Rational Choice Theory Research Paper

Use of RCT in sociology ranges from formulating mathematically deductive models assuming fully informed and flawlessly calculating actors, to a much broader approach, whence actors are seen as ‘doing their best’ (i.e., subjectively optimizing), given certain preferences/desires/affects/utilities, with only a limited understanding of their environment and partial reasoning or calculative ability. It is this flexibility which recommends RCT to many of its proponents, ranging from those for whom it comprises the hallmark of any acceptable social science to those for whom it is pragmatically the best framework currently available.

RCT may be used in two complementary ways. First, in a positive manner by proposing explanations of individual and collective actions (decisions) and consequently responding to the question—how (in context) do actors behave? Second, in a normative manner, responding to the question of how they (again in context) ought, given their preferences, to behave. While strong in economics, the normative approach is not widely embraced in sociology.

Rigorous applications of RCT are almost invariably based upon an assumption of the maximization of (often subjective (expected) utility). Actors are thus conceived as (a) capable, at any instant, of taking one of a series of feasible actions, (b) each of which leads to an estimable set of outcomes, with (c) a known (subjective/objective) probability. Further, each outcome carries affect for the actor. The actor can then form a preference ordering over the actions in terms of affects/utilities. Actions (decisions) under certainty, risk and uncertainty are thus special cases of the general model. Various axiomatic systems are available, though the one attributed to Savage is closest to sociological concerns. It is now recognized that actors do not always abide by the precepts of expected utility theory. Prospect theory and developments thereof, which introduce a wide range of ‘biases,’ are widely accepted as falling under the broad ambit of RCT, as are models based upon limited rationality (like satisficing).

Three related types of RCT modeling frameworks can be distinguished. First, independent decision-making (actions), where the actor’s environment may be treated as fixed (i.e., parametric); second, where the environment involves other actors and where consequently decisions or actions are interactive (i.e., strategic), where one shot and (finitely and infinitely) repeated game theory and their associated Nash equilibria are appropriate. Third, evolutionary game theory, where strategies are selected by imitation of either the relatively frequent or relatively successful strategies of others. While the first two frameworks, analytically speaking, usually commence with the full assumption of well informed and rationally calculating individuals, the latter requires no such initial assumptions. It does, though, usually assume that evolutionary process is driven towards an evolutionary stable equilibrium (or some refinement thereof).

The assumptions of expected utility maximization (more generally RCT) in complex environments (parametric or strategic) prove demanding on both the reasoning power and knowledge (information) of the actors concerned, and are sometimes deemed by critics to be entirely unrealistic in a positive context. RCT is, however, strongly wedded to the idea of equilibrium, and in off-equilibrium adjustments all manner of trial and error may be allowed for. Equilibrium, in this view, is not necessarily arrived at (learned) because actors are rational, but rather because actors appear rational (i.e., discover rationality) once an equilibrium has been achieved. Evolutionary game-theoretic models derive the same conclusion by myopic processes of selection (emulation) eventuating in a evolutionary stable equilibrium.

All manner of RCT models may be said to rely initially upon three assumptions, each of which may be regarded as controversial in some sociological circles. They are:

Individualism. Only individual persons ultimately take actions.

Optimality. Individual actions are chosen optimally (at least at equilibrium).

Self-regard (or egoism). Individuals’ actions are ultimately concerned with their own welfare (and nobody else’s).

Though any one or more of these assumptions may be adjusted in a perspicacious manner (see below), rational choice theory may nevertheless be said to retain its paradigmatic privilege in the sense that it (with its standard assumptions) provides an analytical point of departure.

Despite its explicit individualism, system level or collective/macro relationships between variables are a perfectly acceptable currency for RCT as long as they are construed ultimately as ‘shorthand’ for the con- junction of relationships that ultimately run through individuals. Indeed, a theory about the connection between two system-level variables must ultimately imply a mechanism involving individual actions. The theoretical question is precisely this: How has the system level relationship been generated by the actions of individuals? The assumption of individualism in RCT implies no more than this about the explanation of system or collective level empirical generalizations.

No contradiction exists between RCT and a view which pictures individuals’ actions as ‘shaped’ by their membership in a group (more generally a collectivity). If, as some ‘structuralists’ opine, individuals are often so circumscribed by internalized standards and external constraints upon their actions that they barely have any choice at all, then this may call into question optimality and self-regard, but it does not undermine the assumption of individualism in the sense that it is still the motor power behind individual actions (albeit ‘programmed ones’) that drive the social world.

Things happen in the social world because individuals do and do not (i.e., forbear) to do things, and they are the only things that do and do not do things. All statements that attribute ‘doing’ to other entities can, in principle if not in practice, be translated, without loss, into statements about individuals doing things.

Individuals choose optimally if no other action exists whose consequences are preferred to the chosen action. An actor’s preference-driven action will in turn depend upon that person’s (a) beliefs or reasoning about the consequences of the available actions, and (b) affect (desire for/against) the consequences (and perhaps for the action itself). It is thus possible by derivation to speak of optimal beliefs and affects. It is this broad assumption of optimality that gives RCT its paradigmatic privilege.

The major alternative assumptions to this viewpoint are:

(a) That individuals frequently do not act optimally in terms of their preferences.

(b) That although individuals may act optimally in terms of their preferences, senses exist in which their preferences are themselves not formed optimally (nonoptimal beliefs and affects).

(c) That individuals do not act upon choices but are driven by forces entirely beyond their control (structuralism).

Cases are well documented in which ‘weakness of will’ may be said to prevent people from acting in the way they would ‘really prefer.’ Preferences may also be imprecise or unclear, actions entirely experimental or based upon wishful thinking, and more than one optimal choice may be available. All of this is acknowledged by RCT, although in practice most of the problems reduce to examples of nonoptimal preference formation.

Some take the view (particularly many economists) that preferences are not only stable (compared to constraints) but also are formed in ways that entirely resist the attentions of RCT. Their genesis is either uninteresting (changed constraints accounting for changed action) or beyond rationality. Sociologists reject the first view almost unanimously, but find some sympathy for the second. Not infrequently it is also claimed that the idea of optimality is so weak, when taken by itself, as to constitute an affront to serious understanding. It is observed that, if we are to be permitted any sort of utility arguments, then actions may (certainly ex-post facto) always be made to look as if they are optimal.

The accusation is that, under such circumstances, the assumption of optimality can be protected entirely against potential refutation. In this context, sociologist think in terms of a series of derivative optimality questions:

(a) Are preferences formed optimally with respect to beliefs and affects?

(b) Are beliefs formed optimally with respect to the information available?

(c) Is available information (collected/processed) optimal with respect to affects?

(d) Are affects optimal with respect to individual autonomy?

An action is objectively optimal when conditions (a) to (d) hold true. Such actions may take place under certainty, risk, or uncertainty, according to standard assumptions about the nature of beliefs. These assumptions are collectively very strong, and many social theorists maintain that it is only rarely, if ever, that actions will conform to the ideal pattern.

The issue at stake is whether we may assume that affective dispositions in some sense flow from a balanced or autonomous personality. No doubt forms of behavior/actions exist that flow from affective dispositions that are ‘irrational’ or pathological. When this occurs, it usually implies that individuals resist all occasions to consider the opportunity costs of maintaining the disposition (e.g., constantly striving for what is beyond their grasp or manifestly damages them).

It is points (a), (b) and (c) that have attracted the most attention. It is often proposed that preferences cannot be formed optimally for a variety of reasons, prominent among which are the following:

Satisficing. Beliefs are not performed optimally with respect to information available.

Search. No adequate theory of optimal search exists; therefore, once again, beliefs are not optimal with respect to information, nor information with respect to affects.

Decision biases. Preferences are not optimal with respect to beliefs and affects.

Exclusions. Not all possible actions are considered.

Limited deliberation. Not all consequences of all actions are thought through.

Received styles of reasoning. These prevent optimal belief formation.

Ritual routine habit copying. These systematically lead to exclusions and do not evolve optimally in an evolutionary sense.

In each case a conceptual understanding of the apparent limitations of RCT is established in terms of the departure from full optimality. This is one significant sense in which RCT can make a claim for paradigmatic privilege: It still serves, even in default, as a benchmark. Furthermore, many of the above factors are modelled in terms of limited information environments.

The standard version of RCT usually starts with the assumption that individuals act (optimally) in order ultimately to satisfy their self-regarding preferences (utilities). That is, optimal actions are taken in a way that is indifferent to the welfare/utility/consumption of others, considered either individually or collectively. Although nothing intrinsic to the theory mandates this assumption, it is frequently regarded as a natural analytical starting point, to be discarded only cautiously if and when the facts speak otherwise.

Furthermore, if other-regarding preferences are invoked eventually, not infrequently it is maintained that they must in some way be explained or accounted for, whereas self-regarding sentiments do not invite explanation, being in some way natural or self-explanatory. Thus many theorists adopt the practice of assuming that, if no evidence exists to the contrary, then postulate self-regarding utilities. If, however, independent evidence exists for other-regarding sentiments, it would be a mistake to ignore it and go on to ask for an explanation of the source. However, many RCT models interpret other regarding sentiments as instrumental to self-regard (enlightened self-interest). Such models are usually excluded before other regard is introduced into the utility function (affective orientation).

Evolutionary biologists (and evolutionary game theorists) continue to debate the issue as to whether altruistic motives can evolve at an evolutioning stable equilibrium. Sociologists are required increasingly to relate their own theories to those of colleagues working in these adjacent disciplines. This is particularly so when they adopt the view whereby it is the strategic interaction between two or more actors which constitutes much social interaction. These developments are bringing sociological theory increasingly into contact with game theory and evolutionary game theory. Although games of incomplete information currently play a limited role in sociology, we can expect this to change in the near future. Perfect Bayesian equilibra will thus come centre stage.

The paradigm of sociological investigation which is emerging from an RCT perspective takes the following form:

(a) Initially, assume the actors who are responsible for generating a collective outcome (which may not be intended but only a consequence of what is intentionally brought about) to be rationally self-interested, fully informed and endowed with calculative ability given their (exogenous) objective resources, and (exogenous) preferences/utilities.

(b) Model the structure of the actors interdependencies (social interaction) in the sense that the outcome depends jointly upon what each does. Thus each actor has to take account of what others do (will do, are likely to do, and, if appropriate, have done in the past). Such modeling will usually define a feasible set of actions strategies for each actor.

(c) In the light of (a) and (b), determine the course of action each actor will (rationally) pursue. If these are not unique, then specify their ranges of actions (i.e., multiple equilibria). If the interactions are repeated, adopt the appropriate repeated game theoretical framework.

(d) Thus predict the outcome(s) and test these against descriptions of the outcomes one wishes to explain. Of course, if the model possesses no unique outcome, then all one can require is that the event in question is among the set of possible outcomes predicted by the model (an envelope rather than point prediction). Specific cultural historical occurrences may select one equilibrium (i.e., focal points) and suggests path-dependence on a particular outcome.

(e) If this simple model fails, construct further models which might involve one or more of the following possibilities:

(i) Relax assumptions of full information, probably following recognized procedure in game theory.

(ii) Relax the assumption of self-interest, introducing richer utility/preference functions (e.g., altruism, malice, indifference, envy, etc.).

(iii) Relax the assumption of objectively calculated resources introducing subjective beliefs about resources.

(iv) Relax the assumption of a set of feasible actions, calculated objectively, introducing subjective beliefs about what is feasible.

(v) Re-compute the structure of interdependencies.

When either (i), (ii), (iii), or (iv) are seen to hold, this will inevitably prompt a further question as to why it should be so. Answers to these questions might also be couched in terms of some of the RCT precepts outlined above (e.g., strategic limitations of information). It is only if this very general framework fails to provide the answers we seek that we should then reach for an alternative theoretical framework.

Bibliography:

1. Abell P (ed.) 1991 Rational Choice Theory. Elgar, Aldershot, UK
2. Abell P 1996 Sociological theory and rational choice theory. In: Turner B S (ed.) Social Theory. Blackwell, Oxford, UK
3. Boudon R 1989 Subjective rationality and the exploration of social behaviour. Rationality and Society 12: 173–96
4. Coleman J S, Fararo T J 1992 Rational Choice Theory: Advocacy and Critique. Sage, Newbury Park, CA
5. Moser P K (ed.) 1990 Rationality in Action: Contemporary Approaches. Cambridge University Press, Cambridge, UK