The question of human rationality has been debated since time immemorial. In recent decades, neuroscientific and behavioural evidence seem to suggest that people are not the paragons of rationality as we might have once believed. However, orthodox economic theories of human behaviour appear unable to shake their reliance on assumptions of absolute rationality, and the representative agent.
Therefore, this essay examines the evolution of the treatment of the inquiry— “Are humans rational?”— in economics. It traces the progression from the conception of homo economicus, to boundedly rational agents, and finally, to models of the dual-self. This evaluation also considers the implications and methodological underpinnings of each of these stylisations of rationality.
In addition, defensible grounds for the persistence of rational agent models are provided. Notwithstanding, this paper dismisses the notion that without rationality, there is no economic theory. It also warns against the naive application of theoretic models that employ assumptions of pure rationality. To conclude, I argue in favour of the complementarity between a range of imperfect, yet useful, models of human behaviour.
Figure 1: Thinker by Fredrik Rubensson
Neoclassical economics
The main concern of the field of economics is “a particular subset of man’s behaviours— those having to do with the production, exchange, and consumption of goods and services” (Simon, 1976). The theoretic modelling of man’s behaviours require certain simplifying assumptions, or axioms, to ease computation and interpretation. One such assumption entails the abstraction away from peoples’ idiosyncratic and irrational choices and conduct in the real world. The representative agent that emerges from this axiomatic approach is quite like the perfectly rational man, described by Becker (1974, p.1078), that would eat with his fingers in front of his family:
“He would read in bed at night only if the value of reading exceeded the value (to him) of the loss in sleep suffered by his wife, or he would eat with his fingers only if its value exceeded the value (to him) of the disgust experienced by his family.”
The representative agent, homo economicus, is an expected utility maximiser who presides over perfect information and absolute computational abilities, and who typically operates in frictionless markets for goods and services. He decisively optimizes objective functions that are entirely known to him, and are subject only to the apparent constraints on his budget or productive capacity. These optimisations are made purely in his own self-interest, and in accordance with a set of “well-defined preferences and unbiased beliefs and expectations” (Thaler, 2016, p.1578). To derive these preferences mathematically, they are usually characterised by their stability, completeness, transitivity, strict monotonicity, and time-invariance. Notably, even his expectations of the future are depicted to be consistent (Muth, 1961).
Homo economicus is a predominant character in neoclassical economic theory and models, from which much value has been derived in the study of the allocation of scarce resources. Even Thaler, a well-known critic, describes this type of theory as “essential building blocks for any kind of economic analysis” (2016, p.1577). Rational actor models, whilst not uncommon in other social sciences, are distinct in classical economic theory insofar it specifically characterises the decision-making of an expected utility maximiser. Therefore, this approach could be more aptly described as the study of the rational allocation of scarce resources (Simon, 1976, p.2).
Neoclassical theories of rational resource allocation are also characterised by the concept of equilibrium, or steady state, at which point all representative agents are simultaneously exercising their optimal strategies given constraints. Yet, even at its the inception, the so-called “marginalist revolution” was marred with criticisms levelled against the realism and validity of such orthodoxy, particularly the equilibrium concept of “equating at the margin” (Lester, 1946, p.81). Despite being an ardent supporter of marginal theories, even Machlup (1946, p.520) conceded that a considerable portion of agents’ behaviour may be “non-rational, thoughtless, blindly repetitive, deliberately traditional, or motivated by extra-economic objectives.” Nevertheless, it was often implicitly— and rather blithely— asserted by neoclassical economists that a “theory of the economy must be based on rationality…Otherwise there can be no theory” (Arrow, 1986, p.s385).
Bounded rationality
Many observers could not square the neoclassical stylisation of rationality with actual human behaviour. Hence, the issue became subject to a diverse range of alternative interpretations. One such interpretation, bounded rationality, originated with Herbert A. Simon (1955, 1979). This conception of rationality proposes that actual economic decision-makers are not the perfectly rational agents employed by neoclassical theorists. Instead, it envisions the limited ability of an agent to discover and calculate his optimal behaviour for a given situation (Simon, 1978, p.10).
Therefore, bounded rationality acknowledges constraints on optimal decision-making on two fronts. First, there are limitations on the availability and accessibility to the relevant information for the issue at hand, i.e. problems of search. Second, agents have an imperfect ability to compute this information into an optimal economic choice, i.e. limitations on the procedure of decision. This view gave rise to Simon’s (1955) notion of “satisficing” according to which agents opt for the first satisfactory option available to them, as opposed to optimising. Notably, it was often (mistakenly) argued that the aggregations of the choices of satisficers were not systemically different to that of optimisers (Thaler, 2016, p 1581).
Contributors to the field of the New Institutional Economics have made productive use of this conception of rationality. It is held that the bounded rationality and limited resources of agents give rise to inefficiencies and errors in economic decision-making (Furubotn and Richter, 2005). Error-prone decision-makers are offered as grounds for the existence of market frictions, and reaffirms the notion of transaction costs established by Coase (1937). For example, the bounded rationality of members of a team is externalised as costs of intra-team communication (Simon, 1978, p.11), each instance of which amounts to a transaction. Non-zero transaction costs are a hallmark of this school of economic thought, and constitute a significant departure from the neoclassical assumptions of frictionless markets, rational agents, and perfect information.
Furthermore, contemporary proponents of decision theory also seek to understand and utilise the concept of bounded rationality. Their adoption of rational choice theories recognise the superiority of certain decision-making procedures for a given scenario, provided that such procedures are suitably applied (March, 1978, p.587). However, it is also understood that informational constraints and other human characteristics may cause agents to run afoul of these optimal procedures, thereby resulting in unsound decisions. These inherently human constraints on the procedure of decision are fittingly illustrated in Camus’s The Rebel (1951): “There are crimes of passion and crimes of logic. The boundary between them is not clearly defined.” Similarly, in The Theory of Moral Sentiments, Adam Smith describes these characteristics as the “basic passions of human nature” (1776, p.1).
Behavioural economics
The concept of bounded rationality was also adopted in the development of behavioural economics. The field is distinct insofar it appropriates insights from other social sciences, particularly psychology, in the study of human behaviour under scarcity and uncertainty. Thaler (2016), one of the field’s pioneers, describes it as “the mixture of psychology and economics” and cites psychologists, Daniel Kahneman and Amos Tversky,1 among other contemporaries, for their catalysing contributions to the field [p. 1577]. In short, behavioural economists heed the advice of the famous institutionalist economist, John M. Clark (1918, p.4):
“The economist may attempt to ignore psychology, but it is a sheer impossibility for him to ignore human nature, for his science is a science of human behaviour. Any conception of human nature that he may adopt is a matter of psychology, and any conception of human behaviour that he may adopt involves psychological assumptions, whether these be explicit or no. If the economist borrows his conception of man from the psychologist, his constructive work may have some chance of remaining purely economic in character. But if he does not he will not thereby avoid psychology. Rather he will force himself to make his own, and it will be bad psychology.”
By no longer ignoring human nature, behavioural models of decision-making necessarily take into account the numerous psychological insights that allude to the limited rationality of individuals. Indeed, Simon (1987) points precisely to this contravention of the neoclassical assumptions of rationality as the distinguishing feature of this branch of economics. Behavioural models, therefore, do not entail the pure optimisation by agents, nor does it guarantee that their decisions or beliefs will be “consistent” (Simon, 1983). In addition, such models tend to be derived from data as opposed to axioms, and are distinctly descriptive in the nature. In other words, behavioural economists discard the inductive reasoning of neoclassical economics in favour of a deductive approach to the study of human behaviour. The contrary would arguably have amounted to “bad psychology.”
There is a rich history of economic thought that pertains to the behavioural peculiarities of man. Even Adam Smith— often depicted as the father of classical economics— observed and articulated some of the “irrational” features of human decision-makers that are today considered as the tenets of contemporary behavioural economics. These include weakness of will, overconfidence, loss aversion, and present-bias (Ashraf et al., 2005, pp.132–134). Most notably, Smith identifies a tension between the “passions of human nature” and a “impartial spectator” in our behaviour (Smith, 1776, pp.1, 12). This is remarkably similar to the dual-process, or two-system, frameworks of human judgement and conduct promoted by psychologists (e.g. Metcalfe and Mischel, 1999), neuroscientists (e.g. Daw et al., 2005), and behavioural economists (e.g. Loewenstein and O’Donoghue, 2004).
In sum, there exists an inherent conflict in human self-control between two internal selves, i.e. an impatient, automatic, and “myopic doer” (the passions), and a patient, deliberate, and “far-sighted planner” (the impartial spectator) (Thaler and Shefrin, 1981, p.392). Stanovich and West’s (2000) equivalent “System 1” and “System 2” modes of processing were famously popularised by Kahneman. The dual-self has been presented as a cogent explanation for various instances of seemingly irrational human behaviour and systematically biased decision-making (Fudenberg and Levine, 2006, p.1449), such as time-inconsistent preferences (Laibson, 1997), Rabin’s (2000) paradox of risk aversion, poor self-control under cognitive load (Shiv and Fedorikhin, 1999), and procrastination (O’Donoghue and Rabin, 2001).
In recent years, a host of neuroscientific studies using brain-imaging appear to corroborate this view of human rationality, or the lack thereof (e.g. McClure et al., 2004; Evans, 2008). The advances in neuroscience have enabled us to lift the veil of the “black box” of human brains, and therefore, human rationality. For instance, it has been found that much of the brain initiates and coordinates automatic processes that are beyond the scope of conscious deliberation and volitional control (e.g. Bargh et al., 1996). Moreover, it has been shown that our conscious behaviour is markedly influenced— and often overridden— by sensitive “affective systems” or emotions (e.g. Adolphs et al., 1995). These findings have given rise to a new wave of economics, or neuroeconomics, that imports realistic and scientifically validated impressions of human decision-making, without having to resort to the restrictive stylisations of rationality of the neoclassical sort (Camerer et al., 2005).
Rationality and methodology
Given the overwhelming evidence in favour of seemingly irrational human behaviour, why did homo economicus persist? Thaler (2016) ascribes it to the relative computational ease of rational choice models [p. 1579]. Camerer et al. (2005, p.9) also provide a historical explanation, and point to a degree of path dependency within the field. It is held that certain fundamental features of economic theory were developed under the assumption that we may never uncover the details of the “black box.” Therefore, “the concepts of ordinal utility and revealed preference eliminated the superfluous intermediate step of positing immeasurable feelings” [p. 10].
This approach has not, however, been without its successes, and has been shown to predict certain human behaviours quite well. A crude but illustrative example is that rational agent models predict human behaviour in a game of tic-tac-toe quite well (Thaler, 2016, p.1579). Similarly, it can be contended that the predictions of rational choice models may approximate the behaviour of experts in certain domains with some success (Friedman, 1953, p.21). As Arrow (1986, p.s385) puts it: “It [rationality in economic theory] is most plausible under ideal conditions.”
However, Arrow also warns that when the ideal conditions are no longer present, “rationality assumptions become strained and possibly even self-contradictory” [p. s385]. The accuracy of these predictions do not scale well with the difficulty or complexity of actual economic problems. It is also apparent that real human behaviour is rarely indicative of expertise. Consequently, the danger for economists is then to derive insights from the naive applications of such models. An additional danger is that economists prioritise form over function, and consequently, become guilty of “getting high on their own supply” of stylised rationality, so to speak.2
This does not imply that we should throw the baby out with the bathwater, because the different aims of different economic theories may warrant varying degrees of stylised rationality. Given the “flesh-and-blood” limitations on human decision-making highlighted by Camus, rational choice models may serve as guides to optimal conduct for different economic problems despite its shortcomings in predicting actual human behaviour (March, 1978, p.588). In other words, homo economicus may shed light on the course of intelligent action for a given situation, yet may fail to provide a sufficient explanation for real human activity. Seemingly, this distinction corresponds to Keynes’ (1891) division between a “positive” and “regulative” science.
It is submitted that much of the controversy regarding rationality may be mitigated through the due consideration of the divergence in aims of different economic theories. The conflation of these aims can be observed, for example, in the contention surrounding the efficient market hypothesis.3 It posits— somewhat strongly— that the stock prices in a completely rational market, at any time, fully reflect all the available and relevant information, and therefore, would provide accurate signals for resource allocation. However, empirical examinations of this hypothesis have uncovered numerous anomalies in actual markets that hitherto have been understated by the theory’s proponents (Shiller, 2003, p.84). These anomalies have partly been explained by the pervasiveness of cognitive bias and other forms of systematic irrationality in human judgement by critics such as De Bondt and Thaler (1985, 1989).
Hence, it is worthwhile to highlight a similar dichotomy between types of theories of human behaviour, namely behavioural and prescriptive (or normative) (March, 1978, p.588). The former seeks to intuit the actual behaviour of human agents, whilst the latter seeks to prescribe optimal agent behaviour. The methodologies of the respective types of theory have developed separately over time, and thus, account for different versions of the rationality assumption. The argument, therefore, is that much of the aforementioned contention may be circumvented if the authors (and critics) of a theory assume (grant) a degree of rationality that is commensurate with the respective aims of behavioural or prescriptive theories. In this light, the efficient market hypothesis could be viewed as a purely normative theory of market behaviour, and correspondingly, should assume an absolute version of rationality.
This raises the issue with respect to the predictive capabilities of economic theories of human behaviour, and the bearing that assumptions of rationality may have on it. Through the benefits of the aggregation of a large number of past observations, one may expect to approximate actual future human behaviour with predictions that rely on assumptions of purely rational agents (March, 1978, p.588). These benefits, however, are not accrued when forecasts are only derived from a relatively small sample. Moreover, as the behavioural economists remind us, people are prone to systematic cognitive biases that cannot always be squared with the predictions of empirical economics. It is not clear whether economic modellers can always anticipate, and in turn, compensate for the direction and extent of such biases. Therefore, it is not necessarily the case that individuals’ idiosyncratic behaviours can be aggregated away nor properly adjusted for. This casts doubt on whether economists can forecast human behaviour with exceeding accuracy using any stylisation of rationality.
Yet, it is held that such empirical tools may still be very useful. The assumption of rationality in economics, as with many other assumptions, may never be a perfect description of reality. In the spirit of Friedman (1953), the rationality assumption— whether absolute or relative— may still serve us well as “sufficiently good approximations for the purpose at hand” [p. 15]. In other words, varying degrees of rationality may be suitable for theories with positive or normative ends, the sufficiency of which ought to be determined through ex post empirical evaluation of the predictions it yields.
However, Friedman’s statement ought to be qualified. This work does not advocate for an agnosticism with respect to the assumptions that underlie our theories of human behaviour. Instead, it argues in favour of a complementarity and supplementarity between different theoretic models, whilst cognisant of the fact that not all theories are equal. Coupled with a forthrightness about the shortcomings of each model and its underlying assumptions, a range of theories could be put to good use. It should be taken for granted that theoretical models of rational optimisers are no more than “metaphors for the world around us” (Shiller, 2003, p.96), and are not always adequate descriptions of “non-expert” human behaviour (Thaler, 2016, p.1589).
Conclusion
This essay’s objective was to shed light on the development of the use and interpretation of rationality in the field of economics. It illustrates a progression from an axiomatic and inductive approach to a data-driven and deductive approach to the study of behaviour. Accordingly, the stylization of rationality has evolved from the rational utility maximiser to that of a boundedly rational agent with a dual-self. This shift has been accommodated by the increasing incorporation of insights from other scientific disciplines, and in turn, lends credence to the remarkable observations of irrationality by Adam Smith.
Given its computational ease, and the antecedent interpretation of the brain as a “black box,” rational agent models have persisted in economic theory. A case is made for their continued use, however the dangers of their naive application and the inclination towards infinite regress are highlighted. It is also argued that a due respect for the divergent aims of behavioural and normative theories of behaviour may moderate some of the controversy surrounding rationality in economics. Finally, this paper advocates for the sophisticated use of a host of complementary, yet imperfect, theoretic models of human rationality and behaviour.