JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms
Springer is collaborating with JSTOR to digitize, preserve and extend access to Zeitschrift
für allgemeine Wissenschaftstheorie / Journal for General Philosophy of Science
This content downloaded from 129.49.5.35 on Wed, 06 Sep 2017 00:01:55 UTC
All use subject to http://about.jstor.org/terms
Epistemology of Economics
PAUL T. SAGAL
Summary
Methodological disputes in economics have been with us since Mill and Senior fought
over the nature of economic science in the first half of the 19th Century. Progress has
been extremely slow, and there is good reason for this as the present essay hopes to show.
Three important methodological positions are examined critically: the "ultra-empiri
cism" of T.W. Hutchison, the "moderate empiricism" of Milton Friedman, and the
"extreme a priorism" of Lionel Robbins and Ludwig Von Mises. The argument be
tween Guttierrez and Block in Theory and Decision over praxiology is discussed in con
nection with the last mentioned position. The paper concludes that "extreme a priorism,"
though very much out of fashion is not without its resources. The work of the contem
porary German philosopher, Paul Lorenzen, is enlisted to bolster this position.
This content downloaded from 129.49.5.35 on Wed, 06 Sep 2017 00:01:55 UTC
All use subject to http://about.jstor.org/terms
Epistemology of Economics 145
information about the real world. It is about space, that extended substance
which surrounds us. In short, geometry gave a highly intelligible uniquely
true characterization of a significant portion of reality. It is no wonder that
especially the 17th Century rationalist philosophers were so enamoured
with geometry. For Spinoza, even ethics was to be put in more geom?trico.
By the time arithmetic and generalized arithmetic (arithmetic with letters
? algebra) were put in axiomatic form, something had happened to the
geometrical ideal of knowledge. For one thing, Euclid's geometry contain
ed logical gaps. For another tiling, Euclid's geometry had competition. The
necessity to choose between Euclid and his competitors had serious conse
quences. The certainty of Euclidean geometry seemed to be grounded in
the epistemological transparency of its basic terms and principles. It just
had to be true. Now the whole notion of substantive truth ? truth about
reality ? as grounded in the evidence of basic principles seemed to go by
the boards. How can we be sure that some other geometry isn't ,truec of
reality? The natural move to make here, and the move actually made, is not
difficult to foresee. A distinction is made between internal truth and external
truth. The former is a matter of logical consistency and valid deduction,
the latter a matter of empirical test, of application. What made Euclidian
geometry so attractive epistemologically was its apparent capacity to ob
viate such a distinction. The reader with roots in the philosophical tradition
will of course interpret the history of geometry as involving the birth and
death of the synthetic a priori (I remind the reader that I do not necessarily
subscribe to this more or less official story of geometry).
The epistemology of empirical knowledge too can be viewed as imitating
the axiomatic style of geometry. In fact, 18th Century empiricists differed
from 17th Century rationalists principally in what they chose for basic terms
and principles. The geometrical (axiomatic) ideal remained. In empirical
knowledge what you want to derive is Newton's laws, the laws of physics.
What makes these laws true is experience, and from these laws you can
explain and predict unknown experience. You don't want to take Newton's
laws as axioms, because then you would have to claim just as in the case
of geometry that the truth somehow was grounded in epistemological
transparency. For the empiricists this would make physics too close to geom
etry (to mathematics). Matters of fact are to be sharply distinguished from
relations of ideas. But how can empirical laws be deduced from observation
(experience) ? They can't. They can at best be induced. Then what happens
to the grounding of empirical laws in observation? We can look at the
matter in two different ways : (1) We supplement deductive logic with an
empirical or inductive logic. We then logically derive Newton's laws let us
say from additional axioms. (2) We supplement our observation statement
axioms with axioms of epistemological nature, some version of a uni
formity of nature principle. Then from this expanded set of axioms we
derive the laws of physics etc. Needless to say all this has the air of make
believe about it. Still it wasn't until the demise of logical positivism in the
late 1930s that this inductivist view of empirical knowledge was dealt a
10
This content downloaded from 129.49.5.35 on Wed, 06 Sep 2017 00:01:55 UTC
All use subject to http://about.jstor.org/terms
146 Paul T. Sagal
death blow. (Of course, it is by no means dead for everyone. It still has a
faint heartbeat).
This official story leaves us in a rather uncomfortable position. Ration
alism and empiricism, the keystones of traditional epistemology, seem
vanquished. Something must rise from the ashes to rescue epistemology.
In the Mill/Senior debate, Mill was the empiricist and Senior was the ra
tionalist. The history of epistemology appears to have refuted both. (Neither
can stand without modification.) Inquiry into the foundations of economics
has several heavy crosses to bear : (i) the relation of economics to the empirical
sciences, e.g. physics on the one hand, and the non-empirical sciences e.g.
geometry, on the other hand, is not easy to set forth (ii) the epistemology
of both the empirical and the non-empirical sciences in general is (and
indeed has always been) too controversial to depend on. (iii) Economics
seems to flirt with normative and moral 'sciences' and this further compli
cates its epistemology. Is it any wonder that in economics "questions of
methodology are among the most difficult to reach agreement on or even
to find a basis of discussion for."1
The old empiricism can be labelled (following Sir Karl Popper) inductivist.
It saw the empirical sciences as based upon observation sentences. First
principles were to be somehow derivable from these sentences. But it is a
simple logical point that principles of a universal sort (scientific laws),
are not derivable from a finite set of observations. And since the set of all
verified observation sentences has to be a finite set, that these principles were
not so derivable. The move usually made by the old empiricism was to
supplement the rules of deductive logic with inductive rules, these together
being sufficient for deriving empirical laws. Alas, nothing like the relative
unanimity in deductive logic ever appeared in the inductive case. A related
attempt to salvage the old empiricism was to rest satisfied with high prob
ability of first principles. The inductive logic could serve as a probability
logic and we could provide a foundation for the probable truth of the
principles in question. However, as Hume was the first to point out, the
shift to probability will not do, for the justification of probability rules was
every bit as difficult as the justification of the inductive rules in the stronger
sense, i.e. the sense in which the truth of principles and not merely their
high probability, was derivable. The old empiricism has to give up both
inductive proof and degrees of confirmation. But empirical knowledge seemed
then to be without justification, without foundations altogether. We appear
caught in the throes of Humian scepticism. The empiricism attempted to
draw a sharp line between empirical science on the one hand and logical
or mathematical science on the other. (It also tried to sharply separate both
from non-science and metaphysics.) Now empirical knowledge seemed to
have vanished. The inductivist old empiricism was too ambitious a theory.
Empirical knowledge could not be salvaged inductively. Proof and confirma
This content downloaded from 129.49.5.35 on Wed, 06 Sep 2017 00:01:55 UTC
All use subject to http://about.jstor.org/terms
Epistemology of Economics 147
2 For the development of Popper's philosophy, see Conjectures and Refutations and
Objective Knowledge. Popper's truly fundamental work remains The Logic of Scientific
Discovery. See also I. Lakatos' penetrating discussion of Popper's thought in Lakatos
Musgrove volume, Criticism and the Growth of Knowledge?)
3 T. W. Hutchison, The Significance of Basic Postulates in Economic Theory.
10*
This content downloaded from 129.49.5.35 on Wed, 06 Sep 2017 00:01:55 UTC
All use subject to http://about.jstor.org/terms
148 Paul T. Sagal
But this new empiricism had its problems, some of which were fore
shadowed in the history of western philosphy. The line drawn appeared too
confining. Not all empirical statements, not all parts of empirical science,
seemed equally empirical. The basic postulates of science did not confront
falsification with the same forthrightness as statements like All ravens are
black, and If a two pound weight is placed on this string it will collapse. In New
tonian science we had the laws of motion which seemed barely empirical.
Could one falsify the law of inertia? Still Hutchison would insist that the
problem with such examples was that thinkers had been epistemologically
shifty in treating them. They had played (and still play, for that matter),
the now it's empirical, now it's not, game. The scientist has to make up his
mind. They are one or the other. He cannot eat his cake and have it too.
Kantianism is one important historic line of thought which attempted
to resist the above argument. It is not that scientists are being shifty, but
rather that methodologists have blinded themselves to a significant and
idiosyncratic part of empirical science ? the a priori part of empirical science.
For empiricism, old and new, there just 'ain't' such an animal. We will
come back to Kantianism since it is the epistemological stage on which
Hutchison's methodological enemies stand.
Another critical reaction to empiricism is the Duhem/Quine line. The
operative word here is holism. We will, as has been our custom, paint in
broad strokes. On this view, theories to conjectures do not confront test
and possible refutation in isolation. When a theory is tested, not only it
but an infinite number of background assumptions ? any assumptions
which were necessary to deduce the consequent which turned out false are
also tested. It is not written in heaven what adjustment among all these
assumptions should be made. There is no commandment to give up the
theory. We could also keep the theory and give up something else. Further
more, where we push this line to its limits, there is nothing in the whole
science which we could not in principle give up. So either all science, the
system, a priori and a posteriori, is testable or nothing is. We cannot go
into the empiricist counterattack here - but the point we want to make is that
both from the Kantian and the holistic epistemological perspective, the
methodological strictures of the new empiricism have to be dropped. Hut
chison's work in economics is of course dependent upon the defensibility
of the new empiricism.
A modest reaction to Hutchison's empiricism is reflected in Milton
Friedman's very influential essay, On the Methodology of Positive Economics.*
Friedman is looking for elbow room. He is worried about the demand that
basic postulates themselves be subject to empirical test.
Friedman's influential essay is usefully viewed as a reaction against what
has been termed (by F. Machlup) Hutchison's ultra-empiricism. But a
caveat is in order. Friedman's essay is a difficult one. Certain key terms
This content downloaded from 129.49.5.35 on Wed, 06 Sep 2017 00:01:55 UTC
All use subject to http://about.jstor.org/terms
Epistemology of Economics 149
are not sufficiently explained and there is some question even about the
consistency of the line of argument. Furthermore laying out the line of
argument has its own difficulties since as indicated key terms prove trouble
some. Friedman is afraid that empiricist strictures a la Hutchison would
outlaw much useful economic theorizing. More pointedly, Friedman holds
that assumptions like perfect competition, and profit maximization need
neither be true nor testable to be useful for prediction. Usefulness is a
matter of predictive reliability. This is Friedman's pragmatism, indeed
his instrumentalism. Look at assumptions as tools; do they give us the
results we are after? We are to judge theoretical assumptions by their fruits
and not by their roots (in reality). This view has all the troubles of any
brand of extreme pragmatism. It has difficulty in drawing the line between
scientific liberty (in theorizing) and scientific license. Assumptions needn't
be true, they needn't even be empirically testable. They just have to yield
appropriate predictions. How are we to choose among alternative useful
theories? Here the hard-line pragmatism is softened. We can look for
parallels with the successful theories. We can compare them on some
measure of simplicity. (It is of course not easy to spell this out.) Basically
Friedman relies upon some notion of fit or coherence with other parallel
theorizing. But of course there is something circular about this. Each
competitor theory is to be judged by its coherence with successful parallel
theories. But how did we choose the superior parallel theories from among
its competitors? Some criterion independent of coherence is apparently
required. Yet how are to admit one, without bringing the realism, the truth,
the empirical meaningfulness of assumptions back into the picture?
Lionel Robbins and Hutchison spoke of basic postulates. And it is
clear that they had such things in mind as profit maximization. Friedman
on the other hand speaks of assumptions ? and clearly includes profit
maximization in this category. Are assumptions simply postulates, first
principles ? axioms? They had better be. For the logic of assumptions
otherwise is something terribly murky. Friedman speaks of hypotheses
having both observational implications and assumptions. In testing a
hypothesis the truth or falsity of the implications matter, if the assumptions
do not. A way of clarifying this kind of talk would be the following. An
assumption is a statement to which the hypothesis is a logical consequence.
(It is a logical consequence if the assumption were one of the statements
used in deriving it.) An implication is simply a logical consequence of the
hypothesis ? same characterization of logical consequence as above. Fried
man's position would then amount to this. Epistemological requirements:
truth realism, testability, etc. are passed up to a hypothesis from its impli
cations and not down to a hypothesis from its assumptions. Assumptions
are judged by the fruitfulness of the hypotheses which are logically-conse
quent to them.
This alone would solve some crucial problems in Friedman-interpretation
But the key separation of assumption from hypothesis is not easily made in
all cases. When Friedman speaks of assumption testing or evaluation ?
This content downloaded from 129.49.5.35 on Wed, 06 Sep 2017 00:01:55 UTC
All use subject to http://about.jstor.org/terms
150 Paul T. Sagal
This content downloaded from 129.49.5.35 on Wed, 06 Sep 2017 00:01:55 UTC
All use subject to http://about.jstor.org/terms
Epistemology of Economics 151
shouldn't concern ourselves with. Though this observation about the sur
vival value of profit maximization should perhaps be taken as mere obiter
dictum, it remains the case that Friedman is concerned with more than the
indirect consequences of theories. For one thing, he wants theories to be
simple or as simple as possible. For another, the examples he adduces of
'false' theories which are useful are examples of theories with other attrac
tive traits. They provide analogies or pictures of operation which are rela
tively easy to grasp. They provide a certain kind of understanding. They
tell us that things happen as //"things were the way the theory says. (This is,
it seems to me, the attractiveness of the billiard player explanation in terms
of geometrical optics). There are criteria, certainly, in choosing among
such as if explanations, whether we speak of understanding, heuristic value
or what have you. Indeed quite independently of a theory's pragmatic
value its role in understanding may make it relatively resistant to competi
tors. So although Friedman goes out of his way to emphasize the instru
mental, black box nature of theories, he does recognize what we may term
the epistemological value of theories. Theories render facts intelligible.
They can be effective engines for prediction generation without accom
plishing this.
In the 1955 volume of the Southern Economic Journal'there was an important
and extremely lively discussion of methodological issues. The chief protag
onist was Fritz Machlup. Machlup in his Verification in Economics distin
guishes two extreme positions ? ultra-empiricism which he identified with
the views of Hutchison which we have already encountered, and extreme a
priorism which he identified most importantly with Ludwig Von Mises.
(We will soon examine Von Mises's position) Machlup's own views were
very close to Friedman's. He, however, did not share Friedman's ultra
pragmatic (instrumentalistic) tendencies. Machlup stressed the idealizing
function of basic assumptions. These assumptions were necessary for the
scientist if not for the job of generating testable consequences. For the
latter, a large class of theories would do. But only a subset (perhaps one) of
these theories could serve as an idealization, as an instrument of inter
pretation, if you will. "For the fundamental assumption may be understood
as an idealization with constructs so far removed from operational con
cepts that contradiction by testimony is ruled out; or even as a complete
fiction with only one claim: that reasoning as if it were realized is helpful
in the interpretation of observations. And... "The fundamental assumption
is a resolution to proceed in the interpretation of all data of observation as
if they were the result of the postulated type of behavior."6
Hutchison did not by the way take the ultra-empiricist charge lying
down. He responded forcefully ? to say the least ? to Machlup's paper
pointing to some non ultra-empiricist passages in his work. Machlup, in his
reply ? Rejoinder to a Reluctant Ultra-Empiricist, ? argued that Hutchison
in spite of some statements to the contrary, deserved the ultra-empiricist
This content downloaded from 129.49.5.35 on Wed, 06 Sep 2017 00:01:55 UTC
All use subject to http://about.jstor.org/terms
152 Paul T. Sagal
This content downloaded from 129.49.5.35 on Wed, 06 Sep 2017 00:01:55 UTC
All use subject to http://about.jstor.org/terms
Epistemology of Economics 153
This content downloaded from 129.49.5.35 on Wed, 06 Sep 2017 00:01:55 UTC
All use subject to http://about.jstor.org/terms
154 Paul T. Sagal
This content downloaded from 129.49.5.35 on Wed, 06 Sep 2017 00:01:55 UTC
All use subject to http://about.jstor.org/terms
Epistemology of Economics 155
10 Ibid., p. 105.
11 See F. Hayek's discussions of this point in his Science, Economics, and Politics, first
essay "Degrees of Explanation."
12 What we have in Robbins is what Karl Popper termed the method of essentialist
definition. Popper saw that this Aristotelian procedure was just the kind of procedure
that his own methodological views could not tolerate. The method of essentialist or
somewhat less extravagantly, simply real, definition is of course one of the main ways
of defending something like the synthetic a priori. Popper's most penetrating discussion
of essential definition is found in his discussion of Aristotle's conception of knowledge
This content downloaded from 129.49.5.35 on Wed, 06 Sep 2017 00:01:55 UTC
All use subject to http://about.jstor.org/terms
156 Paul T. Sagal
in the second volume of Open Society and its Enemies. We have begged off confronting
these issues. One suggestion, though, at this point, may be of some value. It might be
that the appeal to real or essential definitions does have some place but only where one
is dealing with the fundamental (defining) terms or concepts of a theory. Beyond this,
they might, as Popper states, simply get in the way of empirical investigation.
13 Ibid. p. 116-117.
14 See J. Agassi, "Tautology and Testability in Economics," Phil. Soc. Sei. 1 (1971),
46-63. 15 p. 123.
This content downloaded from 129.49.5.35 on Wed, 06 Sep 2017 00:01:55 UTC
All use subject to http://about.jstor.org/terms
Epistemology of Economics 157
in connection with space are definitive of the concepts. They are evident
because they are definitional, and they apply to reality because the concepts
which they define are formed in response to the demands of experience
(experience which is too uniform to be illusory, or out of touch with
reality).16 Now this Aristotelian view has had its ups and downs. For a
long time however, only empiricists of sceptical leanings really rejected it.
In the Medieval period the Ockhamist tradition, and Humean empiricism
in the 18th century were the major anti-Aristotelian forces. Kant's Coper
nican revolution, it should be noted, was hardly revolutionary when it
came to this scientific Aristotelianism. In fact the Kantian attempt to legi
timize the synthetic a priori was in substance if not in terminology quite
congenial with Aristotelianism. The important thing in both was to have
science certain yet informative. Empiricism -? logical positivism ? for a while
threatened to bring down Aristotelianism once and for all. It was nour
ished on, among other things, certain developments in geometry. Euclid
ean geometry had competition. If there were any number of geometries,
all definitive, of any number of spaces, definition and self-evidence could not
by itself provide a criterion of truth. Truth had to be a matter of empirical
testability. If this was the situation in geometry, it was a fortiori the situation
with the rest of science. Empiricism seemed victorious. But two caveats are
in order here : (i) It does not follow that the rest of science must follow the
lead of geometry, (ii) It is not clear i.e. the situation is complex and difficult,
that the discovery of non-Euclidean geometry marks the demise of Aristote
lianism even in geometry. In other words Euclidean geometry still might
possess some privileged status. At this point we are interested simply in
keeping questions open. For this is really all we need to first locate Robbins'
position squarely in the Aristotelian locale and point out that Aristotelianism
is still an option philosophically. (In other words there is still room for
reasoned controversy. Aristotelianism has, after all, outlived a number of
fashions.)
The empiricist (ultra or otherwise) sees first principles as subject to test.
The Aristotelian sees first principles as certain ? and consequently as merely
requiring appropriate application. In practice, there really need be little
difference. Extremely well-entrenched (obvious) principles, e.g. the prin
ciple of non-contradiction, are for all practical purposes untouchable.
Robbins and Von Mises to whom he is indebted view the relation between
human action and basic principles as analogous to the relation between object
and the basic principles of logic (not so much propositional logic but what
S. Lesniewski called Ontology, laws of object identity, etc.).
A parallel which the author finds particularly helpful, though admittedly
hardly unproblematic is that between economics and arithmetic. There are
not many arithmetics (though there are different number systems) in the
way there at least seem to be many geometries. The fundamental principles
This content downloaded from 129.49.5.35 on Wed, 06 Sep 2017 00:01:55 UTC
All use subject to http://about.jstor.org/terms
158 Paul T. Sagal
of arithmetic are definitive of (are explicative of) the concept discrete quantity.
There is no question of (i.e. it would be artificial to speak of) testing the
principles of arithmetic. It is just a matter of properly applying them. Here
it is discrete quantity which is analogous to human action. We shall go into some
of these questions with greater depth in the following section. The Misian
approach to economics stands or falls with these fundamental philosophical
affirmations. Mises gives a Kantian turn to Aristotelian methodology and
we shall have to see whether this strengthens or weakens the position. We
move now to "extreme ultra-apriorism".
With all the debate on methodological questions, the extreme a priorist
position of Robbins has hardly received any contemporary attention. The
debate seems to have passed it by. This holds equally well of the methodolo
gical work of Von Mises, which forms much of the basis of Robbins' work.
The logical positivist revolution and its aftermath seem to have left Von
Mises' work by the wayside. What could be worse than an outdated, un
fashionable methodological perspective. But fashions change, revolutions
have their reactions, and perhaps the outdated, non positivist (Aristotelian
Kantian) methodological stance of Von Mises will once again have its day17.
In fact not so long ago in Theory and Decision there occurred a lively debate
over the viability of Mises' Praxiology ? his general theory of human action
of which Economics is a part. The central paper in this discussion was
Claudio Gutierrez's The Extraordinary Claim of Praxiology.18 The "extra
ordinary claim" is that the whole science of human action can be gleaned
from the single a priori category (concept) of human action. "Extraordinary
as this claim sounds, it has not received, as far as I know, a commensurate
rebuttal either from economists or philosophers." (p. 327) But Mises'
work in general has not received much attention, and "extraordinary" is
perhaps a presumptive term. Is it extraordinary that the concept of set, the
predicate of set-membership e is sufficient (with the aid of logic) to yield
all of classical mathematics? It was perhaps treated as extraordinary at one
time, but surely now if it is considered extraordinary it is not unfamiliar or
questionable. In fact, the history of thought reveals many cases where a
single notion, e.g. object for metaphysics or Lesniewski's logical system
for ontology, homogeneity for Euclidean geometry was treated as adequate
(at least thought to be adequate) for an entire science. In fact, the entire
"axiomatic" approach has as its ideal of getting as much as possible from as
little as possible. So I do not find the label "extraordinary" terribly helpful.
But then again, what's in a label? It's the argument that matters.
17 For a full systematic presentation of praxiology, see Ludwig von Mises, Human
Action, Yale, 1963. On specifically methodological issues, see the same author's Episte
mological Foundations of Economics, Van Nostrand. See also two important papers
of Murray Rothbard, "Mises 'Human Action* : A Comment," American Economic Review,
March 1956, "Praxeology: Reply to Mr. Schuller, American Economic Review, December
1951, and "The Defense of Extreme A Priorism," Southern Economic four nal.
18 Theory and Decision 1, (1971), pp. 327-336.
This content downloaded from 129.49.5.35 on Wed, 06 Sep 2017 00:01:55 UTC
All use subject to http://about.jstor.org/terms
Epistemology of Economics 159
19 p. 328.
20 But see, for a discussion of the complexities involved in idealization, Stephan
K?rner's Experience and Theory.
This content downloaded from 129.49.5.35 on Wed, 06 Sep 2017 00:01:55 UTC
All use subject to http://about.jstor.org/terms
160 Paul T. Sagal
This content downloaded from 129.49.5.35 on Wed, 06 Sep 2017 00:01:55 UTC
All use subject to http://about.jstor.org/terms
Epistemology of Economics 161
11
This content downloaded from 129.49.5.35 on Wed, 06 Sep 2017 00:01:55 UTC
All use subject to http://about.jstor.org/terms
162 Paul T. Sagal
Prof. Dr. Paul T. Sagal, Dept. of Philosophy, New Mexico State University, Las Cruces,
New Mexico 88003
This content downloaded from 129.49.5.35 on Wed, 06 Sep 2017 00:01:55 UTC
All use subject to http://about.jstor.org/terms