2. Literature Review
2.1 Parent disciplines, fields and models
2.1.2 Rational Expectations Hypothesis and Behavioural Economics
Simon (1984, p. 36) considers REH one of the three foundations for neoclassical economics. This section discusses REH for two reasons. One, it forms a benchmark for AIE. Two, REH is ubiquitous in neoclassical theory and underpins the EMH, GET, various CGE models, the
‘permanent-income’ and ‘life-cycle’ theories of consumption and ‘policy ineffectiveness proposition’. EMH, GET and CGE models are discussed in sections 2.1.3, 2.1.4 and 2.1.5, respectively, to provide motivation for the development of AIE. Additionally, this section discusses behavioural economics for two primary reasons. One, it shows that REH is derived using assumptions that are empirically inconsistent with the way individuals behave. Two, it provides support for the adaptive-expectations model, which forms a benchmark for AIE and also a component of AIE.
Muth (1960; 1961) introduces REH, which embodies all three neoclassical assumptions discussed in section 2.1.1. REH assumes that a homogenous or representative agent (methodological individualism) uses ‘rational choice’ (Von Neumann & Morgenstern 1944) to maximise their utility (methodological instrumentalism) that assumes there is just one equilibrium (methodological equilibration). Billari (2006, p. 2) describes the rational expectations paradigm that assumes homogenous economic agents who know their preferences, have unlimited computing power to calculate the optimal solution, and have perfect knowledge of all data relevant to calculate the optimal solution. Sargent (2008, p. 1) asserts in rational expectations that outcomes do not differ systematically (i.e., regularly or predictably) from what people expect them to be. This assertion provides for empirical falsification and is used in section 3.7 to make REH operational. Lovell (1986, p. 112) observes that REH explicitly avoids modelling the process by which expectations are formed. This observation does present a gap for behavioural economics.
In the following sections, section 2.1.2.1 discusses criticism of REH from behavioural economics, while section 2.1.2.2 discusses falsification of REH using econometrics.
2.1.2.1 Criticisms of REH from Behavioural Economics
Van der Sal (2004, p. 432) notes that for nearly half a century rational choice theories have been tested against individual behaviour but repeatedly the underlying assumptions and predictive value appear descriptively false. Sargent (2008, p. 1) claims that believers in REH assume that people behave in a way to maximise their utility or profits. Thaler and Mullainathan (2008, p. 1) state that these maximisers in the neoclassical framework ignore virtually all the findings of cognitive and
social psychologists. They highlight three assumptions of REH that are at odds with the findings, unbounded rationality, willpower and selfishness.
Addressing unbounded rationality, Simon (1972; 1979) criticises REH for its assumption that people have unlimited information processing ability. He introduces the term ‘bounded rationality’
to describe a more realistic concept of human problems solving ability. It is perfectly rational for people to use ‘rules of thumb’ to make decisions, thus make the best use of their limited cognitive abilities. He suggests that economic models that do not incorporate some form of bounded rationality are just bad economics. Sargent (2008, p. 1) claims that in rational expectations outcomes do not differ systematically (i.e., regularly or predictably) from what people expect. However, Tversky and Kahneman (1974) and Kahneman and Tversky (1979) provide extensive empirical evidence to show regular and predictable biases in people forming expectations. Importantly these biases do show that people developed expectations in a profoundly different way to that supposed by REH. Their empirical evidence supports Simon’s ‘rules of thumb’ concept. The AIE model uses bounded rationality and the agents use rules to form their expectations. The
‘rules of thumb’ for AIE are developed from adaptive-expectations and interactive-expectations and are discussed in sections 2.1.8 and 2.1.9, respectively. Section 2.1.3.3 further discusses Kahneman and Tversky.
In addressing unbounded willpower, the representative agent in REH has perfect self-control; however it is evident that humans do lack self-control. Most at some time have eaten, drunk or spent too much and have exercised, saved or worked too little. The healthy eating and exercise initiatives and superannuation scheme show that governments recognise bounded willpower.
Addressing unbounded selfishness, REH does not specifically preclude altruistic acts. However, it does assume homogenous self-interested optimising agents. The free-rider problems in economics and game theory particularly would support this approach. However, at odds with unbounded selfishness there are instances where people do contribute to the public good, even if, their own private welfare is not improved. For instance, people give to charities and people refuse offers that they perceive as unfair in ultimatum games.
‘Permanent-income’ and ‘Life-cycle’ theories
These neoclassical theories have drawn criticism from two directions. First, O'Donoghue and Rabin (1999, p. 125) find that people procrastinate when they make investments for their retirement even though they know the importance of the decisions. People use a hyperbolic discounting function
inconsistent with utility maximisation assumed in REH. Second, using the predictions of the life- cycle hypothesis Banks, Blundell and Tanner (1998) find that people fail to save sufficiently for their retirement.
However, Keen (2009) notes, ‘von Neumann and Morgenstern specifically warned against having their model of consumer behaviour interpreted as it has in fact been interpreted in the economic literature, both by those who misapplied it--using it to model behaviour in one-off choices as in financial decisions--and those who criticised it.’ As Von Neumann and Morgenstern (1944 p.19) state, ‘Probability has often been visualized as a subjective concept more or less in the nature of estimation. Since we propose to use it in constructing an individual, numerical estimation of utility, the above view of probability would not serve our purpose. The simplest procedure is, therefore, to insist upon the alternative, perfectly well founded interpretation of probability as frequency in long runs.’ Furthermore, Keen (2009) states if these test were applied as multiple repeats of the same choice, the tests that have ‘contradicted’ von Neumann & Morgenstern, would have instead confirmed von Neumann & Morgenstern. It is the misapplication of Neumann & Morgenstern’s model in terms of subjective rather than objective probability by economists, which is at fault and not von Neumann & Morgenstern. This misapplication of Neumann & Morgenstern’s model beyond the domain of its assumptions is a major flaw in REH.’
2.1.2.2 Falsification of REH using Econometrics
REH uses the assumption in equation (2–1) that E(ε) = 0, which is basically that a prediction is an unbiased estimate of an actual value.
ε = P – A (2–1)
Where P = predicted value A = actual value
Muth (1961) implements the assumption in equation (2–1) and uses the assumptions in equation (2– 2) that the forecast error is uncorrelated with the predicted value.
A = β0 + β1 P + ε (2–2)
In equation (2–2), Lovell (1986, p. 115) notes that REH asserts that variance of actualisation will exceed the variance of predictions. Additionally, ‘the prediction error must be uncorrelated with the entire set of information that is available to the forecaster at the time the prediction is made.’
For instance, Lovell (1986, p. 120) notes that the prediction of some forecasters may be characterised as rational, in other examples the assumption of the rationality is clearly violated. In one example Lovell (1986, p. 115) finds that an alternative regression to that in equation (2–2) contains lagged actualisations and provides a much better fit to the data, thus implying firms are not REH rational because they do not make use of all the available data to make a prediction.
Lovell (1986, p. 115) notes that some firms are perennial optimists, generally overestimating the future, while others are perennial pessimists, usually underestimating sales volume. In aggregate the underestimates of the pessimists roughly cancel out the overestimates of the optimists. The offsetting of systematic error explains why the aggregates of prediction data are more accurate than the predictions of individual firms. This aggregation effect becomes apparent in the EMH, when comparing price movements in individual stock with market indices, see section 2.1.3.1.
Lovell (1986, p. 111) notes that there is disagreement over whether REH should be tested empirically because REH is an assumption, therefore not suitable for empirical analysis. The REH as an assumption argument avoids direct falsification. For instance, Prescott (1977, p. 30) claims, ‘Like utility, expectations are not observed, and surveys cannot be used to test the rational expectations hypothesis. One can only test if some theory, whether it incorporates rational expectations or, for that matter, irrational expectations, is or is not consistent with observations.’
However, theories that incorporate REH have been falsified, for instance the EMH, GET and DSGE. Sections 2.1.3, 2.1.4 and 2.1.5 discuss these theories, finding that the EMH incorrectly models volatility in share market prices and fails to model the momentum effect. The SMD Theorem proves that the basic axioms for GET are logically incongruent; axioms REH also embodies. The stylised facts from the DSGE models fail to match those of the business cycle. Despite the presence of REH in all these conjoint failures, the Duhem–Quine Thesis can still be invoked to avoid falsification, as discussed in section 2.1.1.3. Using such a falsification avoidance construct, Popper (1972a; 1972b; 1972c) would label REH as unscientific. Additionally, Lakatos (1976) would consider REH part of the ‘hard core’ of the neoclassical framework and components of the larger theories as part of a ‘protective belt’. Falsification avoidance is a sign of a degenerative scientific research program. Lovell (1986, p. 122) states that there is sufficient empirical evidence to suspend belief in REH and calls for the empirical testing of REH against
alternatives. This thesis empirically tests REH against the adaptive-expectations and the AIE models.
2.1.2.3 Summary
REH and rational expectations appear to be normative assumptions and not explanatory or predictive theories. As explanatory or predictive theories, they are inadequate in the sense that people frequently fail to follow their prescription and as a normative theories they are indeterminate, as often prescriptions are ambiguous (Elster 1989). Lovell (1986, p. 120) concludes that expectations are a rich and varied phenomenon, which are not adequately captured in REH.