Download Sociology Theory and Rational Choice Theory and more Study notes Sociological Theories in PDF only on Docsity!
Sociological Theory and Rational
Choice Theory
PETER ABELL
We may conveniently start with the words of Weber: "Sociology.….is a science which attempts the interpretive understanding of social action in order thereby to arrive at a causal explanation of its course and effects." Furthermore, "action is social when it takes account of the behaviour of others and is thereby orientated in its course" (Weber, 1947).
Rational choice or action theory may be understood as one possible interpretation of Weber's program, though it must be said a rather special one in that it invites us to adopt the least complex conception of social action that we can analytically get away with in arriving at a "causal explanation of its course and effects." It thus departs from many post-Weberian (and for that matter pre-Weberian) theoretical traditions - particularly those of a phenomenological persuasion - where the purpose seems to tilt in an entirely opposing direction, namely to find ways of conceiving (social) actions which are locally detailed and complex. Why it should be that rational choice theory adopts this heterodox standpoint will detain us below.
Even with all its manifest limitations rational choice theory has arguably proven to be the most successful theoretical framework in those social sciences which, like sociology, deal with explanations of macro or system-level phenomena and we should accordingly be somewhat circumspect before, following the persuasive voice of Parsons (1937), roundly rejecting it. Although we shall examine the assumptions of rational choice theory rather closely later on, a broad working definition (probably a little too broad for some) will help to get us started. In its broadest interpretation rational choice theory invites us to understand individual actors (which in specified circumstances may be collectivities of one sort or another) as acting, or more likely interacting, in a manner such that they can be deemed to be doing the best they can for themselves, given their objectives, resources, and circumstances, as they see them.
It is perhaps appropriate here to insert a word of caution. In my view it would be dangerously partisan to argue for an exclusive reliance upon any single theoretical framework in sociology - the shape of the discipline is just too diverse for this to ring true. Indeed, all I wish to urge here is that rational choice theory may be the least bad theoretical framework at our disposal, for solving certain theoretical puzzles (Abell, 1992). This, as I hope to demonstrate, derives partly from its own strengths (albeit limited ones) but also from the weaknesses of its competitors. In this latter respect it will be up to the reader to compare the arguments contained here, with those in other chapters in this volume.
I shall start by briefly introducing some past attempts to promote rational choice theory as a serious foundational framework for sociological theory. I have
intentionally kept this brief because overwhelmingly the most important developments are taking place currently and they bear only a passing relationship to what has gone before. Indeed, we have in the last decade or so witnessed a modest revolution in rational choice theory. Whether it will come to have any lasting impact upon the future direction of sociological theory only time will tell, but the reader wishing to become acquainted with the mainstream developments might like to refer to the early chapters in Coleman's Foundations of Social Theory (1990); also, for both advocacy and critique, to Coleman and Fararo's edited volume Rational Choice Theory (1992) and to Abell's Rational Choice Theory (1991). Cook and Levi's edited volume The Limits of Rationality (1990) is also a useful read, as is almost anything written by Elster though his Nuts and Bolts (1989a) must be given pride of place as an introductory text strongly coloured by a rational choice theory perspective.
Those readers who prefer their theory diluted by sensitive qualitative empirical research will find Diego Gambetta's The Sicilian Mafia (1993) irresistible and Michael Hechter's Principles of Group Solidarity (1987) persuasive. The journal, with a sociological orientation, which is now established as the leading one in this field is Rationality and Society, published by Sage, though the Journal of Mathematical Sociology also contains many articles informed by a rational choice perspective.
It is, however, important to appreciate that developments in adjacent disciplines may well in the longer term prove of equal importance: in this regard the reader might like to refer to Ordeshook's Game Theory and Political Theory (1986) and Kreps's Game Theory and Economic Modelling (1990). The latter provides a gentle introduction to a burgeoning technical literature which has taken economics by storm but which addresses many ideas, like trust and reputation, which were previously thought to be the exclusive province of the sociologist.
This chapter is designed to afford the uninitiated reader an overview of the issues, which arise when adopting a rational choice theory perspective. By its very nature, though, it can only touch upon many of these. Furthermore, a number of the issues are quite technical; 1 have, however, chosen not to present things at all technically but, rather, where appropriate, to refer the reader to additional literature. It must be said, though, that the best way to gain an appreciation of the analytical potential of rational choice theory is to work through a few examples where it has proven to be successful. In this respect Rationality in Action (1990) edited by Paul K. Moser may prove useful.
The Foundations of Rational Choice Theory in Contemporary Sociology
Although one can easily find the roots of the very conception of a feasible systematic social inquiry in close association with Enlightenment precepts of rational conduct, it is to the nineteenth century that we must look for ideas, which have a strong contemporary resonance. Utilitarianism and Marxism both, in their own ways, make use of assumptions about individual rational actors, though Marxism has only recently regained a full recognition of its debts in this regard (Roemer, 1982; Elster, 1985). It is, nevertheless, to Pareto that we must look for a systematic analysis of inter-actor exchange, an idea, which now lies at the centre
about the dynamics of influence and propositions about the structure of small groups." He introduces the ideas of the benefits and costs of alternative courses of action and declining marginal effects, along with a conception of balance or equilibrium in exchange. Peter Blaus book, Exchange and Power in Social Life, followed shortly afterwards, in 1964, and though it suffered from a number of technical deficiencies (see, for instance, Heath, 1968) it has taken on the mantle of a minor classic in the field. Ever a since, exchange theory has attracted its adherents and has become one of the few areas in sociological theory which is at the same time broadly cumulative and technically well founded. In this latter respect the association which is ever increasingly engineered with network analysis is of prime importance, as is the movement into the laboratory with carefully controlled experiments. Indeed, there are many who claim a broad allegiance to rational choice theory who now see the combination of networks and rational choice theory as the most promising avenue ahead for systematic sociological theory. Meanwhile, the juxtaposition of exchange and rational choice theories finds its contemporary apotheosis in Coleman's Foundations of Social Theory (1990) Mancur Olson's The Logic of Collective Action (1965) was published a year after Blaus book and has proved to be one of the most seminal works in modern sociological theory. It has spawned a copious literature about collective action, much, of which is based upon an analysis of the prisoner's dilemma game. This literature has brought the issue of public goods, externalities, and club goods, formerly the sole preserve of economists, into the centre of sociological theory (Coleman, 1990).
Although systematic claims for game theory as the theory of social interaction are relatively recent (Raub and Weesle, 1992; Abell, 1993), the use of a game theoretic framework in theoretical analysis is increasingly common. An early advocate was Fararo (Mathematical Sociology (1973)) and we can now, with the benefit of hindsight, see Schelling's two books Strategy of Conflict (1960) and Micromotives and Macrobehavior (1978) as important, and conspicuously inventive, influences. Ullmann-Marglit's The Emergence of Norms (1977) brought the game theoretic analysis of the genesis of norms into the picture. The idea which is probably having the most profound impact upon sociological theory, though, is the shift from an emphasis upon one-shot games (e.g. a one-off prisoner's dilemma) to repeatedly iterated games. In this respect Axelrod's The Evolution of Cooperation (1984) is the turning point, but here we increasingly find a merging of sociological with economic theory (Kreps, 1990).
Economists have given pride of place to non-cooperative games, usually reducing any negotiation between actors to a prior move in an extended game (the so-called Nash Program). Whether, however, this stratagem will turn out to suit the analytical purposes of sociological theorists only time will tell. What seems certain, though, is that both cooperative and non-cooperative games will play a significant future role in the formulation of sociological theory.
In my view serious sociological theory is at a crossroads: not only has it to separate from much recent tradition (found elsewhere in this volume) by explicitly becoming both propositional and deductive, it will often need to take the strategic implications of social interactions into account. Coleman (1990) has provided us with the seminal text for the analysis of non-strategic interactions (competitive
exchanges), but my feeling is that, in the future, strategic exchange will be more at the centre of things. We now turn to analyse the implications of this feeling.
The Nature of Theoretical Problems in Sociology
A valid theory amounts to a reasoned conjecture about the nature of a more or less latent mechanism that purports to account for a puzzling "empirical" event or relationship. Although rational choice theorists have no reason to deride the oft-touted observation that empirical events and relationships are themselves cc theoretically described," in the sense that there is no pre-theoretical realm of facts at our disposal, they play down the importance of "descriptive theory" in our theoretical endeavours. In the spirit of the Weberian program with which 1 opened, the mechanisms may be further conjectured to involve the social actions of individuals. A simple diagram (see figure 9.1) adapted from one popularised by Coleman (1990) is helpful in giving some initial orientation.
This diagram may initially be seen as a pointer to four generic types of causal relationship, each of which may require theoretical treatment. They are: 1 Macro, or system, level relationships: for example, how Protestant religious octrine causes the economic organisation of society (the example through- out is from Coleman (1990) courtesy of Weber). 2 Macro (or system) level to individual level (or micro) relationships: for example, how Protestant religious doctrine causes individual values. 3 Micro, or individual, relationships: for example, how individual values cause individual economic practices.
Macro Causes 1 Macro outcomes
Micro (individual level) Micro (individual level) Causes action
Figure 9.1 The Coleman-Lindenberg diagram
4 Micro, or individual, level to macro level relationships: for example, how individual economic practices cause the economic organisation of society.
Although the diagram is disarmingly simple, and we shall have occasion to enlarge upon it, a few initial points may be made. First, from a rational choice theory point of view the explanation of type (1) relationships is usually made through the conjunction of the other three types. In other words, the latent mechanisms which generate the puzzling event at the macro level are usually deemed to operate through the micro (individual) level. Thus, rational choice theory's reductionism or methodological individualism.
theory is proffered not as a descriptively accurate (realistic) model of individuals or their interactions but as a simplification thereof designed to render highly complex mechanisms which cannot be observed directly theoretically tractable. Rational choice theory might prove to be the wrong theoretical simplification in this respect; if so another one will be needed, but the point is that it would normally be wrong, if this happens, to reach for the rich descriptive vocabularies of interactionists and phenomenologists or psychologists. Of course if these vocabularies are empirically accurate then the sparse abstractions of rational choice theory must necessarily not be inconsistent with them. But how should we decide upon what sort of simplification to settle upon which is not inconsistent with received wisdom from these directions?
Lindenberg (1992) draws our attention to two important criteria: minimum information per actor and the adoption of models that can be systematically and progressively complicated (i.e. made more "realistic") as need be. But how should we interpret the phrase "need be" and how are we to know where the minimum information per actor requirement places us? Although I don't think these sort of issues have been fully worked out we can perhaps make a tentative start (Lindenberg is essential reading on these matters).
Again we return to the overall explanatory objective - namely to account for a macro level outcome. Coleman (1990: 2) as always puts it well:
The focus must be on the social system whose behaviour is to be explained. This may be as small as dyad or as large as a society or even a world system, but the essential requirement is that the explanatory focus be on the system as a unit, not on the individuals or other components which make it up.
Now, intuitively, the simplicity of the models we should adopt of both individuals and their interactions will be directly proportional to the complexity of the "system whose behaviour is to be explained." If the system itself is simple - e.g. a dyad - we may be in a position to select a fairly rich model and vice versa. But, in addition, simplicity will be a function of the distinctions we wish to preserve at the system level. We may, as it were, proceed backwards making these distinctions one constraint, among others, in determining the optimal simplification of our models.
The maxim of no emergent distinctions is central. That is to say, a distinction (i.e. the boundaries of an equivalence class) at the system (macro) level implies (is necessary and sufficient for) a distinction in its generating mechanisms (i.e. the interacting individuals) at the micro level. Putting it succinctly, there are no macro distinctions without micro distinctions. The reverse is, though, not true: there may be micro distinctions without there being any macro distinctions, in the sense that diverse micro processes may produce equivalent macro outcomes. But we are not theoretically interested in the former and, therefore, may select models of individuals and their interactions, which elide such distinctions. Thus, the complexity of such models is determined largely by the level of discrimination we require at the macro level. We should adopt models, which are as complex as is necessary to determine distinctions drawn at the macro level.
One common criticism which rational choice theorists level at much contemporary sociological theory is precisely that this lesson has not been heeded. Be this as it may, it is important when seeking to criticise rational choice theory to do so upon grounds where it is vulnerable (and there are many), not where it has no pretensions.
The second general issue concerns the idea of interdependent actions. Weber was satisfied with a definition of social action whereby an action by one actor "takes account of the behaviour of others." We need to be somewhat more precise. First, it is often not the behaviour but the action or social action of others, which is taken into account (Parsons would agree). Second, we need to tense the definition - so an action by one actor may depend upon what others have done (and/or not done), are doing, or are expected to do. When actions are interdependent in this manner the actions or choices made by the individuals are said to be strategic and, as we shall see, the appropriate analytic framework is game theory. However, strategic interdependence (social action) is by far the most complex to model and in the spirit of searching for the simplest model is resorted to only when necessary (below).
In practice much of the choice involved in modelling the micro-macro bridge (the right-hand side of figure 9.1) centres around the selection of the appropriate model of social interaction or, more generally, action interdependencies.
It is perhaps helpful to make this explicit by replacing the diagram in figure 9.1 by that in figure 9.2, which draws our attention to the fact that it is individual social actions (or interdependent actions) which generate the macro outcomes that require explanation. I have added a type (5) explanation here to allow for institutional constraint (rules of the game in game theory) upon social action.
One way of appreciating the contribution rational choice theory makes to sociological theory is to specify the various ways in which actions can be (inter) dependent upon one another. Doing so gives an indication of the differing theoretical frameworks which rational choice theory can bring to bear. We will, however, postpone a consideration of these matters until later when we have examined the concept of rational choice or action rather more closely.
Assumptions of Rational Choice Theory
There are many technical introductions to rational choice theory and the interested reader is referred to the copious literature (a good start is Moser, 1990). We will, however, adopt a rather informal approach which will, I trust, integrate the ideas presented into more sociological ways of thinking. On these matters Elster (1989a) is, as always, a source of inspiration.
For our purposes rational choice theory can be regarded as starting with the following assumptions, though as we shall see they may be relaxed in one way or another whilst still staying faithful to the spirit of rational choice theory:
1 Individualism - it is only individuals who ultimately take actions and social actions and it is, thus, social actions, which cause the macro social outcomes or events,
The uninitiated reader might quite legitimately feel entirely dismayed when faced with this list. First, a number of the constituent terms have not yet been defined. Second, even if they were, surely, it is easy to think of situations where actions are neither self-regarding nor, most likely, optimally chosen, whatever this latter phrase might mean. Third, is it not the case that sociologists often wish to embrace super-individual entities (collective actors) in their theoretical endeavours? A partial answer to these worries might be forthcoming if we analyse each of the assumptions a little more closely.
Turning first to individualism; it would be inappropriate here to rehearse tired debates about methodological individualism, reductionism, and related ideas as these have on balance not served sociology well (O'Neill, 1973). Many sociologists, however, following Durkheimian leads, will no doubt wish to part company from the start. This eventuality would be unfortunate, for all that rational choice theorists are tied to is the assertion that macro or system level causal relationships (type (1) in figures 9.1 and 9.2) are shorthand for the conjunction of type (2), (3), and (4) relationships. A puzzling connection between two system level variables must imply mechanisms involving a more micro level of analysis. This will usually, though not invariably, be the individual level, as rational choice theory is perfectly at one with a nested version of figure 9.2 which allows for intermediate levels lying between the individual and the chosen macro levels (e.g. groups between individuals and "societies"). Then the perspective allows for different degrees of reduction, although if a relationship at any level above that of the individual is deemed puzzling then ultimately a reduction must be carried out down to the individual level.
The theoretical question is this: how has the system level relationship been generated by the social actions of individuals? The individualism of rational choice theory necessitates no deeper a commitment than this, and what should be clear is that it does not, contrary to some critical portrayals, disavow the importance of system level relationships, or assert an asocial model of the individual or, indeed, deny the exogenous "primacy" of a group or institutions in accounting for a given social outcome. In the face of persistent institutions, however, the rational choice theorist will usually seek an account of their genesis. So, institutions, themselves, become macro outcomes in need of explanation, although - once again in the spirit of figure 9.2 - such an explanation may invoke prior institutions. In this context and many others it is useful, in order to fix the nature of the explanatory problems in hand, to conceive of a backwards (in time) iteration, as in figure 9.2.
Next, consider "optimality;" let us say, given the set of opportunities faced, an individual chooses optimally if no other (social) action exists whose consequences he or she prefers to the chosen course of action.
A word about "transitivity:" most formal treatments of rational choice theory make assumptions whereby, if A is preferred to B and B to C, then A will be preferred to C. This does seem rather reasonable, almost, we might say, following from the idea of preference itself. Experimental evidence, however, suggests that people do not always preserve transitivity (Kahneman and Tversky, 1990). Furthermore,
people are sometimes not able either to compare options (partial ordering) or, if they can, to choose between them (weak ordering or indifference). Although problems like this, which might be quite common in sociological phenomena, create difficulties for rational choice theory they are not insuperable and need not detain us here. Certainly, in the spirit of selecting "simple" models of the actor we will often not be far off the mark in assuming actors themselves select a course of action which dominates their ranked alternatives.
If people were to act insistently in a way that is inconsistent with their preferences this would certainly bring rational choice theory to its knees. The major alternative assumptions seem to be that:
1 individuals characteristically do not act optimally in terms of their preferences; 2 although individuals may act optimally in terms of their preferences, the preferences themselves are not optimally formed; 3 individuals do not act upon preferences at all, but are driven by forces beyond their control.
In order to judge the warrant for the strong claims of optimality which rational choice theory enjoins upon us, it is helpful to look at these alternatives.
Acting in Terms of Preferences
To maintain that, in the general course of things, actions are knowingly chosen so, as not to optimise satisfaction is surely otiose. Certainly, some cases are well documented, by Elster (1989b) among others, in which weakness of will may be said to prevent people from acting in the way they prefer - that is, according to their better judgement. Furthermore, at times we act impulsively, which in retrospect we find distasteful, and we may also be possessed by our desires which then cloud our reason. All of this and perhaps more must be acknowledged, but to raise these peripheral matters to a position of central importance in modelling micro-macro mechanisms would be, to say the least, counterproductive.
Note that nothing that has been said so far holds one to the view that the dominant preferences upon which the actor is deemed to act necessarily represent his or her best course of action in the sense that the preferences themselves are optimally constructed. We are held merely to the view that, whatever preferences the actor happens to have, he or she pursues a course of action, which is optimal with respect to them. It the second point above which is more problematic for rational choice theory.
Formation of Preferences
So let us now examine the formation of preferences. Some, particularly economists, take the view that preferences are not only fairly universal and stable (when compared with opportunities which, it is supposed, account for variations in action) but are also formed in ways that entirely resist the attentions of rational choice theory. Their genesis is either uninteresting or, if not, entirely beyond the
redirected, away from preferences toward affects. To my knowledge there is no good theory of the genesis of and shaping of affects, though rational choice theorists are inclined to invoke some social contagion or learning model. Indeed many of those persuaded by the virtues of rational choice theory see the combination particularly of game theory and learning theory as the way ahead (Kreps, 1990).
Structural Explanation
Turning finally to the idea that actions are not driven by preferences, it should be clear that type (1) explanations (figures 9.1 and 9.2) are the main candidates for structural explanation and we have already seen the rational choice theorists' negative response to taking these sui generis. If, however, in establishing the links through the micro level, we are driven to the view that the actor has only one alternative, a type of structural explanation which is independent of any concept of preference is acquired. Social norms or institutions are sometimes deemed to operate in this manner, reducing the scope of action to a fixed point. So the rational choice theorist regards "structuralism" as a special case of his or her own theory (Elster (1989a; 198.9b) is a particularly rewarding read on these matters; Roemer (1982) considem them from a Marxist perspective).
Although a modern rational choice theorist is prepared to adopt a rather relaxed standpoint about optimality, some idea that actors do the best they can, albeit often in circumstances they do not fully comprehend, is essential to the theory. Even in a relaxed form it is the major assumption which confers upon the theory its deductive and explanatory powers. It is accordingly this assumption of the theory, which must be proved wrong to bring it into disrepute.
Self-Regard
Let us now briefly turn to the assumption of self-regard. Although the standard version of rational choice theory starts with the view that individuals act (optimally) to satisfy their self-regarding preferences, there is nothing intrinsic to the theory which excludes other-regarding sentiments directed towards either individuals or collectivities. There are now many attempts to incorporate altruism, malice, and, more generally, relative utilities into a rational choice theory framework (e.g. Margolis, 1982). Rational choice theorists remain, nevertheless, ever cautious about invoking such sentiments and are inclined to search for hidden self-regard when faced with apparent other-regard.^ Further, where there exists indisputable evidence for, say, altruism, they will often mobilise an evolutionary "test" based upon repeated interactions to show how self-regarding populations could have generated and maintained the altruism through a process of random variation, selection, and copying (e.g. Frank, 1988). It is this tireless attempt to promote human beings as almost innately self-regarding which is most offensive to many sociologists who feel that both benign and malign attitudes toward others are central to social life. But there is nothing intrinsic to the theory, which hastens us to adopt this stance. Indeed, it must largely be an empirical matter as to where the truth lies. Nevertheless, the caution which many rational choice theorists enjoin upon us can be turned to our advantage; it is all too easy when confronted by a puzzling macro phenomenon to invoke a convenient assumption about individual
affects which renders the explanation tantamount to a tautology (Hechter, 1987). At least the conservatism of rational choice theory can protect us against this eventuality.
Do groups form because it gives people the opportunity to serve the group or because they serve a self-regarding need - or, indeed, for both reasons? The rational choice theorist's approach should be as follows: if there is independent evidence for group sentiments, then construct a theory appropriately and also diary theory accounting for the origins - and perhaps persistence - of the sentiments. In the absence of such evidence construct a theory on the basis of simple self-regard and confront its deductive consequences with, as wide a range of evidence as is possible. Only if this fails should one search elsewhere.
This might sound a little trite, but there is an important message here. On the one hand, we should not ignore any independent evidence at our disposal concerning other-regarding sentiments (that is, independent of the outcome we are trying to explain). Many rational choice theorists do run foul of' this dictum in their zeal to promote an exclusive reliance upon self-regard -and as a consequence construct overelaborate theories when there is no need and where the explanatory focus should shift to the sources of the sentiments themselves. On the other hand, we should beware the cavalier invocation of "exotic sentiments. Paradigmatic Privilege
Finally, we turn to paradigmatic privilege - although this need not detain us for very long. It will 1 trust be fairly clear at this stage in the argument why rational choice theory may claim explanatory privilege. It does so not on the grounds of its exclusivity but, rather, in terms of its claim to be our first choice of framework for theory construction and thereby, by its own canons, generating the need for auxiliary theories when it has done its best. Currently these auxiliary theories appear to be a complex mixture of network theory, learning theory, mimetic theory, and evolutionary theory. We will encounter some of these below.
Interdependence and Social Action
If, following Weber, we place social action at the centre of things, the ways in which actors are dependent upon one another also become central. It is rather surprising that despite the avowed importance of social interactions to many theoretical traditions in sociology we are not overburdened with theoretical typologies of interaction.
For the purposes of rational choice theory it is useful to distinguish between independent actions (Weber's action) and two types of (inter) dependent actions (Weber's social actions) - parametric social actions and strategic social actions. Independent actions call for no special comment, except to note that it is rare for sociologists to take an interest in situations where individuals act independently of one another.
Social actions are parametric where the action of others can be taken as independent of what the focal actor does, so he or she does not have to calculate what others will do as a consequence of what he or she does or in anticipation of
scores but the mean ability and/or achievement scores in the individuals' group or social environment. There are good rational choice theories of this finding (E.g. De Vos, 1989), which is usually deemed to be rather puzzling.
Structural or compositional models offer us a rather gross picture of the parametric dependence of each individual upon others in his or her social environment but do not fundamentally alter the aggregation process. The units are still regarded as independent, but with individual level scores derived from those who surround them.
A more detailed picture of individual interdependence is provided by models of endogenous and exogenous autocorrelation (Anselin, 1988'). Here the units of analysis (e.g. actors) are explicitly linked to one another in the sense that the scores on their endogenous variable are a function of the scores of at least some of the other actors' endogenous and/or exogenous variables.
Endogenous autocorrelation may be used to model the interdependence of the actions of individuals as a consequence of their location in an "influence structure." Aggregation to the macro level still produces only a distribution of the endogenous variable, but one that now depends crucially upon the pattern of connectivity of the actors at the micro level. Rational choice theorists take a particular interest in "influence structures" when, in highly uncertain environmerits, it becomes rational (the best thing to do) for actors to copy those around them - rational mimetics (Abell, 1991). Modelling of this sort links rational choice theory into theories of social contagion as well as into network theories.
Although these models are used largely to study the parametric dependence of an actor upon those other actors to whom the first actor is connected, and where the structure is exogenously given, there is no intrinsic reason that this should be the case. Markov models, which purport to show how rational actors construct connections to others, are possible (Leenders, 1993) and the two-way interdependence so generated may be given a strategic interpretation (see below).
Social exchange interdependencies are perhaps the most intensively studied in the sociological literature. In the spirit of our basic distinction, it is useful to distinguish between parametric and strategic models. Though the latter are the more fundamental, the former are the more thoroughly analysed, Coleman's Foundations of Social Theory being the canonical source. Parametric exchange models largely ape competitive models in economic theory by assuming that exogenously endowed individuals trade to a Pareto Equilibrium under given (i.e. parametric) rates of exchange (prices). The details of the movement to the equilibrium - that is, the macro outcome - are abstracted away. Strategic models, on the other hand, invite us to analyse the tactics of the actors in arriving at rates of exchange. Here, the precepts of game theory become central in understanding the macro outcome (see below).
Coleman has also provided extensive analysis of actions and social actions, which generate externalities (i.e. impacts), for other parties; it is externalities, which create the need for social norms. Coleman has thus promoted a theory of "the
demand for and supply social norms, whilst working from a rigorous rational choice perspective. Again this theory is parametric in structure.
One further way in which social actions can come about is where actors actually have (inter) dependent utilities/preferences. We encountered other regarding sentiments above. These sorts of utilities in effect build externalities into the reasoning of actors.
So far, we have concentrated almost exclusively upon parametric social actions; however, current research strongly directs our attention toward strategic situations. Of course, in the light of our earlier plea for simple models, it is always important to see how far parametric models can take us, before moving onto strategic ones, as the former are generally more simple. If they fail, however, the complexities of strategic situations must be addressed.
Social Action and Game Theory
Game theory extends the theory of rational individual action to situations where actors take other actors' actions into account when acting themselves. So the fate of an actor depends not only upon his or her own actions but also upon the actions of others, and the outcome, one of a set of possible outcomes, is generated by more than one actor.
Many micro-macro mechanisms have a complex strategic structure and it is to game theory that we have to look in attempting to unravel their complexity Although game theory increasingly lies at the centre of modern sociological theory it has unfortunately not yet become staple fare in sociological theory courses and is therefore little understood. This creates a difficulty in that the space constraints of a single chapter preclude a full-blown introduction, yet it is inappropriate to assume too much. The best approach 1 can think of is to sketch the potential role of game theory making use of standard concepts, each of which is italicised in the text. Readers who are conversant with these and wish to know more can refer to any standard text on game theory. I have tried to cover the minimum set of concepts that the contemporary social theorist needs in order to gain some orientation in this field. For those who want to take things further, Binmore's Fun and Games (1992) is both readable and takes the reader near to the intellectual frontiers; others may prefer to go back to Games and Decisions (1957) by Luce and Raiffa, or to look at Myerson's Game Theory: Analysis of Conflict (1991) or Games and Information (1989) by Ramusen. Hamburger's Games as Models of Social Phenomena (1979) gives an impression of the range of possible applications. Rationality and Society, vol. 4, no. 1 (1992) is concerned with game theory in the social sciences. Raub and Weesie's Sociological Applications of Game Theory (1992) offers a very concise introduction to the field.
At the most general level game theory can be used to clarify the nature of strategic situations and wherever possible to point to an equilibrium solution (sometimes a dominant equilibrium, usually a Nash Equilibrium or some refinement thereof) which predicts the actions (strategies) which each rational
accommodated this pattern of line of criticism by assuming players learn by trial and error (perhaps involving a chance component) to master the game and find an equilibrium.
A parallel approach which I suspect will find a secure home in the sociological theory of the future does not even require players (actors) to learn how to play rationally, but only for each to select any one of the strategies available and then to repeatedly interact with others (either randomly or on a selective basis). If players then switch strategies to emulate those who are more 'successful' according to some dynamic specification we have what is termed as evolutionary game (Weiball 1996, for a good though technically demanding introduction). Matching ideas from evolutionary game theory with numerical simulation of the imitative dynamics (deterministic or stochastic) has recently become a most exciting prospect for the aspiring theoretician (Skyrms 1997, Bahr and Passerini 1998). It enables the theoretician to study the dynamic behaviour of systems which are invariably too complex to lend themselves to analytic solutions and, in effect, to conduct quasi-experiments where genuine experiments would prove impossible.
Evolutionary game theory has its own concept of equilibrium (first formulated by evolutionary biologists) in terms of an evolutionary stable strategy (ESS) - a strategy (perhaps mixed in a population) which the dynamics leads to and which cannot be supplanted by 'invading' strategies. The relationship between ESS and Nash equilibria is rather complex and an area of active research (Bandor and Sarstak 1995, 1997). Deterministic dynamics of an evolutionary games will select a particular Nash equilibrium (where there is more than one) in path dependent manner. Surprisingly the introduction of stochastic dynamics often leads to path independence (Sandholm 1998 for an accessible review).
From the sociologists point of view the most promising model is one where individuals (playing a strategy) interact selectively with 'neighbours' only. Though analytical solutions are rare, simulation can give an impression of the range of possibilities.
If I care to pin-point any one analytical framework which might prove to be the most beneficial in the development of a genuine social theory I would select evolutionary game theory. It models situations (1) with a strategic structure (ii) which do not depend upon hyper-rational assumptions and (iii) where dynamics select equilbrium outcomes. The dynamics can be modelled in various ways and can even introduce self-organisation and chaotic collection behaviour into the picture (Skyrms).
Building the Sociological Theory of the Future
The great challenge for sociological theorists is to model the bridge between, on the one hand, individual social actions and macro outcomes and, on the other, between macro outcomes and individual social actions. To achieve these twin objectives we need a model of the individual actor, which is simple and flexible enough to give us a reasonable chance of success. Rational choice theorists believe that rational choice theory fits the bill. All this holds us to, in the final
analysis, is an assumption whereby individuals do the best they can, given their preferences over the opportunities that they see as facing them.
The mechanisms which link individual social actions to social outcomes (not necessarily intended) are frequently complex (Abell, 1993), involving many actors and intricate patterns of strategic and parametric interdependence. The only framework we possess at the moment for engaging this complexity is game theory
- even with all its limitations. However, the mechanisms characteristically possess an additional complexity, which calls for special attention:
By complex I mean the sort of sequential material which a rich historically based case study characteristically provides. Case studies of this sort usually possess a number of features. First, they often involve many actors (units of analysis) the actions of which are parametrically or strategically interdependent; second, these actors/units may be at different levels of aggregation, e.g. individuals and collectivities of one sort or another; third, the aggregation may change in the course of the narrative, e.g. the creation of an effective collectivity; fourth, there may be complex patterns of parallel processing, that is branching and recombination of sub-plots; fifth, the time signature of parallel processing might be on a different scale. (Abell, 1993)
If this sort of complexity is to be theoretically grasped we have to develop ways of dealing with it. From a game theoretic point of view, it is likely that a number of games are going on at the same time, the outcomes of which are parametrically linked. We need ways of describing these complex processes such that they can be treated as "sequential facts" and then, hopefully, as equilibrium paths in the constituent games.
In recent years a number of techniques have been developed which direct our attention towards sequences of interactions as the factual basis of sociology (e.g. Abbot, 1984: sequence analysis; Fararo and Skvoretz, 1984: productionsystems; Abell, 1987; 1993: narrative analysis; Heise, 1990: event structure analysis). Though these techniques are not necessarily tied to a rational choice theory framework, the connection can easily be made. Directing sociological theory back towards the provision of explanations of sequences of interactions (which generate macro outcomes) is perhaps the most significant recent turn in the world of social theory (Fararo, 1989). Similarly, the complex macro to micro processes which establish our beliefs, affects, and norms/values (constitutive of our preferences) and structure our opportunities are also notably sequential and will eventually require parallel treatments. Armed with these sorts of analysis we will finally begin to realize Weber's ambitions to achieve an interpretative understanding of social action in order thereby to arrive at a causal explanation of its course and effects.
