.e., a state in which the system equations are mutually consistent.
From this perspective, then, it is not surprising that, with the repeated failure of the project to enlighten, and with various novel specifications (consistent with the intrinsic and extrinsic closure conditions setout) being explored, a range of novel "solution concepts" or types of equilibrium are being continually elaborated. For clearly different specifications are required to fit the specific details of the model (including game) in question. And given that, at the end of the day, all methods or formulations that presuppose or entail a closure are bound to be unenlightening of a social world insusceptible to closure, the one question that will continually attract attention, and when answered provide some semblance of an achievement, is whether it is possible to demonstrate that the formal system at hand is at least internally coherent. From this perspective, the orthodox attention to equilibrium notions, or more generally "solution concepts," if largely besides the point, is nevertheless not surprising.
I have argued that the various features of the economic theory project traditionally held as essential, along with the more recently suggested revisions of perspective, can easily be rendered intelligible once the unquestioning, uncritical, orthodox adherence to the deductivist mode of reasoning is acknowledged. Given the flexibility in that project ultimately revealed at the level of substantive premises, including axioms as well as assumptions, and the lack of flexibility apparent at the level of its mode of explanation, I suggest that adherence to deductivism in the context of attempting to render intelligible economic (or generally social) phenomena be recognized not merely as fundamental to, but actually constitutive of, the orthodox "economic theory" project. In other words, in the absence of any obvious alternative hypothesis that appears capable of rendering intelligible the nominal or manifest characteristics of orthodox contributions examined throughout, I suggest that the mainstream "theorizing" project be recognized just as the uncritical adherence to or persistence with the deductivist mode of explanation in the context of economics.
If this assessment is correct, moreover, it may be possible to anticipate the direction of future tendencies encouraged by those economists within the orthodoxy who are most concerned to render the orthodox project more realistic or at least to "explain" actual events. Specifically, concern about the continued explanatory failure of the project is likely to induce one or both of two types of general response: either to suppose that the extrinsic condition for a closure has been inadequately addressed, with the obvious conclusion that more (possibly context specific) influences must be explicitly elaborated, or to suppose that the intrinsic closure condition has yet to be adequately addressed, with the obvious inference that the individuals of the analysis have yet to be given a sufficiently atomistic or micro formulation.
In other words, the maintenance of the deductivist project, combined with the goal of being realistic, are likely to set in train, as tendencies, two regresses - the limit points of which are systems so large that they exclude nothing couched in terms of single individuals so small that they include nothing. In short, they set us searching in the wrong direction. The recent recommended turn to computer simulations, indeed, appears to be an early manifestation of this tendency. The alternative, efficacious solution, of course, is to posit an autonomous, structured agent somewhere between the universe and the atom that acts in an equally structured and fundamentally open and changing world. In other words, the point is to recognize the superior adequacy of the perspective provided by transcendental realism and accordingly accept its reformulation of the possibilities and tasks of economics as science and explanation. It is this issue, in brief, to which I must now return.
Transcendental Realism and Economic Explanation
It has been argued that there are grounds for questioning the relevance to economics of the traditional Humean ideal or image of science as the elaboration of universal constant conjunctions of events, at least as a guiding conception for all contexts and occasions. What though can be said about the transcendental realist-grounded conception of science, and specifically economics, that I am proposing in its place? This, needless to say, is too big an issue to elaborate in very great detail here [but see Lawson 1989a, 1989b, 1992, 1994a, 1994b]. However, there are certain basic matters that, at least in some general way, must now be addressed.
The first point to consider is an obvious one. I have argued, along with Kapp, that the world is open in the sense that universal constant conjunctions of events are a rare event even in natural science and more so in the social realm. But if this is so, how can economic mechanisms, etc., be identified? The point that warrants emphasis is that just because universal constant conjunctions of the form "whenever event x then event y" are unlikely to be pervasive, it does not follow that the only alternative is an inchoate random flux. These two possibilities - strict(14) event regularities or a completely nonsystematic flux - merely constitute the polar extremes of a potential continuum. Although the social world is open, certain mechanisms can come to dominate others over restricted regions of time-space, giving rise to rough and ready generalities or partial regularities, holding to such a degree that prima facie an explanation is called for. Thus, just as autumn leaves do still fall to the ground much of the time, so women are concentrated in secondary sectors of labor markets, and productivity growth in the United Kingdom over the last century has frequently been slower than that of otherwise comparable industrial countries. Such "stylized facts" can serve both to initiate investigation and to assess the relative explanatory power of competing hypothesis that may, in due course, be constructed.
A consequence of the possibility of a continuum of outcomes (i.e., ranging from closed systems of constant conjunctions of events to an inchoate random flux) is that economic explanation (i.e., the account of it that can be sustained via transcendental realist analysis) must, in effect, be divided into two distinct moments or separate modes of activity. The discussion above, in fact, has been referring to a mode of inference that should really be termed pure or abstract explanation - the identification of underlying structures and mechanisms. A necessary condition for this activity is that certain mechanisms do at some time and place dominate others, so that rough or partial event regularities are discernable. However, it is equally clear that a second mode of inference, which is appropriately termed applied or concrete explanation, is also called for and in economics must prove significant. For, to the extent that the concrete phenomena of experience are relatively unique or novel, being conjunctures or resultants of numerous countervailing tendencies, their explanation entails drawing upon antecedently established knowledge of relatively enduring structures and mechanisms (rather than revealing them) and investigating ex post the manner of their joint articulation in the production of the novel event in question.
Nothing yet has been said explicitly about the events of the social realm, the social phenomena to be explained. What then are these? What form do they take? It follows from the outline above, of course, that social structures govern, so that the events, or event-analogues, of the social realm are precisely social activities. On the transcendental realist conception, then, the object of economics is primarily to identify the structures governing, the conditions surrounding, facilitating as well as being transformed through, some human activities of interest. Social explanation entails providing an understanding of certain practices and activities of interest, that is, identifying and understanding the unacknowledged conditions of these practices, their unconscious motivations, the tacit skills drawn upon, and unintended consequences. While society and economy are the skilled accomplishments of active agents, they remain to a degree opaque to the individuals upon whose activities they depend. The task of economics, then, is to describe all that must be going on (whether or not adequately comprehended by the agents involved) for some manifest social phenomenon, for some set of practices or activities, to be possible.
As an example, consider the productivity growth in the United Kingdom over the last 100 years, which, as noted, has quite frequently been less than that of otherwise comparable industrial countries. The activities involved here include those bearing upon the rate of introducing new techniques of production, levels of staffing of machinery, the ability of an organization to respond flexibly to change, and so forth. In Britain, however, there are definite differences in the structures drawn upon in these activities compared with, say, the countries of the continent of Europe. For in Britain, unlike other countries, the legality of collective worker organization was recognized before the introduction of mass-production techniques, when work was organized upon a craft-system basis. In consequence, norms, relationships, and practices, etc., associated with this highly localized and sectional organization of work became built into the British industrial system. And once legitimized, this craft-oriented basis has tended to be reproduced. For the United Kingdom, unlike many other countries in the last 100 years, has not experienced fascist government with its associated prohibition on trade union organization and activities, and it has not been defeated in world wars, while, until relatively recently at least, UK industry has had the opportunity to turn to protected imperial markets to offset the otherwise revolutionizing forces of (increasing) worldwide competition. For such reasons, among others, the United kingdom's craft-oriented practices, localized system of collective bargaining, relatively strict demarcation lines, and other such structures that can work to inhibit flexibility, have been subject to less (pressure for) transformation than would otherwise have been expected.
Needless to say, this summary is much too brief to provide more than an overview. (For a more detailed account of all this, see Kilpatrick and Lawson 1980). Nor does the transcendental realist account stand or fall according to the accuracy of this (or indeed of any other) substantive analysis. All that is intended is an indication of the sort of reasoning that is legitimized and supported. Of course, once a hypothesis, or a set of competing hypotheses, of mechanisms is formed, such hypotheses can then be further assessed or selected among on the basis of the empirical support for any further contrasting claims that are entailed. The hypothesis entertained above is that, where bargaining structures are more decentralized, there is an inbuilt potentiality for productivity growth to be that much slower. Of course, any such potential may or may not be continually activated, and if or when it is activated, it may rarely be precisely actualized because of countervailing tendencies. Even so, where countervailing factors are comparable across regions (or whatever is) being contrasted, or where they ebb or flow in their intensity, etc., it is to be expected that, if the hypothesis in question is correct, the tendencies of interest will, on a (perhaps significant) number of occasions, shine through. In consequence, the hypothesis has implications that can be empirically assessed (albeit not according to any strict set of a priori rules, procedures, or criteria) both within a region like the United Kingdom, as well as across countries and, or over, time [again see Kilpatrick and Lawson 1980] and so on.
The preceding discussion is, I think, sufficient to justify the optimistic assessment made at the outset that the perspective opened up by realist analysis and critique, though debilitating of contemporary "economic theory" as currently understood, is empowering in its implications generally for economics as science. It has been found that the traditional Humean conception of science underpinning deductivism and including orthodox "pure theory," etc., presupposes rather special conditions that are unlikely to arise often in the social realm. The failure to realize this does, of course, lead to a pessimistic outlook and no doubt explains the ex posteriori hesitancy of some orthodox theorists even to acknowledge that, or how, their goal is the illumination of reality at all. However, once the transcendental realist perspective is accepted, and it is recognized that the social realm, just like the natural one, is structured and intransitive, then it follows that economics, in principle, can be just as illuminating and revelatory as any other science.
As a final point, something should be said about the further optimistic claim made at the outset regarding the possibilities for policy analysis…