<<

. 23
( 25 .)



>>

which it is assumed. Since all assumptions, all observations, are in some something in the real world. If our understanding is to be an improvement
way dependent on the acceptance of certain universals, the acceptance of over past understanding the new understanding must contradict some of the
assumptions, theories and observations is in this sense ad hoc. Ad hocery in old understanding. Any ad hoc modification which avoids such
this fundamental sense can neither be criticized nor recommended (because contradictions can only be a loss, a backward step.
the criticism or recommendation would also be ad hoc in the same sense). In summary, ad hoc specifications that limit further the conceivable
The ad hocery that might be criticized is that which arises when counter- states of the real world (which possibly can be compatible with the model
examples are arbitrarily ruled out when the theorist narrows the or theory) are ˜good™ since they increase testability. Ad hoc specifications
˜applicability™ of his or her theory “ for example, by assuming that our which increase the content by increasing the number of exogenous
theory applies only to ˜normal goods™. Such ad hocery might be criticized variables that might affect the determination of the endogenous variables
because it avoids criticism or it handicaps the theorist™s understanding of can also increase the testability since more possible counter-examples can
the objects of his or her study. In general, we can say that any ad hocery be deduced from the model and thus be used as indirect tests of the model.
which reduces the testability of a theory is considered ˜bad™ by most With regard to the ad hoc models of consumer theory being considered
theorists today. Conversely, any ad hocery which increases the testability is in this and the previous chapter, we can say the following: Hicks™
considered ˜good™. assumption that extremely inferior goods are less likely than slightly
The question arises as to how one increases the testability of a theory. I inferior goods is probably false. But that does not jeopardize the original
have previously dealt with the subject of how model-building assumptions consumer theory if it is still possible to exclude Giffen goods by specifying
can affect testability [Boland 1989, Chapters 2 and 3] where I have set out directly the nature of preferences. However, all specifications of
an analysis of the ingredients of a model (viz. the number of parameters, preferences need not be improvements. Some of the specifications may
standard-form coefficients, exogenous variables, endogenous variables, increase the ˜likelihood™ of Giffen goods, but those specifications which do
etc.) and demonstrated a measure of a model™s testability such that it is increase the ˜likelihood™ may themselves be ˜unlikely™, since they may be
very special (ad hoc) cases.11
possible to say when a model is ˜more testable™. The basic idea is that the
more information needed to test a newly modified model than was needed
without the modification, the less testable the model becomes. Such a
GIFFEN GOODS AND THE TESTABILITY OF DEMAND THEORY
modification would constitute ˜bad™ ad hocery. Testability, however, need
not be viewed as an ad hoc test of ad hocery. Testability is closely linked A couple decades ago the issue of the testability of demand theory itself
with the explanatory power of any theory, or with its empirical was actually publicly debated. The debaters were Cliff Lloyd [1965, 1969]
˜meaningfulness™ as followers of Paul Samuelson™s methodology [1947/65] and Gordon Welty [1969]. The importance of Giffen goods for the
like to say. An ad hoc specification of a theory which would make it testability of demand theory was only implicitly raised in their debate.
possible to test the theory with less information would be considered an However, Giffen goods were the explicit topic of Welty™s [1971] critique
improvement “ that is, it would be ˜good™ ad hocery. Testability, however, of Louis De Alessi™s [1968] views on the Giffen paradox. I will comment
can only be viewed as a means to an end, never as an end in itself. Even here on the Welty“Lloyd debate and Welty™s critique of De Alessi™s views
when the good ad hoc modification produces a model which turns out to be in hopes of furthering the understanding of the significance of Giffen goods
false (when tested), we still do not know whether it is the modification or it or upward sloping demand curves.
is something in the original model which is yielding the contradictions Lloyd [1965] discusses the general issue of the falsifiability of demand
between the modified model and the test evidence. 10 theory. Lloyd seems to think that ˜traditional demand theory™ can be tested.
Now ad hoc modifications such as limiting the applicability of a model For him a prerequisite of testability would be falsifiability. He outlines
© LAWRENCE A. BOLAND
208 Principles of economics Giffen goods vs market-determined prices 209
hatch to avoid almost any conceivable refutation. 13
what he considers to be testable ˜implications™ of demand theory. Basically,
if one can determine whether a good is not an inferior good, we can test the Welty™s arguments concerning ceteris paribus clauses are based on a
Slutsky equation (which presumes maximization of utility). An upward simple matter of logic. Adding extra clauses to any theory can insulate that
sloping demand curve for a non-inferior good is clearly contrary to theory from refutation. By the well known property of logic called modus
traditional demand theory. Whether one can actually test demand theory in tollens, we know that a false conclusion derived from a valid logical
this case would depend on the acceptance of the conventions used to argument implies the existence of at least one false statement contained in
establish the non-inferiority of the good in question and to measure the that argument. Unfortunately, modus tollens cannot usually indicate which
slope of the demand curve. The test will only be as good as the testing statement (of the valid argument) is false. If the argument consists of the
conventions used. But, as a matter of logic, Lloyd argues that demand original theory plus some additional clauses, then a false conclusion (or
theory is falsifiable, hence not untestable for reasons of internal logic of the prediction) does not tell us whether it is the original theory or the added
clause which is at fault.14 However, if the added clause can be
individual consumer.
Many economists may think that limiting any testing of demand theory independently tested, then this matter of logic “ the ambiguity of modus
tollens “ need not concern us.15
to non-inferior goods renders the theory irrefutable. As De Alessi put it in
1968, Implicit in this debate and criticism is the view that the existence of the
possibility of deducing upward sloping demand curves from a given theory
The theoretical admission that the income effect may dominate the
of the consumer is evidence of the failure of demand theory. Not everyone
substitution effect in the case of inferior goods implies that the
would accept this view. Many seem to think that neoclassical
demand curve of an individual, derived holding money income
(microeconomic) consumer theory can be refuted without negating either
constant, may be either positively or negatively sloped; it follows that
the Law of Demand or neoclassical price theory. For example, observance
the sign of the slope of the corresponding aggregate demand curve is
of conceivable counter-evidence would lead Lloyd to reject Ordinal
also indeterminate, and thus cannot be refuted by experience. [p. 287]
Demand Theory, yet, as he said, there are an infinity of possible theories of
If Lloyd™s proposed test is only a test of an individual™s behaviour, De the consumer. What one replaces it with need not be anything like the
Alessi claims, original consumer theory. However, the given refuting evidence would now
have to be explained by the replacement. Lloyd™s notion of convincing
Under no circumstances a single observation pertaining to a single
refuting evidence is the observation of upward sloping demand curves for
individual would provide a test of any economic hypothesis. [p. 290]
non-inferior goods (as well as Giffen goods). But, if my arguments in this
And further, chapter are correct, his counter-evidence would overturn neoclassical price
theory as well. Of course, neoclassical price theory can be false and the
in the final analysis, ... economists accept negatively sloped demand
traditional demand theory true without any need for ad hoc modifications.
curves ... because empirical evidence suggests that negatively sloped
In this case the indeterminacy that De Alessi and others point out would
demand curves work. [p. 291]
not matter. It would not matter because if price theory were false and
It seems that De Alessi sides with George Stigler [1950] in accepting Giffen goods were considered possible, demand theory would no longer be
negatively sloped demand curves as a fact until hard evidence to the interesting as it would not have any intellectual purpose. However, in this
contrary is provided. And until this occurs, the job of any demand theorist latter case, if price theory is false, there are no market-determined prices in
is to explain the implicit regularity “ the non-existence of Giffen goods. De the neoclassical sense.
Alessi suggests a possible modification of traditional demand theory. 12 In the absence of a successful test of demand theory as suggested by
Welty argues that Lloyd and De Alessi are both wrong as the former™s Lloyd, what are we to conclude? Should we accept ad hoc modifications in
testing conventions and the latter™s modification would each make the order to explain the presumed regularity inferred from the absence of
traditional theory unfalsifiable. The basis of Welty™s critique of Lloyd is conclusive evidence of the Giffen paradox? De Alessi seems to think we
the role of ceteris paribus clauses and to what extent such clauses refer to should. Others such as Stigler can argue that there is no independent
unspecified variables. If one were to say the Law of Demand is true ceteris evidence of such a regularity either and thus we can drop the necessity of
paribus then one could always use the ceteris paribus clause as an escape being able to deduce only negatively sloped demand curves. Welty seems
© LAWRENCE A. BOLAND
210 Principles of economics Giffen goods vs market-determined prices 211
to think that such a weak approach would make demand theory untestable NOTES
but his conclusion is based on what may be a mistake, the alleged 1 It should again be noted that Marshall™s concern for Giffen goods was due to
indeterminacy of the slope of the demand curve. Lloyd and others have doubts not about his theory of demand but instead about the ability to calculate
shown, however, that the slope may always be determinate. It is only that consumers™ surplus since such a calculation would require a downward sloping
demand curve [see further Dooley 1983].
the slope is indeterminate with the a priori conditions placed on utility
2 For example, we could have publicly or privately administered prices. And with
functions or indifference maps.
(1) excess demand does not necessarily lead to a rising price without someone
having the notion that by raising the price the situation will somehow be
improved.
CONCLUDING REMARKS
3 The existence of a counter-example (a case where the world is as described
here, but there still is no movement toward equilibrium) will be sufficient
This brings us full circle. I have argued that the Giffen paradox is contrary
evidence for the insufficiency of the combination of (1) and (2). Their necessity
to our market equilibrium theory of prices. Apart from our neoclassical
has never been asserted except by those who might wish to claim that is the
price theory, the existence of the Giffen paradox would not be a refutation way the world should be.
of consumer theory. Lloyd™s positively sloped individual demand curve for 4 To successfully criticize the necessity we would have to produce a successful
a non-inferior good would be a refutation of both traditional consumer and theory that did not explicitly or implicitly use both of these assumptions (1) and
(2).
traditional price theories, but that is still not a case of a Giffen good in the
5 I will ignore the cases that cannot be represented as ˜well defined functions™
Hicksian sense. Giffen goods themselves are still consistent with Ordinal
(viz. vertical and horizontal lines) and those cases of parallel demand and
Demand Theory. The problem is that Ordinal Demand Theory which supply curves which imply a covariance that would contradict independent
allows Giffen goods may not be consistent with our individualist theory of decision-making.
market prices. 6 For example, if both curves are positively sloped (e.g. a case involving a Giffen
good), Walrasian stability would not be assured if the market is characterized as
If the existence of Giffen goods has never been empirically established
case (f). Thus we must be able to explain why the supply curve will be steeper
then a realistic theory of demand should at least explain the fact of their
than the demand curve as in case (c).
non-existence. Any demand theory which does not explain that ˜fact™ (if it 7 In other words, if the price is above marginal cost, the firm will increase the
is agreed that it is a fact) has not done its empirical job, let alone whether quantity produced.
or not it has done its intellectual job with regard to explaining the demand 8 It is interesting to note that one can argue that both Marshall and Walras used
both stability concepts. So-called Walrasian stability must hold in the short run
side of price theory consistent with laissez-faire individualism. More
and Marshallian stability in the long run [see Davies 1963]. In this light, note
subtly, in any given demand theory, if Giffen goods are allowed as a
also that most neoclassical arguments involving prices in applied economics
possibility for the individual but not for the aggregate demand curve, then presume the existence of a long-run equilibrium. And since the long run is but a
such a theory puts the desired independence of decision-makers into special short-run equilibrium, both stability conditions must hold in applied
jeopardy whenever market-determined prices are to be the ˜given prices™ neoclassical economics based on market-determined prices. Some Post-
Keynesian economists may wish to dismiss the long-run aspect but the
upon which the individual consumers base their demand decisions. If
fulfillment of Marshallian stability is already built into the neoclassical short-
Giffen goods are allowed in consumer theory but not in price theory, then
run theory of supply. Other more mathematically minded economists may argue
some explanation must be provided concerning the given income that neither condition needs to hold if one merely adds an appropriate time-
distribution. That is to say, we would have to explain why income is differential function for price changes to assure convergence to an equilibrium
sufficiently well distributed such that the kind of income“expenditure price over time. The stability of such a market determination of price depends
entirely on an arbitrarily chosen coefficient representing the speed of response
situation Hicks and Marshall describe for the Giffen paradox could never
[see Lancaster 1968, p. 201]. For a discussion of the methodological problem
occur. Of course, that theory of income distribution must also avoid
posed by this ad hoc dynamics strategy, see Boland [1986a, Chapter 9].
contradictions with our laissez-faire individualism. If the so-called 9 Or at least not positively sloped if the supply curve is not vertical. Note that the
Cambridge controversy over capital and distribution is any indication, the argument would hold even if we were only concerned with one type of stability
possibility of such a neutral theory of income distribution does not seem as we would still have to distinguish between (a) and (d) or between (c) and (e)
of Figure 14.1.
promising.
10 For a discussion of using models to test theory, see further Boland [1989,
Chapters 1 and 7].
© LAWRENCE A. BOLAND
212 Principles of economics
11 In particular, Lipsey and Rosenbluth [1971] argue that Giffen goods are more
Epilogue
likely when we base utility on ˜characteristics™ rather than the goods
themselves. Unfortunately, they use Lancaster™s linear model of the
relationship between goods and characteristics and it is the linearity alone
Learning economic theory
which produces their result. There are many possible non-linear models of
characteristics production which would yield the Hicksian conclusions
through criticism
concerning ˜likelihood™.
12 He suggests that we assume that ˜individual utility functions [are such] that the
absolute value of the deduced income effect is less than the absolute value of
the deduced substitution effect in the case of inferior goods™ [De Alessi 1968, p.
293]. This would seem to be as testable as Lloyd™s considerations, only a little
more complicated.
13 For example, the Giffen paradox can be avoided by assuming ceteris paribus
the constancy of the marginal utility of money and then with an additive utility
function using diminishing marginal utility we can explain the Law of Demand. Some opponents of neoclassical economics will complain that my
Any substitution as the result of a change in price would change the marginal
exploration of ways to criticize neoclassical theories was not exhaustive. I
utility of money, hence rendering this theory of demand untestable. With regard
welcome them to take up any other line of criticism they might have in
to such counter-critical uses of ceteris paribus clauses Welty would be quite
mind. My interest has been to develop a clear understanding of neoclassical
correct but Lloyd does not use ceteris paribus in this manner.
14 In philosophy literature, this is known as the ˜Duhem“Quine™ thesis, see theory by determining the essential ideas that are used to form any
further, Boland [1989, Chapter 7]. neoclassical explanation. Trying to pin down the essential ideas is
15 De Alessi™s added clause might not be independently testable or it might only
sometimes difficult because neoclassical economics always seems to be a
be more difficult to test than other statements contained in the traditional theory
moving target. I remember conversations (arguments?) with radical
(such as the fixity of money income, fixity of prices of other goods, etc.). On
Marxist students in the 1960s who often would claim to have the definitive
this matter De Alessi™s modification may not seem to be very problematic. The
only criticism Welty can give reduces to the accusation that De Alessi offers a critique of neoclassical economics. Whenever they explained their criticism
˜demonstrably arbitrary™ modification of traditional demand theory. That is, De to me it always seemed that they were criticizing economics as it was
Alessi™s modification is ad hoc.
understood about 1870. These conversations convinced me that if the
critics really wanted to form effective criticisms of neoclassical economics
they should learn more about how neoclassical economics is understood
today. The more they understand neoclassical economics the better will be
their critiques. The fear in the 1960s was always that one would be
indoctrinated if one went through a formal process of learning neoclassical
economics. Indoctrination might be possible but nevertheless I cannot see
how one can form an effective criticism of neoclassical economics without
a clear understanding of neoclassical theory.
When it comes down to its essential ideas, neoclassical economics
seems now to have settled down into the clear research programme which
was fairly well defined in the 1930s. Of course, the techniques of
modelling neoclassical theories have changed significantly over the last
fifty years and it is all too easy to confuse advancements in techniques with
improvements in essential ideas. While some of the rhetoric is different,
there are two identifiable streams. On the one hand there is the approach of
Marshall and his followers. On the other there is the one developed by
Hicks and Samuelson which follows Walras. Both are based on the
neoclassical maximization assumption. Both are concerned with the
© LAWRENCE A. BOLAND
214 Principles of economics Epilogue 215
necessary conditions which follow from the existence of a competitive neoclassical model would seem not to require uniqueness or completeness.
equilibrium. While over the years the means of determining the necessary But the question remains whether a neoclassical model based on local
conditions have varied widely, the necessary conditions of interest are the maximization can be a basis for understanding why prices are what they
same for both. are and not what they are not. Unless one has shown that the prices are
The source of the necessary conditions is the maximization assumption consistent with global maximization the possibility exists that there are
and the details are due to the particular form assumed for the objective multiple local optimal prices that could have been obtained. Whenever
functions (utility or profit). But Marshall, the mathematician, had a deeper there are many possible sets of general equilibrium prices within an offered
understanding of necessary conditions than mere technical questions explanatory model, the question is begged as to why the world faces the
concerning the form of the objective functions. The questions that have one set of equilibrium prices rather than any other logically possible set of
preoccupied the followers of Walras are almost exclusively concerned with equilibrium prices. If we understand prices by being able to explain them
what assumptions one must make about the form of the objective functions then the basis of our explanation is a critical issue. The basis for
to assure an equilibrium. Marshall clearly understood that one cannot understanding is not just the neoclassical maximization hypothesis but, I
explain an individual™s behaviour as a matter of choosing the optimum am arguing, it also includes the assertion that those are the only possible
unless there is sufficient freedom to choose other options. This he prices.
expressed with his Principle of Continuity which is a reflection of his What I am saying here about the requirement of understanding is not
approach that focuses on the necessary conditions by analyzing the widely accepted by economic theorists today. This is partly because most
calculus-based neighbourhood properties of any equilibrium. For Marshall economists today think that if there is any problem with neoclassical
the idea of the availability of alternative options translates into the economics it is most likely a technical modelling issue. Few economists
requirement of a continuum of options. So, from Marshall™s perspective, think there is anything fundamentally wrong with their notion of
one says that one understands phenomenon X because one has assumed explanation or understanding. Unfortunately, if the question of uniqueness
that X is the logical result of maximization given that the decision-makers and completeness is considered to be a mere technical modelling question,
had numerous alternative options from which to choose. Moreover, prices it can be dismissed since any model which might provide uniqueness or
must matter in the individual™s choice if the logic of the choice process is to completeness is usually ˜intractable™. So much for tractable models! The
be used to explain prices. If one™s choice is limited to an extreme point on question I ask is just how do we understand prices?
the continuum then one can explain the choice without reference to prices Put in more methodological terms, how do we know when a neoclassical
and thus prices do not matter. Clearly, one cannot explain or understand explanation of price is false? If we say we understand prices with a
prices with a model in which prices might not matter! neoclassical explanation then conceivably we must be recognizing the
Marshall™s [1920/49, p. 449] understanding that one cannot generally possibility that such an explanation could be false “ otherwise it would be a
assume that knowledge is perfect implicitly recognizes that knowledge is vacuous tautology! Any claim that says you know why the world is what it
important. Yet few if any neoclassical models try to explain how the is must entail an assertion that you know why the world is not what it is
maximizing individual decision-maker knows the prices or income or even not. Whenever people claim to have explained something, the challenge is
knows the utility or profit functions. Attempts to give a neoclassical for them to explain what evidence it would take for them to admit that their
explanation of knowledge by explaining the economics of information [e.g. explanation is false (if it is false). This is my challenge to believers in
Stigler 1961] begs the question of how information becomes knowledge “ neoclassical economic explanations of prices. What would neoclassical
do we always have to assume knowledge is acquired inductively? economists accept as a situation that would force them to admit that they
Leaving aside the difficult question of explaining knowledge, to what might not actually understand why prices are what they are?
extent do we understand fundamental things like prices with neoclassical
models? If our understanding is that all prices are general equilibrium
prices then at least logically the explanatory basis will be adequate but only
if those prices are the only prices implied by our model. This raises the old
problem of whether one must require uniqueness or completeness in
models. If we are only interested in local maximization then a successful
© LAWRENCE A. BOLAND

Bibliography


<<

. 23
( 25 .)



>>