Авторы: 147 А Б В Г Д Е З И Й К Л М Н О П Р С Т У Ф Х Ц Ч Ш Щ Э Ю Я

Книги:  180 А Б В Г Д Е З И Й К Л М Н О П Р С Т У Ф Х Ц Ч Ш Щ Э Ю Я


5.3.1 The rational expectations hypothesis

One of the central tenets underlying new classical macroeconomics is the

rational expectations hypothesis (REH) associated with the work of John

Muth (1961) initially in the context of microeconomics. It is, however, interesting

to note that Keuzenkamp (1991) has suggested that Tinbergen was a

precursor to Muth, having presented a model of rational expectations nearly

30 years earlier. We should also note that it was Alan Walters (1971) who first

applied the idea of what he called ‘consistent expectations’ to macroeconomics.

However, it was John Muth’s (1961) seminal paper that proved to be most

influential on the research of the young new classical Turks during the early

1970s. In his seminal article, Muth suggested ‘that expectations since they

are informed predictions of future events are essentially the same as the

predictions of the relevant economic theory’.

Expectations, which are subjective, are fundamental to the behaviour of

economic agents and all economic activities have an informational/expectational

dimension. For example, expectations of the future value of economic variables

will clearly influence demand and supply decisions. As Carter and Maddock

(1984) note, ‘since virtually all economic decisions involve taking actions now

for uncertain rewards in the future, expectations of the future are crucial in

decision making’. An obvious example where expectations of inflation will

influence behaviour concerns wage negotiations between trade unions and

employers. Should a trade union negotiator underestimate the rate of inflation

prevailing over the period of the negotiated wage contract, then workers are

likely to find that they have suffered a nominal wage increase, but a real wage


An expectation of the future value of some key economic variable need not

be confined to a single predicted value but can more realistically take the

form of a probability distribution of outcomes. Therefore, there are two key

questions facing macroeconomists with respect to incorporating expectations

into macroeconomic models:

1. how do individuals acquire, process and make use of information in

order to form expectations of key variables?

2. what form of expectations hypothesis should we use in macroeconomic


During the 1970s, the rational expectations hypothesis replaced the adaptive

expectations hypothesis as the dominant way of modelling endogenous expectations

(in his General Theory, published in 1936, Keynes had stressed the

importance of expectations for understanding macroeconomic instability, but

in Keynes’s theory expectations were exogenous, being driven by ‘animal

spirits’; see Chapter 8 and Keynes, 1937). One great appeal of the rational

expectations hypothesis is that alternative (non-rational) hypotheses of expectations

formation involve systematic errors, a situation that does not sit

comfortably with the rational calculating agents that populate orthodox neoclassical


The rational expectations hypothesis has over the years been presented in

the literature in a number of different forms and versions (see Redman,

1992). At the outset it is important to note the distinction between weak and

strong versions of the hypothesis. The main idea behind the weak version of

the hypothesis is that, in forming forecasts or expectations about the future

value of a variable, rational economic agents will make the best (most efficient)

use of all publicly available information about the factors which they

believe determine that variable. In other words, expectations are assumed to

be formed ‘rationally’ in line with utility-maximizing behaviour on the part

of individual economic agents. For example, if economic agents believe that

the rate of inflation is determined by the rate of monetary expansion, they

will make the best use of all publicly available information on rates of

monetary expansion in forming their expectations of future rates of inflation.

The strong version of the rational expectations hypothesis is captured in the

above quotation taken from Muth’s (1961) article and it is the Muthian

version that has been taken up by leading exponents of the new classical

school and incorporated into their macroeconomic models. In the Muthian

‘strong’ version, economic agents’ subjective expectations of economic variables

will coincide with the true or objective mathematical conditional

expectations of those variables. Using the example of economic agents’ expectations

of inflation (P˙t ),

e the rational expectations hypothesis may be

expressed algebraically in the following way:

where ˙Pt is the actual rate of inflation; E(Pt t ˙ |−1) is the rational expectation

of the rate of inflation subject to the information available up to the previous

period (t−1). It is important to emphasize that rational expectations does not

mean that agents can foresee the future exactly. Rational expectations is not

the same as perfect foresight. In order to form a rational expectation of

inflation, agents will need to take into account what they believe to be the

‘correct’ macroeconomic model of the economy. Agents will make errors in

their forecasts, since available information will be incomplete. Indeed, this is

an essential element of Lucas’s monetary surprise model – see sections 5.3.3

and 5.5.1. However, such forecast errors will be unrelated to the information

set at the time the expectation (for example of inflation) was formed. With

rational expectations, agents’ expectations of economic variables on average

will be correct, that is, will equal their true value. Furthermore, the hypothThe

esis implies that agents will not form expectations which are systematically

wrong (biased) over time. If expectations were systematically wrong, agents

would, it is held, learn from their mistakes and change the way they formed

expectations, thereby eliminating systematic errors. More formally, the strong

version of the rational expectations hypothesis implies that:

e = expected rate of inflation from t to t + 1; ˙Pt = actual rate of

inflation from t to t + 1; and t = random error term, which (i) has a mean of

zero, and (ii) is uncorrelated with the information set available at the time

when expectations are formed, otherwise economic agents would not be fully

exploiting all available information. In summary, the forecasting errors from

rationally formed expectations will (i) be essentially random with a mean of

zero; (ii) be unrelated to those made in previous periods, revealing no discernible

pattern: that is, they will be serially uncorrelated over time; and (iii)

have the lowest variance compared to any other forecasting method. In other

words, rational expectations is the most accurate and efficient form of expectations


The rational expectations hypothesis contrasts with the adaptive expectations

hypothesis initially used by orthodox monetarists in their explanation of expectations-

augmented Phillips curve (see Chapter 4, section 4). In the adaptive

expectations hypothesis, economic agents base their expectations of future

values of a variable (such as inflation) only on past values of the variable

concerned. One of the main problems with this ‘backward-looking’ approach to

forming expectations is that, until the variable being predicted is stable for a

considerable period of time, expectations formed of it will be repeatedly wrong.

For example, following the discussion of Chapter 4, section 4.3.2, on the

accelerationist hypothesis, if unemployment is held below the natural rate,

inflation will accelerate and inflation expectations will be biased in a downward

direction. This problem results from (i) the assumption that economic agents

only partially adjust their expectations by a fraction of the last error made; and

(ii) the failure of agents to take into consideration additional information available

to them other than past values of the variable concerned, despite making

repeated errors. In contrast, in the ‘forward-looking’ approach, rational expectations

are based on the use of all publicly available information, with the

crucial implication of the strong version of the hypothesis being that economic

agents will not form expectations which are systematically wrong over time;

that is, such expectations will be unbiased.

A number of criticisms have been raised against the rational expectations

hypothesis and we now consider three common ones. The first of these

concerns the costs (in time, effort and money) of acquiring and processing all

publicly available information in order to forecast the future value of a

variable, such as inflation. It is important to note that the weak version of the

hypothesis does not require, as some critics have suggested, that economic

agents actually use ‘all’ publicly available information. Given the costs involved

in acquiring and processing information, it is unlikely that agents

would ever use all publicly available information. What proponents of the

weak version of the hypothesis suggest is that ‘rational’ economic agents will

have an incentive to make the ‘best’ use of all publicly available information

in forming their expectations. In other words, agents will have an incentive to

use information up to the point where the marginal benefit (in terms of

improved accuracy of the variable being forecast) equals the marginal cost (in

terms of acquiring and processing all publicly available information). In this

case, expectations would be less efficient than they would be if all available

information were used. Furthermore, the weak version of the hypothesis does

not require, as some critics have suggested, that all individual agents directly

acquire and process available information personally. Economic agents can

derive information indirectly from, for example, published forecasts and

commentaries in the news media. Given that forecasts frequently differ, the

problem then arises of discerning which is the ‘correct’ view.

A far more serious objection concerns the problem of how agents actually

acquire knowledge of the ‘correct’ model of the economy, given that economists

themselves display considerable disagreement over this. The issue of

whether individual agents operating in decentralized markets will be able to

‘learn’ the true model of the economy has been the subject of considerable

debate (see, for example, Frydman and Phelps, 1983; Evans and Honkapohja,

1999). With regard to this particular criticism, it is important to note that the

strong version of the hypothesis does not require that economic agents actually

know the correct model of the economy. What the hypothesis implies is

that rational agents will not form expectations which are systematically wrong

over time. In other words, expectations, it is suggested, will resemble those

formed ‘as if’ agents did know the correct model to the extent that they will

be unbiased and randomly distributed over time. Critics of the hypothesis are

not, however, convinced by arguments such as these and suggest that, owing

to such problems as the costs of acquiring and processing all available information,

and uncertainty over which is the correct model, it ‘is’ possible for

agents to form expectations which are systematically wrong. There is some

evidence that agents do make systematic errors in expectations (see, for

example, Lovell, 1986).

A third important criticism, associated in particular with the Post Keynesian

school, relates to the problems of expectations formation in a world of fundamental

uncertainty. To Keynesian fundamentalists, a major achievement of

Keynes was to place the problem of uncertainty at the centre stage of macrThe

oeconomics. In the Post Keynesian vision, the world is non-ergodic; that is,

each historical event is unique and non-repetitive. In such situations the rules of

probability do not apply. We are in a world of ‘kaleidic’ change and fundamental

discontinuities (Shackle, 1974). Accordingly, Post Keynesians argue that it

is important to follow both Keynes (1921) and Knight (1933) and distinguish

between situations involving risk and situations involving uncertainty. In situations

of risk the probability distribution is known. In contrast, in situations of

uncertainty there is no possibility of formulating any meaningful probability

distribution. Because the rational expectations hypothesis assumes that economic

agents can formulate probability distributions of outcomes of various

economic changes and situations, it belongs to the world of risk. In new

classical models the problem of fundamental uncertainty is ignored since Lucas

(1977) interprets business cycles as repeated instances of essentially similar

events. Hence, in Lucas’s ergodic world, meaningful probability distributions

of outcomes can be gauged by intelligent and rational economic agents. Unfortunately,

according to Post Keynesians, the real world is one characterized by

fundamental uncertainty and this means that conclusions built on models using

the rational expectations hypothesis are useless. Likewise, the Austrian school

are also very critical of the rational expectations hypothesis (see Snowdon et

al., 1994, and Chapters 8 and 9).

The various influences on expectations have recently been investigated by

the Bank of England (2003). Reporting the results of a recent ‘inflation

attitudes survey’ the Bank of England finds the following interesting results:

1. disaggregating the data reveals that different people and groups have

different attitudes to inflation;

2. the expectations of ‘professional’ groups cluster around the mean expectation;

3. younger respondents have lower expectations of inflation than older


4. mortgage holders have lower inflation expectations than respondents

who rent accommodation;

5. people in the south of Britain have higher expectations of inflation than

those living in the north; and

6. lifetime experience of inflation influences expectations of inflation.

Thus expectations of inflation are influenced by age, geographical location,

education and occupation, and housing status. Clearly those old enough to

have lived through the ‘Great Inflation’ of the 1970s have not been entirely

able to remove that experience from their judgement.

Notwithstanding these criticisms, during the 1970s there was undoubtedly

a ‘rational expectations revolution’ in macroeconomics (Taylor, 1989; Hoo230

ver, 1992). However, it should be noted that Muth’s idea was not immediately

taken up by macroeconomists, maybe because during the 1960s the orthodox

Keynesian model was ‘the only game in town’. It took almost ten years

before Lucas, Sargent and other leading new classical economists began to

incorporate the hypothesis into their macroeconomic models.

Evidence of this lag can be gleaned from citation counts for Muth’s (1961)

paper. In an interesting comparison of the relative influence of Muth’s paper

with that of Axel Leijonhufvud’s (1968) famous book, On Keynesian Economics

and the Economics of Keynes (see Chapter 2), Backhouse (1995) has

shown how during the 1970s and 1980s citations of Muth’s paper exploded

while citations of Leijonhufvud’s book declined as interest in Keynesian

economics waned (see Snowdon, 2004a). While Leijonhufvud’s book had an

immediate impact, but ultimately failed to transform macroeconomics in the

direction of coordination failure stressed by Leijonhufvud, in contrast, Muth’s

paper got off to a slow start but ultimately played a key role in transforming

macroeconomics (see Leijonhufvud, 1992, 1993, 1998a, 1998b on the need

for macroeconomics to reconsider, among many other things, the coordination

question in macroeconomics).

One final point is worth making. The use of the word ‘rational’ in the

presentation of the hypothesis proved to be an important ‘rhetorical’ weapon

in the battle to win the minds of macroeconomists during the 1970s. As Barro

(1984) has pointed out:

One of the cleverest features of the rational expectations revolution was the

application of the term ‘rational’. Thereby, the opponents of this approach were

forced into the defensive position of either being irrational or of modelling others

as irrational, neither of which are comfortable positions for an economist.

For a more detailed discussion of the rational expectations hypothesis and its

application in macroeconomics, the reader is referred to Begg (1982); Carter

and Maddock (1984); Shaw (1984); Attfield et al. (1985); Redman (1992);

Sheffrin (1996); and Minford (1997). On the use of rhetoric in new classical

economics, see Backhouse (1997a).