24
This article was downloaded by: [Temple University Libraries] On: 18 November 2014, At: 13:27 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK Journal of Economic Methodology Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/rjec20 A structure of the consumption function HsiangKe Chao a a National Tsing Hua University , Taiwan E-mail: Published online: 12 Jun 2007. To cite this article: HsiangKe Chao (2007) A structure of the consumption function, Journal of Economic Methodology, 14:2, 227-248, DOI: 10.1080/13501780701394102 To link to this article: http://dx.doi.org/10.1080/13501780701394102 PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub- licensing, systematic supply, or distribution in any form to anyone is expressly

A structure of the consumption function

Embed Size (px)

Citation preview

Page 1: A structure of the consumption function

This article was downloaded by: [Temple University Libraries]On: 18 November 2014, At: 13:27Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH,UK

Journal of EconomicMethodologyPublication details, including instructions for authorsand subscription information:http://www.tandfonline.com/loi/rjec20

A structure of the consumptionfunctionHsiang‐Ke Chao a

a National Tsing Hua University , Taiwan E-mail:Published online: 12 Jun 2007.

To cite this article: Hsiang‐Ke Chao (2007) A structure of the consumption function,Journal of Economic Methodology, 14:2, 227-248, DOI: 10.1080/13501780701394102

To link to this article: http://dx.doi.org/10.1080/13501780701394102

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all theinformation (the “Content”) contained in the publications on our platform.However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness, orsuitability for any purpose of the Content. Any opinions and views expressedin this publication are the opinions and views of the authors, and are not theviews of or endorsed by Taylor & Francis. The accuracy of the Content shouldnot be relied upon and should be independently verified with primary sourcesof information. Taylor and Francis shall not be liable for any losses, actions,claims, proceedings, demands, costs, expenses, damages, and other liabilitieswhatsoever or howsoever caused arising directly or indirectly in connectionwith, in relation to or arising out of the use of the Content.

This article may be used for research, teaching, and private study purposes.Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly

Page 2: A structure of the consumption function

forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

Dow

nloa

ded

by [

Tem

ple

Uni

vers

ity L

ibra

ries

] at

13:

27 1

8 N

ovem

ber

2014

Page 3: A structure of the consumption function

A structure of the consumption function

Hsiang-Ke Chao

Abstract It is claimed in the structural realism in philosophy of science thatscientists aim to preserve the true structure, represented by the equations in theirmodels. We reinterpret structural realism as a doctrine involving representation.Proving the existence of a representation theorem secures the problem of lackingindependent criteria for identification between structure and non-structure. Thispaper argues that a similar realist view of structure can be found in the theory ofconsumption in which the Fisherian framework of intertemporal choices isregarded as the true structure of the consumption function. Unlike the passivestrategy of inducing the structure contained in structural realism, economistsdefine structure in terms of invariance under intervention. Such a definition servesas a crucial device to examine and develop models for the adequacy ofrepresenting the structure of the consumption functions.

Keywords: consumption function, Euler-equation approach, invariance,representation, structure, structural realism

JEL Classification: B22, B41, C50, E21

1 INTRODUCTION

The ‘stylized’ history of macroeconomic theories of consumption is usually

presented as follows. In the ‘fundamental psychological law’ in his General

Theory (Keynes 1936), John Maynard Keynes argued that the level of current

consumption is determined by current income. Keynes’s absolute income

hypothesis also implies that the marginal propensity to consume is between 0

and 1, and is less than the average propensity to consume; the averagepropensity to consume falls, and the average propensity to save rises with

income, so ‘rich people save more’. While the absolute income hypothesis

accounted for short-run phenomena, it was contradicted by long-run

phenomena, such as stable average propensities to consume and to save,

equivalence of marginal and average propensities to consume, that were found

by Simon Kuznets in his (1946) empirical survey of the US time series data

since 1896. In order to reconcile these findings, James Duesenberry’s (1949)

relative income hypothesis assumed that consumptions are interdependentamong different people and for the same people at different periods.1

Journal of Economic Methodology ISSN 1350-178X print/ISSN 1469-9427 online

# 2007 Taylor & Francis http://www.tandf.co.uk/journals

DOI: 10.1080/13501780701394102

Journal of Economic Methodology 14:2, 227–248 June 2007

Dow

nloa

ded

by [

Tem

ple

Uni

vers

ity L

ibra

ries

] at

13:

27 1

8 N

ovem

ber

2014

Page 4: A structure of the consumption function

In the 1950s Milton Friedman and Franco Modigliani independently

proposed a new theory of consumption. Friedman’s permanent income

hypothesis was published in his seminal 1957 book A Theory of the

Consumption Function; Modigliani’s life-cycle hypothesis was established in

a series of papers that he had written with Albert Ando and Richard

Brumberg (Modigliani and Brumberg 1954; Ando and Modigliani 1963).

Both the permanent income hypothesis and the life-cycle hypothesis were

based on Irving Fisher’s intertemporal choices theory and claim that the

level of aggregate current consumption is determined by a long-term

expected income. But after the rational expectations revolution and the

Lucas critique of traditional econometrics (Lucas 1976) – in the 1970s, bothapproaches were criticized for not dealing properly with expectations.

Robert Hall’s famous ‘random-walk’ model (Hall 1978) then incorporated

the idea of rational expectations into the intertemporal choices framework.

In recent consumption models the representative consumers do not smooth

their consumption. But since Hall, the modeling strategy of deriving

consumption functions directly from the first-order Euler equation – the

Euler-equation approach – has come to dominate the theory of consumption

in today’s macroeconomics.Although this stylized history, at least in its early years, has been criticized

for failing to represent the ‘true’ history of the consumption function (e.g.

Thomas 1989), the first half of this stylized history is cited by economic

historians as an example of Thomas Kuhn’s scientific revolutions in

economics (Mayer 1972: 7–8; Hynes 1998). They see the change from

Keynes’s consumption theory to the theories of Friedman and Modigliani as

a Kuhnian paradigm shift. Whether or not the rest of the history can also be

seen in this way, for instance, whether or not the movement that we calledthe ‘rational expectations revolution’ is really a Kuhnian scientific

revolution, the introduction of rational expectations has changed econo-

mists’ view of many fields of macroeconomics, including that of the

consumption function.

Yet, if we take a deep look into the development of the theory of the

consumption function, we find both continuity and discontinuity. In

philosophy of science, realists and anti-realists hotly debated theory change.

Anti-realists conclude from the presence of radical theory change (orparadigm shift) in the history of science that any theory will eventually be

proven false. This is the argument of ‘pessimistic meta-induction’. In

contrast, realists assert that all theories are completely or partially true or

approximately true so that the success of past theories is not accidental. This

‘no-miracles’ argument clashes with the argument of pessimistic meta-

induction. In order to solve the puzzle, philosopher of science John Worrall

(1989) accounts for both assertions in his theory of ‘structural realism’.

Philosophers and historians of science regard Worrall’s structural realism asproviding ‘the best hope for realism’ (Papineau 1996: 13).2 However, one

228 Articles

Dow

nloa

ded

by [

Tem

ple

Uni

vers

ity L

ibra

ries

] at

13:

27 1

8 N

ovem

ber

2014

Page 5: A structure of the consumption function

major problem, as Worrall admits, is that there are only ‘limiting cases’

for structural realism. Whether or not Worrall’s account of structural

realism fully explains the history of science, it does point out scientists’undeniable concern with preserving structure in their scientific practices.

The same concern about structure, as this article is going to demonstrate,

can also be clearly found in the history of the consumption function in

macroeconomics.

This article argues that a methodological position similar to Worrall’s

structural realism can be found in the history of the consumption function.

We call this position ‘realism about structure’.3 Realism about structure

reveals two points that are pertinent to structural realism. The first is theconcept of structure that Worrall does not clearly define. Structure usually

goes beyond the formal–mathematical level that concerns Worrall, and is

particularly construed in terms of invariance in the study of consumption

function. Precisely, the notion of structure in economics is construed in

terms of invariance under intervention. The other point is the relations

between model and structure. Worrall’s structure contains two tiers of

representations. We refer to the discussions in philosophy of science and

economic methodology to demonstrate the concerns of these models, inwhich certain independent criteria are necessary for choosing between

structural and non-structural relations, since both relations could be

preserved during theory change.

2 STRUCTURAL REALISM

Realism has many faces, but the most commonly recognized one is

‘scientific realism’. According to scientific realism, the objects ofscientific knowledge exist independently of scientists’ minds or acts. In

addition, scientific theories are true to the objects. The first assertion

leads to a metaphysical claim for the independent existence of certain

entities while the second one leads to the epistemological principle that we

can know the independent existence of these entities (Papineau 1996; Fine

1998).

Scientific realism asserts that the past and present mature scientific

theories have successfully explained or predicted phenomena because theyare (approximately) true to the world. If these theories were false, then their

empirical success would be a miracle. As Hilary Putnam (1975: 73) puts it,

‘[t]he positive argument for realism is that it is the only philosophy that does

not make the success of science a miracle’. This ‘no-miracles’ argument,

however, was challenged by the historical fact that many theories which

were once successful were later proven false. From this fact, we can induce a

conclusion that presently accepted scientific theories will likewise be proven

false. Hence, this ‘pessimistic meta-induction’ argument (Laudan 1981)rejects scientific realism, and theory can by no means be regarded as true.

A structure of the consumption function 229

Dow

nloa

ded

by [

Tem

ple

Uni

vers

ity L

ibra

ries

] at

13:

27 1

8 N

ovem

ber

2014

Page 6: A structure of the consumption function

To adopt the realist position and to respond to the pessimistic meta-

induction argument, Worrall (1989) appeals to the continuity of ‘structure’

across theory change. He argues that even though pessimistic meta-induction is right about the facts about radical theory changes and the

discontinuity of ‘contents’, there is still a continuity of ‘structure’ or ‘form’

between the old and new theories. Worrall proposes an account of structural

realism, originating with Henri Poincare, Bertrand Russell and Grover

Maxwell among others,4 arguing that in structural realism even though the

new theory replaced the old, the old theory was previously accepted because

it captured the true structure, expressed in mathematical terms. The

evidence of the old theory capturing the true structure is the presence of thesame mathematical equations in the superseding theories. In this way

Worrall modifies the no-miracles argument, asserting that the success of

mature theory is due to its truth to the structure. Worrall’s paradigm

example for structural realism is that of the development of the theory of

light in the nineteenth century, starting with Augustin Fresnel’s successful

re-establishment of the wave theory. Later when James Maxwell’s

electromagnetic theory replaced Fresnel’s, Fresnel’s assumption that the

luminiferous aether – the hypothetical medium for propagation of lightwaves – was rigid was overthrown in favor of James Maxwell’s

electromagnetic field. Yet the equations contained in Fresnel’s theory (also

known as ‘Fresnel’s laws’ or ‘Fresnel’s equations’) were preserved in

Maxwell’s theory (‘Maxwell’s laws or ‘Maxwell’s equations’). Thus Worrall

states:

it seems right to say that Fresnel completely misidentified the nature of

light; but, none the less, it is no miracle that his theory enjoyed the

empirical predictive success that it did; it is no miracle because Fresnel’s

theory, as science later saw it, attributed to light the right structure.

(Worrall 1989: 117; emphasis in original)

As Poincare stated: ‘if the equations remain true, it is because the relations

preserve their reality’ (quoted in Worrall 1989: 118).

However, Worrall later admitted that the Fresnel–Maxwell example was

‘unrepresentative’ because it was a rarity in the history of science for the

mathematical equations of the old theory to be ‘completely taken overintact’ into the new theory (Worrall 1989: 120). But Worrall then argued

that ‘[t]he much more common pattern is that the old equations reappear as

limiting cases of the new’; this means that ‘the old and new equations are

strictly inconsistent, but the new tends to the old as some quantity tends to

some limit’ (Worrall 1989: 120). In this sense, structural realism is concerned

with the existence of approximate rather than complete continuity during

theory change.

Worrall further argues that structural realism is meaningful not onlybecause it reconciles radical theory change with the no-miracle argument,

230 Articles

Dow

nloa

ded

by [

Tem

ple

Uni

vers

ity L

ibra

ries

] at

13:

27 1

8 N

ovem

ber

2014

Page 7: A structure of the consumption function

but also because it can play an active role in the ‘correspondence principle’

of scientific discovery. Some philosophers see the correspondence principle

as evidence of realism (Boyd 1984; see Worrall 1989: 120), and as a heuristicdevice for developing the new theory by inferring the old, as well as the new

theory explaining the empirical success of its predecessor (Worrall 1985). An

example is the argument that Newton arguably deduced his laws from

Kepler’s (Worrall 1985: 314–18). Worrall’s structural realism paper treats

this correspondence principle as operating ‘purely at the mathematical level’

(Worrall 1989: 120; emphasis in original). This means that from the

mathematical equations that express the structure of the old theory that was

once successful, we can infer and construct a new theory, one which containssimilar equations expressing the same structure that is thought to be real.

In sum, Worrall’s account of structural realism requires (1) a preserved

mathematical equation to be passively observed; and (2) an equation that

plays an active heuristic role in developing new theories. Yet Worrall’s

characterization of scientific practices on the importance of maintaining and

using the structure can likewise be found in the history of the consumption

function. In particular, the framework of intertemporal choices is regarded

as the structure for the consumption function.

3 STRUCTURE OF THE CONSUMPTION FUNCTION

In his work on interest, Irving Fisher (1930) developed the concept of

intertemporal choices that he traced to John Rae and Eugene Bohn-Bawerk.

Fisher argues that people tend to prefer present over future goods. This

‘time preference’, also known as ‘human impatience’ or ‘impatience’ (Fisher

1930: 62), in its marginal form, determines the rate of interest as thepremium on the exchange between present and future goods (Fisher 1930:

61). In this light, Fisher’s theory sharply contrasts with Keynes’s

fundamental psychological law, or the absolute income hypothesis, which

asserts that current consumption is only determined by current income at

the aggregate level.5

After Keynes’s revolution, we have two well-known theories of the

consumption function directly derived from Fisher’s intertemporal choices

theory. Friedman referred to Fisher’s intertemporal choices as the ‘puretheory of consumer behavior’ and the building block of his permanent

income hypothesis. Modigliani also wrote that his life-cycle hypothesis was

built on the ‘received theory of consumer’s choice a la Fisher’ (Modigliani

1975: 5). Even though the aggregation problem of deriving the macro-

economic consumption function from an individual’s optimal decisions is

not addressed in either Friedman or Modigliani, the success of the

permanent income and the life-cycle hypotheses was attributed to the

application of the Fisherian framework, in addition to their empiricalsuccesses over the Keynesian theory.6

A structure of the consumption function 231

Dow

nloa

ded

by [

Tem

ple

Uni

vers

ity L

ibra

ries

] at

13:

27 1

8 N

ovem

ber

2014

Page 8: A structure of the consumption function

One crucial element of the framework of intertemporal choices is the

expectations of future income when consumers face uncertainty. Friedman

adopts the theory of adaptive expectations in forming expected income as a

proxy of permanent income. But this is challenged by the hypothesis of

rational expectations, which is seen as a more appropriate account of how

consumers form their expectations. John Muth (1960) was the first to show

the adaptive expectations formula as a special case of the rational

expectations hypothesis. He argues that Friedman’s adaptive expectations

formula only coincides with the rational expectations hypothesis under some

circumstances; otherwise they are inconsistent.7 This point is taken by Lucas

(1976). Lucas, following Haavelmo (1944) and the Cowles Commission

scholars such as Marschak (1953) and Hurwicz (1962), is concerned with

invariant relationships under policy intervention. The ‘Lucas critique’

challenges the standard econometric models which do not exhibit invariant

relationships because agents’ expectations are modified in the face of policy

change. In the case of consumption function, Lucas (1976) shows that

Friedman’s permanent income hypothesis, based on the assumption of

adaptive expectations, incorrectly forecasts the consumption for any

expected policy change. Moreover, the permanent income hypothesis does

not apply to the case of unexpected policy change. Since Friedman’s

permanent income hypothesis does not yield a stable relationship between

consumption and income, Robert Hall suggests that the Lucas critique

states that ‘there is no such thing as a consumption function’ (Hall 1990:

135).

Hall’s famous ‘random walk’ paper (Hall 1978) attempts to

construct a consumption function that satisfies the Lucas critique. Hall’s

approach is compatible with the traditional permanent income and life-cycle

hypotheses, for Hall’s consumption theory contains a Fisherian framework

in which the representative consumer intertemporally allocates his/her

wealth on consumption. But what distinguishes Hall’s consumption

function from previous theories is its accommodation of the Lucas critique.

He introduces expectations to the consumer’s maximization problem in the

face of uncertainty, sets up the problem in which the consumer maximizes

expected utility, and keeps the expected marginal utility constant. Hall’s

aggregate consumption function is derived from the representative

agent’s intertemporal optimizing behavior under the assumptions of

rational expectations and a representative agent. This derivation of the

consumption model is the Euler-equation approach, which refers to the first-

order condition of the representative agent’s intertemporal choices of

consumption.

To illustrate, consider a two-period case, in which the representative

consumer maximizes his/her utility function subject to the intertemporal

budget constraint. The Euler equation is derived as

232 Articles

Dow

nloa

ded

by [

Tem

ple

Uni

vers

ity L

ibra

ries

] at

13:

27 1

8 N

ovem

ber

2014

Page 9: A structure of the consumption function

E1u0

c2ð Þ~1zd

1zr

� �u0

c1ð Þ ð1Þ

where d is the subjective time preference rate, and r is the constant real

interest rate. E1 denotes the conditional expectation given all the

information available at period 1. When the utility function is quadratic,

and d5r, the Euler equation (1) yields:

E1c2~c1 ð2Þ

By the definition of expectations, we have the random walk consumption

c2~c1ze2 ð3Þ

where e2 is a true regression disturbance whose expectation as of period 1 iszero. Equation (3) suggests that consumption is a random walk: the

consumption in this period is only the function of the consumption in the

previous period plus an innovation.

Hall’s random-walk model, as Clive Granger claimed, is easy to state,

easy to understand and easy to test. It is like ‘manna from heaven to

macroeconometricians’ (Granger 1999: 43). Granger’s view is shared with

macroeconomists who apply the Euler-equation approach to consumption

research. The literature cites four reasons why the Euler-equation approach

is the dominant consumption theory in macroeconomics:

First of all, the Euler-equation approach is operationally simpler than the

conventional ‘solved-out’ way, which requires deriving consumption

functions by solving all period-to-period budget constraints. In contrast to

the solved-out approach, the Euler-equation approach only needs to set up a

representative-agent model, write down the Euler-equation, and use the

equilibrium relationships between the expected marginal utility of futureconsumption and the marginal utility of current consumption like equation

(1) to derive the ‘consumption function’.8 Some economists regard the

avoidance of solving the consumer’s optimization problem as the most

appealing element of the Euler-equation approach (Attanasio and Low

2004: 407).

Secondly, the Euler equation is a more useful measuring tool than the

solved-out consumption function. In many empirical studies, the standard

procedure is to log-linearize the Euler equation given that the utility

function is constant-relative-risk-averse (Hansen and Singleton 1983). In

doing so, it allows a focus on directly estimating the ‘structural parameters’

of the consumption function (Attanasio 1998: 21), for example, the elasticity

of intertemporal substitution.9

Thirdly, the random walk consumption highlights the rational expecta-

tions hypothesis and the Lucas critique in many ways. Besides the

representative consumer’s optimization setting, the random-walk modelsuggests that the consumer uses today’s consumption as the best predictor of

A structure of the consumption function 233

Dow

nloa

ded

by [

Tem

ple

Uni

vers

ity L

ibra

ries

] at

13:

27 1

8 N

ovem

ber

2014

Page 10: A structure of the consumption function

future consumption because any available information has been included in

today’s consumption. Hence other variables, particularly current and past

incomes, can be excluded from the consumption function.Finally, the Euler-equation approach preserves the ‘structure’ of the

consumption function. Even though Lucas criticizes the unsustainable

relationships between consumption and income in the traditional consump-

tion function, Hall argues that there still exists something in the Euler-

equation approach that can be considered as structure:

Although Lucas was scornful of existing econometric policy evaluation

models, his message was not completely destructive of all model-building

or empirical research. There are structural relationships in the economy,

but the consumption function is not among them. For consumption, the

structural relation, invariant to policy interventions and other shifts

elsewhere in the economy, is the intertemporal preference ordering.

(Hall 1990: 135; emphasis added)

For Hall, the intertemporal preference ordering, or the Fisherian

intertemporal choices, is the structure of the consumption function because

this relation is invariant to exogenous changes. As mentioned above, theEuler equation per se is not really a consumption function but an

equilibrium condition that can be regarded as a structure of the

consumption function. The final model exhibiting the relationships between

the current consumption and other variables, such as Hall’s random walk,

depends strongly on the auxiliary assumptions that the modeler employs.

(For Hall’s random-walk model, the most crucial assumption is the

quadratic utility function.) In this sense, Hall’s random-walk model is only

a special case of the Euler-equation approach.

4 STRUCTURAL REALIST INTERPRETATION OF THE

CONSUMPTION FUNCTION

In the context of the history of the consumption function and structural

realism, we do not find that Worrall’s structural realism fits. While

Worrall’s strategy requires observing continuity in structure during theory

change, the history of the consumption function shows that thesimultaneous discontinuity in theory and continuity in structure is hard to

find. Economists are unanimous in seeing the discontinuity between

Keynesian theory and the theories of Friedman and Modigliani. But

usually, they refer to Hall’s theory as a life cycle–permanent income

hypothesis – in fact Hall himself (1978) regards his random-walk model as a

justification of the Friedman–Modigliani theory. In this sense, it seems that

no theory change was induced by the rational expectations revolution.

Yet, consider the consumption theories before and after the rationalexpectations revolutions; if we regard adaptive and rational expected

234 Articles

Dow

nloa

ded

by [

Tem

ple

Uni

vers

ity L

ibra

ries

] at

13:

27 1

8 N

ovem

ber

2014

Page 11: A structure of the consumption function

marginal utilities of future consumptions as different substances because of

drastic changes in the nature and the formulation of expectations, then there

is the change from Friedman and Modigliani to Hall. Therefore the Eulerequation retains the structure of the consumption function during this

theory change. This interpretation is not incongruous in the sense that we

often construe the introduction of rational expectations as a ‘revolution’.

Whether or not this is parallel to Kuhn’s ‘scientific revolution’, it marks a

sea change in macroeconomics that could parallel the ‘Keynesian revolu-

tion’ which is more commonly regarded as fitting Kuhn’s account. Gilbert

(1991: 156) sees it in this way: ‘The [rational expectations hypothesis] is seen

as an improvement over adaptive expectations in the same way that the[permanent income hypothesis] was seen as an improvement over Keynes’s

‘‘fundamental psychological law’’.’ Whereas Gilbert’s ‘improvement’ is not

as strong as a ‘revolution’, he does imply the existence of the change in

consumption theory after rational expectations. Consequently, we can see

that Friedman’s and Modigliani’s theories have been replaced by Hall’s;

Worrall’s structural realism seems to be an appropriate way to depict such a

shift.

But generally in the case of consumption theory, economists tend toclassify a theory by its structure. This view is best illustrated by the title of

Thomas Kuhn’s article ‘Theory-change as structure-change’ (Kuhn 1976).

In addition to Hall’s case, note that recently there has been a vast number of

studies on consumption that ‘go beyond’ the life cycle–permanent income

hypothesis. These studies cast doubt on the appropriateness of the

assumption of consumption smoothing, arguing that the assumption is

not fulfilled because the representative consumer may not want to smooth

out consumption owing to his/her precautionary motive of saving, orbecause s/he cannot do so due to liquidity constraints (see section 6).10

Consumption may be neither smoothing according to David Laibson’s

(1997, 1998) hyperbolic Euler equation that discount rates are declining over

time, nor constant in the standard exponential Euler equation. These

consumption theories are different from the life cycle–permanent income

hypothesis, either because consumers cannot freely borrow and lend to

smooth their consumptions, or because consumers’ behavior is based on

Laibson’s more radical adoption of time-inconsistent preferences. Thisindicates that economists usually assume complete consumption smoothing,

designated by the unrestricted Fisherian intertemporal choices framework,

and the life cycle–permanent income hypothesis as the same thing – even

though the final consumption functions, such as Friedman’s distributed lag

model and Hall’s random-walk model, are different. Therefore, theories are

distinguished by the structure they impose, so we see the discontinuity

among the Keynesian theory, Friedman–Modigliani life cycle–permanent

income hypothesis, and the new hypotheses of precautionary saving andliquidity constraints. Yet, when contemporary economists regard the Euler

A structure of the consumption function 235

Dow

nloa

ded

by [

Tem

ple

Uni

vers

ity L

ibra

ries

] at

13:

27 1

8 N

ovem

ber

2014

Page 12: A structure of the consumption function

equation as representing the true structure, they apply it to future theories.

So we see that the models based on these new consumption theories are

derived from Euler equations, but with auxiliary assumptions that aredissimilar to Hall’s.11 Worrall’s correspondence principle seems to fit this

description: the framework of intertemporal choices is used in developing

new models in which their Euler equations are considered as ‘limiting cases’

of the Euler equation in the random-walk model.

It seems the structural realist account only partially interprets the theory

of consumption function. However, regardless of the questions of the

existence of theory change and the possibility of a full application of

Worrall’s structural realism to the history of the consumption function,economists themselves have emphasized that the framework of intertem-

poral choices is maintained as the structure of the consumption function and

is employed for further research – exactly the ideas that Worrall stresses.

The economists’ realist attitude toward structure will be discussed in depth

in section 6. But first it needs to be pointed out that a difference between

Worrall’s structural realism and economists’ realism about structure lies in

the fact that in economics and econometrics the notion of invariance under

intervention is usually applied to identify the existence of structure, while inWorrall’s account there is no such independent criterion for identifying the

structure. This is the main criticism of Worrall’s account in the philosophy

of science literature.

5 STRUCTURAL REALISM AND REPRESENTATION

Worrall’s structural realist strategy involves an ‘optimistic induction’

concerning the discovery of mathematical structure in the history of science(Worrall 1994: 336). Stathis Psillos (2001: S23) criticizes the strategy as

being only a ‘modest epistemic thesis that emerges from looking into the

history of scientific growth’ that does not tell us what makes unchanged

mathematical equations a real structure. Elsewhere Psillos (1999: Ch. 7)

raises some unique challenges to Worrall’s account; two of them are related

to economic methodology and to the consumption function.

First, Psillos observes that apart from mathematical equations, theoretical

assumptions and principles are preserved during theory change, even inWorrall’s case of light. This point is similar to what Marcel Boumans (1999)

says about business-cycle model building. In economics, Boumans argues,

models are usually built by integrating many ingredients. He shows that

successful business cycle models integrate concepts and facts from economic

theories and practices in a mathematical mould. One of Boumans’s case

studies is Lucas’s business cycle model. He argues that Lucas’s model

inherits many ingredients from previous studies of business cycles theories.

When the model is regarded as a successful integration, its ingredients arepreserved in the later models (Boumans 1999: 89). Usually in economics,

236 Articles

Dow

nloa

ded

by [

Tem

ple

Uni

vers

ity L

ibra

ries

] at

13:

27 1

8 N

ovem

ber

2014

Page 13: A structure of the consumption function

also indicated in Boumans’s case studies, some, maybe not all, ingredients

are relational. Since the preserved ingredients are not all relational, it is not

possible for structural realists to identify the structure merely by employingWorrall’s inductive strategy. In addition, even though relational and non-

relational ingredients can be distinguished, not all relational equations are

regarded as structure. As the previous section stated, only the equation of

intertemporal choices is regarded as structure for the consumption function.

Psillo’s and Boumans’s studies show that in the natural and social sciences,

some ingredients preserved under theory change are non-structural. Hence

structural realism’s inductive strategy is not capable of distinguishing

structure from non-structure – this strategy seems to be ‘pessimistic’ ratherthan ‘optimistic’.

Secondly, Psillos tries to see off structural realism by arguing that Worrall

does not provide an independent argument to link mathematical equations

with the structure of the world; otherwise we cannot tell whether the

mathematical equations are preserved because they represent the true

structure or because, say, it is just convenient to apply them to build a

model. This argument seems true to our example of the Euler-equation

approach in which economists adopt the Euler-equation approach becauseit is operationally convenient, easy to measure, fitting rational expectations

hypothesis and representing the structure. It seems that we equally cannot

derive the structural realist conclusion from the employment of the Euler

equation. This again indicates that Worrall’s inductive strategy is by no

means optimistic because alone it cannot reach the existence of structure.

Both objections to Worrall’s strategy show that to claim structural

realism, some independent criteria are required to distinguish structure from

non-structure. For, as van Fraassen (2006: 293) puts it, ‘if there is no non-structure, there is no structure either’. Conventionally, like just as

economists use invariance under intervention to distinguish between

structure and non-structure, scientists apply the notions of invariance, for

example, symmetry, to serve as such criteria (van Fraassen 2006). However,

it seems that such criteria are implied in Worrall’s account if his structural

realism is understood in terms of representation.

Worrall’s structural realism involves a task of representation.

Intrinsically, Worrall is concerned with the relation between models andthe real world, that is, mathematical equations represent the true structure.

Worrall does not state that a set of equations is a structure; he says that the

mathematical equations of a theory express a structure of the phenomena

(e.g. Worrall 1989: 122). Michael Redhead, Worrall’s colleague at the

London School of Economics, points out a three-tier task of representation

in structural realism to clarify Worrall’s account (Redhead 2001). The first

task is to group the physical relations of interest as concrete structures. Then

the concrete structures are related to each other up to an isomorphism.Finally, the mathematical equations are constructed as the representation of

A structure of the consumption function 237

Dow

nloa

ded

by [

Tem

ple

Uni

vers

ity L

ibra

ries

] at

13:

27 1

8 N

ovem

ber

2014

Page 14: A structure of the consumption function

the abstract structure that maps the concrete structures up to the same

isomorphism (Redhead 2001: 345). Structural realism, as Redhead states, is

the claim that ‘this abstract structure associated with physical reality is whatscience aims, and to some extent succeeds, to uncover, rather than the true

physical relations of that reality’ (Redhead 2001: 345). The last step, the

mathematical representation of the abstract structure, is what Worrall is

aiming at. Redhead’s restatement also clearly presents the necessity of types

of isomorphism – a one–one mapping – between models. It can be stated that

there exists an isomorphic mapping between mathematical equations of the

old and new theories. When the isomorphism of equations is found, we can

say that such equations represent a real structure in Worrall’s sense.But if there is a representation, we must ask what makes some

mathematical equations a good representation of something – a structure

in our case. A formal account of representation can be found in Patrick

Suppes’ work on the semantic approach of theories (e.g. Suppes 2002), in

which isomorphism is a key concept. Suppes’ account of representation

suggests that representation is in mathematical form, is usually understood

in terms of models. A mathematical representation is empirically adequate

to an aspect of the world only if we prove a representation theorem formodels. The idea of representation theorems is as follows. If we want an

empirical model of the world (consisting of Redhead’s concrete structure) to

be represented by an abstract mathematical model (consisting of Redhead’s

abstract structure), we need the empirical model to satisfy a certain set of

axioms. Given that the axioms at the empirical end can in principle find

their counterparts at the mathematical end, we find the mathematical

model, and an isomorphism can be established between the empirical and

the mathematical models. In this sense a representation is made. To prove arepresentation theorem is to prove that such a type of isomorphism exists.

The role of representation theorems and isomorphic mappings is crucial

in two respects. On the one hand, isomorphism, as Suppes puts it, ‘makes

the intuitive idea of same structure precise’ (Suppes 1967: 59; emphasis in

original). This means that if a representation theorem is proven, the

mathematical model can be used to represent the empirical model with

respect to structure. In other words, the abstract structure and the concrete

structure are two of a kind. Seen in the way of representation, Worrall’sargument is compatible with Suppes’ account in the sense that a

mathematical equation of a theoretical model expresses the structure of

the world if there is an isomorphic mapping between the model and the

world. This also means, on the one hand, that if representation in Worrall’s

structural realism is regarded as or assumed to be satisfactory, a

representation theorem can always be proven to confirm the structural

representation. On the other hand, isomorphism indicates invariance under

transformation.12 When there is an isomorphic mapping between the sametypes of models, those models are said to be related by a permissible

238 Articles

Dow

nloa

ded

by [

Tem

ple

Uni

vers

ity L

ibra

ries

] at

13:

27 1

8 N

ovem

ber

2014

Page 15: A structure of the consumption function

transformation. Examples are the symmetry of geometric figures under

certain rotations, and the uniqueness of scales of measurement up to various

transformations (see Suppes 2002). The structural realists’ observation onthe preservation of mathematical equations between old and new theories

can be interpreted as the transformation between old and new theories.

Since the models of old and new theories are isomorphically related, these

theories contain the same structure. Hence structural realism can be

understood as a proposition derived from sequential models which are

invariant under a certain transformation.

Why does the task of representation need to be stressed for structural

realism? Because the concept of invariance that consists in representationtheorems can, at least indirectly, serve as a criterion for distinguishing

between structure and non-structure that structural realism is criticized for

failing to hold. Consider two theories, T1 and T2. Each is regarded as a

mature and successful theory because each contains a structure S1 and S2,

respectively, that represent the same empirical structure SE. Therefore the

mappings between S1 and SE (S1«SE), and between S2 and SE (S2«SE) are

the same. But since S1 and S2 are the same structure, the relationships of

S1«SE and S2«SE must be the same as well. In other words, when Worrallobserves that in the history of science the mathematical equations of the new

and old theories are invariant (isomorphism among theoretical models),

given that in the successful or mature scientific theories, the theoretical

model flourishingly represents a feature of the world (isomorphism between

theoretical and empirical models), we can conclude that the represented

empirical models are also invariant under the same type of transformation

as the theoretical one.

In this interpretation the represented empirical model is invariant andtherefore can be regarded as a structure. Moreover, when this inference

from the invariance of theoretical models to the invariance of empirical

models is established, Worrall’s inductive strategy for realism is better

supported as associated with the conventional philosophical view of ‘the

invariant as the real’ (Hooker 1991; Weinert 2004: 62).13

We interpret Worrall’s structural realism as securing invariance through a

satisfactory representation. In contrast, for economists, when structure is

explicitly and (almost) unanimously construed as invariance underintervention in macroeconometrics for those who see models as structural

(e.g. the Colwes Commission) or as non-structural (e.g. the VAR

approach),14 we can secure a good representation if the model is invariant.

Hoover (1994: 65) suggests seeing the Lucas critique as a criticism of the

traditional macroeconometric models for not being ‘accurate representa-

tions of the current structure of the economy’ because the typical

macroeconometric models are not invariant under policy change.

Therefore, the traditional ‘structural’ models do not refer to any structureas their proponents claim. When seen in this way, the Lucas critique not

A structure of the consumption function 239

Dow

nloa

ded

by [

Tem

ple

Uni

vers

ity L

ibra

ries

] at

13:

27 1

8 N

ovem

ber

2014

Page 16: A structure of the consumption function

only clearly asserts that the aim of econometrics is to use models to

represent the true structure of the economy, but also that econometrics

revives the notion of invariance as a criterion for an adequate representationof the structure of the world by models. When specifying models, invariance

is thought to be a necessary and sufficient condition for a model

representing the structure.15

6 REALISM ABOUT STRUCTURE

The realist attitude toward structure that economists hold can be argued as

the following. Given that the existence of structure of the consumptionfunction, namely the Euler equation, is perceived, macroeconomists accept

or reject the theory of consumption accordingly. They accept the life cycle–

permanent income hypothesis and the Euler-equation approach precisely

because they capture the right structure. To say that Keynes’s and

Duesenberry’s theories were abandoned for lack of theoretical foundation

is to say that they do not consist of the right structure. This implies a

‘realism about structure’ among economists. Realism about structure, on

the one hand, is consistent with the invariance view of reality: invariantrelations are structural and real. On the other hand, it does not subscribe to

the falsificationist account of theory testing but asserts that we accept a

theory if it yields a model containing the right structure. If a model is not

supported by empirical data when the model is considered as containing the

true structure, we do not reject the model but instead construct a new model

with the same true structure.

Methodologically, the realism about structure is compatible with two

characterizations of the new classical macroeconomics. One is argued byHoover (1994), who asserts that the new classical macroeconomists have a

strong prior belief in economic theories that are based on well-specified

optimization behavioral assumptions. Hoover labels this a ‘strong aprior-

ism’ that is similar to the position represented in Koopmans’ (1947) article

‘Measurement without theory’. The new classical economists, as also

exemplified in Hall’s model, adopt a strategy of developing an optimizing

representative-agent model taking ‘deep parameters’ as given (Hansen and

Sargent 1980; Lucas and Sargent 1979; and see Hoover 1988, 1994). Suchmodels are secured for invariance problems and can then be regarded as

accurate representations of the structure. The other is stated by John Sutton

(2000), who claims that economists usually hold ‘strong priors’ in favor of

some basic ideas. One of the examples that Sutton uses is our history of

the consumption function. Sutton, borrowing extensively from Gilbert

(1991), observes that Duesenberry’s relative income hypothesis fits

empirical data better than Friedman’s permanent income hypothesis, but

the relative income hypothesis ‘suffered primarily from its lack of atheoretical underpinning in individual maximizing behavior, rather than any

240 Articles

Dow

nloa

ded

by [

Tem

ple

Uni

vers

ity L

ibra

ries

] at

13:

27 1

8 N

ovem

ber

2014

Page 17: A structure of the consumption function

shortcomings at the empirical level’ (Sutton 2000: 95). Similarly, he points

out that Hall’s random-walk model is not usually falsified by empirical data.

This means that when the theory is inconsistent with data, economists whohold a strong prior belief in the framework of Fisherian intertemporal

choices accept the theories which contain the structure that is thought to be

true, and reject those which do not.

Sutton’s case of holding strong prior belief in the Euler-equation

approach can be explored in more detail. Testing the new classical

consumption theory, as addressed in literature, resolves the correspondence

problem between theory and data. Theoretically, the life cycle–permanent

income hypothesis involves intertemporal choices as to the structure of theconsumption function, where the assumption of intertemporal choices

implies that the representative consumer optimally smoothes the consump-

tion over time. This motive has been pointed out by Friedman (1957: 7) that

a consumer tends to ‘straighten out’ the stream of consumption

expenditures. This assumption is known as consumption smoothing. The

life cycle–permanent income hypothesis predicts that consumption exhibits

a smooth pattern, because permanent income does not fluctuate in response

to the short-term income fluctuation; permanent income is smoother thanmeasured income. Consumption is smoother than measured income because

consumption presumably depends upon permanent income. In some cases,

the permanent income hypothesis could lead to the conclusion that the

consumption in each period equally divides the permanent income. This

outcome of consumption function can be called consumption smoothness,

which is also a stylized fact about consumption. The problem of

correspondence lies between consumption smoothing and consumption

smoothness. The existence of such a correspondence is a strong support forHall’s random-walk model. Since the random-walk model exhibits the

relationship between current and previous consumption, smoothness

can be defined as ct+1 > ct and thus is measured by the parameter on ct.

Hall’s own estimation seems to support this interpretation: the

estimate of the coefficient for ct of the random-walk model is 0.983.

Consequently, random walk consumption is an empirical justification for

the consumption smoothing assumption. This implies the following

argument: if consumption is not a random walk, i.e. consumption is notsmooth, then the correspondence between consumption smoothing and

consumption smoothness does not exist. The failure of such correspondence

is probably due to the inadequacy of the assumption of consumption

smoothing.

Macroeconomists have long been considering the empirical issues of the

Euler-equation approach. Hall (1978) used 1948(1)–1977(1) seasonally

adjusted quarterly data to test his random-walk model against four other

consumption functions, in order to see whether variables other thanprevious consumption are significant for current consumption. The results

A structure of the consumption function 241

Dow

nloa

ded

by [

Tem

ple

Uni

vers

ity L

ibra

ries

] at

13:

27 1

8 N

ovem

ber

2014

Page 18: A structure of the consumption function

show that higher lag variables have no significant predictive power for

current consumption. Moreover, the F-statistics for jointly predictive

powers for variables other than ct-1 are all below critical values of the F-

distribution at 0.05 level. Thus the random walk consumption is not

rejected. But Hall also found that when stock prices are included as proxies

of wealth, they are statistically significant, hence the random walk

consumption function should be rejected. Hall’s findings inspire further

empirical studies based on the Euler-equation approach to reconcile theory

with data. In John Campbell and Gregory Mankiw’s (1989) model,

consumers are divided into two groups: one follows the assumption of

consumption smoothing, and the other follows the rule of thumb that

consumers’ decisions are based on current income. As a result, consumption

does not follow a random walk because current income is significant to the

next period’s consumption. Other studies show that the life cycle–permanent

income hypothesis fails to explain the facts of ‘excess sensitivity’ that

consumption is too sensitive in response to expected changes in income

(Flavin 1981), or of ‘excess smoothness’, meaning that consumption is too

smooth in response to unexpected changes (Campbell and Deaton 1989) of

consumption. Consequently many have attempted to explain the anomalies

(excess sensitivity and excess smoothness) by providing models with

different modifications.

However, Hall insists that since most of the predictive value of the stock

prices in his random-walk model comes from the change of the immediately

preceding quarter, it is compatible with the implication of the life cycle–

permanent income hypothesis that the change in consumption is related to

the change in permanent income (Hall 1978: 985–6). Hall also contends that

the improving predictive power by including stock prices ‘while statistically

significant, is not numerically large’ (Hall 1978: 985).

It can be argued that even though many other economists take

this empirical finding as evidence against the fundamental assumption of

complete consumption smoothing in the random-walk model, Hall

insists that the random-walk model is a reasonable approximation of

consumption because of his strong belief, as Sutton indicates, in the

fundamental assumption of consumption smoothing. Hoover’s view

provides reason for such a belief. The belief in the new classical model is

well grounded in its satisfaction of the Lucas critique on invariance. In this

vein it can be interpreted that although the correspondence between

consumption smoothing and consumption smoothness does not exist,

Hall still has reason to believe both in the random-walk model and in the

Euler-equation approach. The very reason, as the quotation above

shows, lies in the model’s inclusion of an Euler equation, a true structure

of the consumption function that characterizes an invariant relation of

consumption.

242 Articles

Dow

nloa

ded

by [

Tem

ple

Uni

vers

ity L

ibra

ries

] at

13:

27 1

8 N

ovem

ber

2014

Page 19: A structure of the consumption function

7 CONCLUSION

For Worrall, passively retained mathematical equations are taken as evidence of

the existence of the true structure and they can play a heuristic role in developing

new theories. In a similar realism about structure, observed in the history of the

consumption function, both representation and invariance are central to

structuring the consumption function. For economists, the structure of the

consumption function is the Fisherian intertemporal choices, or the Eulerequation. Euler equations are passively observed not only as being preserved

across changes in consumption theories, but also because they satisfy the

definition of structure in macroeconomics, invariance under intervention. The

Fisherian framework of intertemporal choices is therefore conceived by

macroeconomists as the real structure and actively involved in theoretical and

empirical research. Moreover, this structure, that is thought to be real, provides

macroeconomists with grounds for accepting and rejecting a consumption

theory. Consequently, the poor correspondence between the theoretical assump-tion of consumption smoothing and the empirical fact of consumption smooth-

ness does not imperil the Euler-equation approach. Anomalies may reject the

permanent income hypothesis or the life-cycle hypothesis, but they only

motivate modelers to modify the Euler equations instead of abandoning them.

Yet Worrall’s structural realism is criticized for not involving criteria of

distinguishing between structure and non-structure. But we can understand

the structural realism as involving a task of representation in the following

way. The mathematical equations in a theory represent invariant empiricalrelations. When these equations are retained during theory change, our

belief in their truth to the world is based on their invariance across

transforming theories. Hence the notion of invariance under transformation

is embedded in the representation and can be applied to judge the existence

of structure.

Hsiang-Ke Chao

National Tsing Hua University, Taiwan

[email protected]

ACKNOWLEDGEMENTS

An early version of this paper was titled ‘A structural realist interpretation

of the euler-equation approach in macroeconomics’, published as a research

memorandum of the Amsterdam Research Group in the History and

Methodology of Economics, Faculty of Economics and Econometrics,

University of Amsterdam (No. 01-9). I thank Marcel Boumans and Mary

Morgan for their advice and encouragement. I am also most grateful to

Kevin Hoover and an anonymous referee for their constructive commentsand suggestions.

A structure of the consumption function 243

Dow

nloa

ded

by [

Tem

ple

Uni

vers

ity L

ibra

ries

] at

13:

27 1

8 N

ovem

ber

2014

Page 20: A structure of the consumption function

NOTES

1 Modigliani (1947, 1949) has shown a Keynesian consumption function, whichconsists of the ‘highest previous income peak’, that is similar to Duesenberry’srelative income hypothesis.

2 See, inter alia, Psillos (1999, esp. Ch. 8), Votsis (2004), and papers that appearin the special issue of Synthese in 1999 on structural realist interpretations ofquantum field theory.

3 Note that some philosophers of science, for example, French and Saatsi (2004),use the terms ‘structural realism’ and ‘realism about structure’ interchangeably.

4 See Grower (2000) and Votsis (2004) for the history of structural realism.5 The contrast between Keynes’s consumption function and Fisher’s account can

be seen as early as Holden (1938). Note that in the General Theory, before heproposes the formal definition of the fundamental psychological law, Keynesdiscusses the ‘objective factors’ that he thinks would affect the marginalpropensity to consume. These objective factors include ‘changes in the rate oftime-discounting’ and ‘changes in expectations of the relation between thepresent and the future level of income’ (Keynes 1936: 93–6). But he goes on tostate that ‘ [f]or whilst the other factors are capable of varying (and this mustnot be forgotten), the aggregate income measured in terms of the wage-unit is,as a rule, the principal variable upon which the consumption-constituent of theaggregate demand function will depend.’ (Keynes 1936: 96). I thank a refereefor pointing out this to me. Also see Keynes’ (1938) reply to Holden (1938).

6 See Deaton (2005).7 See Sheffrin (1983: 105–6).8 The term ‘solved-out approach’ appears in Muellbauer (1994). In this way,

the consumption function is solved, for example, by substituting the Eulerequation (equation (1)) into the intertemporal budget constraint

c1E1c2

1zr~A0 1zrð Þzw1zE1w2

1zr , where At is the asset in period t, ct is the

consumption in period t, Et is the mathematical expectation conditional on allinformation available in period t; r is the constant real interest rate, and wt is

income in period t. c1~1

1z 1=1zrð Þ A0 1zrð Þzw1zE1w2

1zr

� �.

9 Hall (1988) measures the elasticity of intertemporal substitution.10 Discussions on the hypotheses of precautionary saving and liquidity

constraints can be seen in Carroll (2001) and the references therein.11 Laibson (1998) further shows that the standard exponential Euler equation is a

special case of his hyperbolic Euler equation.12 One major difference between invariance under intervention and invariance

under transformation is that the former account explicitly addresses theexistence of an outside force affecting the structure and hence associates itselfto the structural account of causality firmly. In modern terminology, a causalstructure consists of superexogenous relations that are invariant to interven-tions. In this way, the requirement for invariance and the Lucas critique can beunderstood as a type of statement on using models to represent a causalstructure satisfactorily. The structural account of causality in macroeconomicshas been extensively explored in Hoover (2001). Also note that causal relationsand invariant relations cannot be treated as equal. See Hoover (2001) andChen (2002).

13 Also see van Fraassen (1989, 2006) and Nozick (2001) for discussion of theontological issue of invariance in science.

14 The introduction to Hendry and Morgan (1995) provides an excellentreview on the development of the notion of structure in econometrics

244 Articles

Dow

nloa

ded

by [

Tem

ple

Uni

vers

ity L

ibra

ries

] at

13:

27 1

8 N

ovem

ber

2014

Page 21: A structure of the consumption function

before the 1950s. Sims’s acceptance of the Cowles Commission’s definitionof structure, especially in Hurwicz (1962), can be seen in Sims (1980,1982).

15 Hendry (1995: 34) argues that invariance is only a necessary – but notsufficient – condition for structure because he requires the structuralparameters to be non-reducible: structural parameters are those which areinvariant and not derived from more fundamental parameters.

REFERENCES

Ando, A. and Modigliani, F. (1963) ‘The ‘‘life-cycle’’ hypothesis of saving: aggregateimplications and tests’, American Economic Review 53: 159–78.

Attanasio, O. P. (1998) ‘Consumption demand’, NBER Working Paper No. 6466.Attanasio, O. P. and Low, H. (2004) ‘Estimating euler equations’, Review of

Economic Dynamics 7: 405–35.Boumans, M. J. (1999) ‘Built-in justification’, in M. S. Morgan and M. Morrison

(eds) Models as Mediators: Perspectives on Natural and Social Science,Cambridge: Cambridge University Press, pp. 66–96.

Boyd, R. (1984) ‘The current status of scientific realism’, in J. Leplin (ed.) ScientificRealism, Berkeley: University of California Press, pp. 41–82.

Campbell, J. Y. and Deaton, A. S. (1989) ‘Why is consumption so smooth?’, Reviewof Economic Studies 56: 357–74.

Campbell, J. Y. and Mankiw, N. G. (1989) ‘Consumption, income and interest rate:interpreting the time series evidence’, NBER Macroeconomic Annual, 185–216.

Carroll, C. D. (2001) ‘A theory of the consumption, with and without liquidityconstraints’, Journal of Economic Perspectives 15: 23–45.

Chen, S.-T. (2002) ‘Economic theorizing: a causal structuralist account withexamples from international trade theory’, Ph.D. thesis, Department ofPhilosophy, Logic and Scientific Method, London School of Economics.

Deaton, A. S. (2005) ‘Franco Modigliani and the life cycle theory of consumption’,Working Paper, Princeton University.

Duesenberry, J. S. (1949) Income, Saving and the Theory of Consumer Behavior,Cambridge, MA: Harvard University Press.

Fine, A. (1998) ‘Scientific realism and antirealism’, in E. Craig (ed.) RoutledgeEncyclopedia of Philosophy, Vol. 8, London: Routledge, 581–4.

Fisher, I. (1930) The Theory of Interest, New York: Macmillan.Flavin, M. A. (1981) ‘The adjustment of consumption to changing expectations

about future income’, Journal of Political Economy 89: 974–1009.French, S. and Saatsi, J. (2004) ‘Realism about structure: the semantic view and non-

linguistic representations’, Paper presented at Philosophy of Science AssociationMeeting 2004, Austin, TX, 18–20 November.

Friedman, M. (1957) A Theory of the Consumption Function, Princeton, NJ:Princeton University Press.

Gilbert, C. L. (1991) ‘Do economists test theories? – Demand analysis andconsumption analysis as tests of theories of economic methodology’, in N. deMarchi and M. Blaug (eds) Appraising Economic Theories: Studies in theMethodology of Research Programs, Aldershot: Edward Elgar, pp. 137–68.

Granger, C. W. J. (1999) Empirical Modeling in Economics: Specification andEvaluation, Cambridge: Cambridge University Press.

Grower, B. S. (2000) ‘Cassirer, Schlick, and ‘‘structural’’ realism: the philosophy ofthe exact sciences in the background to early logical empiricism’, British Journalfor the History of Philosophy 8: 71–106.

A structure of the consumption function 245

Dow

nloa

ded

by [

Tem

ple

Uni

vers

ity L

ibra

ries

] at

13:

27 1

8 N

ovem

ber

2014

Page 22: A structure of the consumption function

Haavelmo, T. M. (1944) ‘The probability approach in econometrics’, Econometrica12(Supplement): 1–118.

Hall, R. E. (1978) ‘Stochastic implications of the life cycle–permanent incomehypothesis: theory and evidence’, Journal of Political Economy 86: 971–87.

Hall, R. E. (1988) ‘Intertemporal substitution in consumption’, Journal of PoliticalEconomy 96: 339–57.

Hall, R. E. (1990) ‘Survey of research on the random walk of consumption’, in TheRational Consumer, Cambridge, MA: The MIT Press, pp. 131–57.

Hansen, L. P. and Sargent, T. J. (1980) ‘Formulating and estimating dynamiclinear rational expectations models’, Journal of Economic Dynamics and Control 2:7–46.

Hansen, L. P. and Singleton, K. J. (1983) ‘Stochastic consumption, risk aversion andthe temporal behavior of asset returns’, Journal of Political Economy 91: 249–65.

Hendry, D. F. (1995) Dynamic Econometrics, Oxford: Oxford University Press.Hendry, D. F. and Morgan, M. S. (eds) (1995) The Foundations of Econometric

Analysis, Cambridge: Cambridge University Press.Holden, G. R. (1938) ‘Mr. Keynes’ consumption function and the time-preference

postulate’, Quarterly Journal of Economics 52: 281–96.Hooker, C. A. (1991) ‘Projection, physical intelligibility, objectivity and complete-

ness: the divergent ideals of Bohr and Einstein’, British Journal of the Philosophyof Science 42: 491–511.

Hoover, K. D. (1988) The New Classical Macroeconomics: A Skeptical Inquiry,Oxford: Basil Blackwell.

Hoover, K. D. (1994) ‘Econometrics as observation: the Lucas critique and thenature of econometric inference’, Journal of Economic Methodology 1: 65–80.

Hoover, K. D. (2001) Causality in Macroeconomics, Cambridge: CambridgeUniversity Press.

Hurwicz, L. (1962) ‘On the structural form of interdependent systems’, in E. Nagel,P. Suppes and A. Tarski (eds) Logic, Methodology and Philosophy of Science:Proceedings of the 1960 International Congress, Stanford: Stanford UniversityPress, pp. 232–9.

Hynes, J. A. (1998) ‘The emergence of the neoclassical consumption function: theformative years, 1940–1952’, Journal of History of Economic Thought 20: 25–50.

Keynes, J. M. (1936) The General Theory of Employment, Interest and Money,London: Macmillan.

Keynes, J. M. (1938) ‘Mr. Keynes’s consumption function: reply’, Quarterly Journalof Economics 52: 708–9.

Koopmans, T. C. (1947) ‘Measurement without theory’, Review of Economics andStatistics 29: 161–72.

Kuhn, T. S. (1976) ‘Theory-change as structure-change: comments on the Sneedformalism’, Erkenntnis 10: 179–99.

Kuznets, S. (1946) National Income: A Summary of Findings, New York: NationalBureau of Economic Research.

Laibson, D. (1997) ‘Golden eggs and hyperbolic discounting’, Quarterly Journal ofEconomics 62: 443–79.

Laibson, D. (1998) ‘Life-cycle consumption and hyperbolic discounting functions’,European Economic Review 42: 861–71.

Laudan, L. (1981) ‘A confutation of convergent realism’, Philosophy of Science 48:19–49.

Lucas, R. E. (1976) ‘Econometric policy evaluation: a critique’, in K. Brunner and A.H. Meltzer (eds) The Phillips Curve and Labor Markets. Carnegie-Rochester

246 Articles

Dow

nloa

ded

by [

Tem

ple

Uni

vers

ity L

ibra

ries

] at

13:

27 1

8 N

ovem

ber

2014

Page 23: A structure of the consumption function

Conference Series on Public Policy, Vol.1, Amsterdam: North Holland, pp.19–46.

Lucas, R. E. and Sargent, T. J. (1979) ‘After Keynesian macroeconomics’,Federal Reserve Bank of Minneapolis Quarterly Review 3: 1–16.

Marschak, J. (1953) ‘Economic measurement for policy and prediction’, in W. C.Hood and T. C. Koopmans (eds) Studies in Econometric Method, CowlesCommission Monograph 14, New York: Wiley, 1–26.

Mayer, T. (1972) Permanent Income, Wealth, and Consumption: A Critique of thePermanent Income Theory, the Life-Cycle Hypothesis and Related Theories,Berkeley: University of California Press.

Modigliani, F. (1947) ‘Fluctuations in the saving–income ratio: a problem ineconomic forecasting’, Social Research 14: 413–20.

Modigliani, F. (1949) ‘Fluctuations in the saving–income ratio: a problem ineconomic forecasting’, in Studies in Income and Wealth, Vol. 11, New York:National Bureau of Economic Research, pp. 371–441.

Modigliani, F. (1975) ‘The life cycle hypothesis of saving twenty years later’, in M.Parkin (ed.) Contemporary Issues in Economics, Manchester: ManchesterUniversity Press, pp. 2–35.

Modigliani, F. and Brumberg, R. (1954) ‘Utility analysis and the consumptionfunction: an interpretation of cross-section data’, in K. K. Kurihara (ed.)Post Keynesian Economics, New Brunswick: Rutgers University Press, pp.383–436.

Muellbauer, J. (1994) ‘The assessment: consumer expenditure’, Oxford Review ofEconomic Policy 10: 1–41.

Muth, J. F. (1960) ‘Optimal properties of exponentially weighted forecasts’, Journalof the American Statistical Association 55: 299–306.

Nozick, R. (2001) Invariances: The Structure of the Objective World, Cambridge,MA: Belknap Press.

Papineau, D. (1996) ‘Introduction’, in D. Papineau (ed.) The Philosophy of Science,Oxford: Oxford University Press, pp. 1–20.

Psillos, S. (1999) Scientific Realism: How Science Tracks Truth, London: Routledge.Psillos, S. (2001) ‘Is structural realism possible?’, Philosophy of Science

52(Supplement): S13–24.Putman, H. (1975) Mathematics, Matter and Method: Philosophical Papers, Vol. 1,

Cambridge: Cambridge University Press.Redhead, M. (2001) ‘‘‘Quests of a realist’’, review of Psillos (1999)’, Metascience 10:

341–7.Sheffrin, S. M. (1983) Rational Expectations, Cambridge: Cambridge University Press.Sims, C. A. (1980) ‘Macroeconomics and reality’, Econometrica 48: 1–48.Sims, C. A. (1982) ‘Policy analysis with econometric models’, Brooking Papers on

Economic Activity 1982(1):107–152.Suppes, P. (1967) ‘What is a scientific theory?’, in S. Morgenbesser (ed.) Philosophy

of Science Today, New York: Basic Books, pp. 55–67.Suppes, P. (2002) Representation and Invariance of Scientific Structures, Stanford:

CSLI Publications.Sutton, J. (2000) Marshall’s Tendency: What Can Economists Know?, Cambridge,

MA: The MIT Press.Thomas, J. J. (1989) ‘The early econometric history of the consumption function’,

Oxford Economic Papers 41: 131–49.Van Fraassen, B. (1989) Law and Symmetry, Oxford: Oxford University Press.Van Fraassen, B. (2006) ‘Structure: its shadow and substance’, British Journal for the

Philosophy of Science 57: 275–307.

A structure of the consumption function 247

Dow

nloa

ded

by [

Tem

ple

Uni

vers

ity L

ibra

ries

] at

13:

27 1

8 N

ovem

ber

2014

Page 24: A structure of the consumption function

Votsis, I. (2004) ‘The epistemological status of scientific theory: an investigation ofthe structural realist account’, Ph.D. thesis, Department of Philosophy, Logic andScientific Method, London School of Economics.

Weinert, F. (2004) The Scientist as Philosopher: Philosophical Consequences of GreatScientific Discoveries, Berlin: Springer.

Worrall, J. (1985) ‘Scientific discovery and theory-confirmation’, in J. C. Pitt (ed.)Change and Progress in Modern Science, Dordrecht: Riedel, pp. 301–31.

Worrall, J. (1989) ‘Structural realism: the best of both worlds?’, Dialectica 43:99–124.

Worrall, J. (1994) ‘How to remain (reasonably) optimistic: scientific realism and the‘‘luminiferous ether’’’, in D. Hull, M. Forbes and R. M. Burian (eds) PSA 1994.Vol. 1, East Lansing: Philosophy of Science Association, pp. 334–42.

248 Articles

Dow

nloa

ded

by [

Tem

ple

Uni

vers

ity L

ibra

ries

] at

13:

27 1

8 N

ovem

ber

2014