Economics’ massive magnet
Behavioural economics challenges orthodox economics theory and its foundational assumptions regarding human behaviour, its institutional underpinnings, its poor prediction power, and its intrinsic non-falsifiability. In orthodox theory, economic agents are assumed to be fully rational and completely informed. It’s not that they do know everything, but that they can know everything and there are means to learn – epistemology – and they know how to make the best choices for themselves (even if only probabilistically, and even if the choice (sic!) of the precise foundations of the theory of probability that underpins expected utility maximisation is colourfully ad hoc).
Individuals are assumed to have underlying orders of preference for all the alternatives which are knowable, although the means of getting to know them is never specified. These rational preferences are often represented by a utility function, which is assumed to be well-behaved. The “non-satiation” assumption promises that the satiation point will never be reached, at least in the economic domain. Thus, the individuals are always in a state where “more is better”.
Behavioural economics originated, almost fully developed, during the 1950s, and can be classified into at least two streams – Classical and Modern. We would argue that Classical behavioural economics (CBE), pioneered by Herbert Simon (1953), presents a more radical break with the tradition than Modern behavioural economics (MBE) originating in work by Ward Edwards (1954), respectively. The two streams have different methodological, epistemological and philosophical aspects.
First, MBE assumes economic agents maximising utility with respect to an underlying preference order – to which “an increasingly realistic psychological underpinning” is attributed. The “realistic psychological underpinning”, however, is not itself based on any computational foundation, in contrast to Classical behavioural economics, in which the cognitive psychology of choice was intrinsically constrained by a machine model of computation. CBE assumes no underlying preference order. An economic agent’s decision-making behaviour, at any level and against the backdrop of every kind of institutional setting, is subject to bounded rationality and exhibits “satisficing” behaviour – a word Herbert Simon coined from “satisfy” and “suffice” to describe a strategy for reaching a decision the decider finds adequate, even if it’s not optimal in theory. Put another way, MBE remains within the orthodox neoclassical framework of optimisation under constraints; CBE is best understood in terms of decision.
Second, MBE concerns the behaviour of agents and institutions in or near equilibrium; CBE investigates disequilibrium or non-equilibrium phenomena.
Third, MBE accepts mathematical analysis of (uncountably) infinite events or iterations, infinite horizon optimisation problems and probabilities defined over s-algebras and arbitrary measure spaces; CBE only exemplifies cases which contain finitely large search spaces and constrained by finite-time horizons.
There is no doubting the success of MBE. You could characterise it as a massive magnet which attracts different resources, new tools and ways of explanations. In fact you could almost claim that MBE has already become a new mainstream economics, as a consequence of it playing the role of a revised approach of orthodox economics rather than an alternative approach. CBE on the other hand, is developed on completely different grounds from MBE.
MBE is fostered by orthodox economic theory, game theory, mathematical finance theory and recursive methods (Not, however, recursive in the rigorous sense of recursion theory, which forms a key foundation in the development of classical behavioural economics), experimental economics and neuroeconomics, computational economics and subjective probability theory. It preserves the doctrine of utility maximisation and does not go beyond it or discard it (the consumer tries to get the most value possible from the smallest amount of money). Though the behavioural models do consider more realistic psychological or social effects, economic agents are still assumed to be optimising agents, whatever the objective functions may be. In other words, MBE is still within the ambit of the neoclassical theories, or is in some sense only an extension of traditional theory, replacing and repairing the aspects which proved to be contradictory.
CBE is based fundamentally on a model of computation – hence, computable economics – computational complexity theory, nonlinear dynamics and algorithmic probability theory. Unlike MBE, CBE does not try to endow the economic agent with a preference order which can be represented by utility functions; nor do equilibria or optimisation play any role in the activation of behavioural decision-making by CBE agents.
Classical behavioural economics exploits the powerful notion of “bounded rationality” proposed by Simon in 1953. Simon’s definition of bounded rationality encapsulates different notions, such as limited attention, limited cognitive capacity of computation sequential decision-making, and satisficing. For Simon, it is not evident and admissible to assume that human beings are able to exhaust all the information and make the “best” choice out of it. To put it simply: Simon took the limits of human cognition into account and devised mathematical means of describing the roles of memory, experience and intuition in solving problems. His agents do not think in terms of infinite horizon optimisations (nobody in their right mind would!) rather they try to make good decisions for only the near future, but with long-term targets in mind.
“Behavioural economics: Classical and modern”, Ying-Fang Kao & K. Vela Velupillai, (2013): Behavioural economics: Classical and modern, The European Journal of the History of Economic Thought, DOI: 10.1080/09672567.2013.792366
 Whether there are ‘unknowable alternatives’ is never clearly specified – especially when the set of alternatives has the cardinality of the continuum, implying the invoking of some form of the axiom of choice, even in the routine implementation of an optimisation exercise. As a result, ‘the means of getting to know them’ cannot be specified in any constructive way.
 See, however, the caveat about probability in the first paragraph.