The Online Library of Liberty

A project of Liberty Fund, Inc.

• Editor’s Preface
• Ludwig M. Lachmann, an Austrian Stocktaking: Unsettled Questions and Tentative Answers
• John B. Egger, the Austrian Method
• Mario J. Rizzo, Praxeology and Econometrics: a Critique of Positivist Economics
• Israel M. Kirzner, Economics and Error
• S. C. Littlechild, the Problem of Social Cost
• D. T. Armentano, a Critique of Neoclassical and Austrian Monopoly Theory
• Gerald P. O’driscoll, Jr., Spontaneous Order and the Coordination of Economic Activities∗
• Murray N. Rothbard, Austrian Definitions of the Supply of Money
• Laurence S. Moss, the Emergence of Interest In a Pure Exchange Economy: Notes On a Theorem Attributed to Ludwig Von Mises
• Roger W. Garrison, Austrian Macroeconomics: a Diagrammatical Exposition
• Louis M. Spadaro, Toward a Program of Research and Development For Austrian Economics

Praxeology and Econometrics: A Critique of Positivist Economics

Mario J. Rizzo

• New York University

The ultimate goal of a positive science is the development of a “theory” or “hypothesis” that yields valid and meaningful (i.e., not truistic) predictions about phenomena not yet observed.¹

Although written a quarter of a century ago, Milton Friedman’s “The Methodology of Positive Economics” remains the immediate philosophical justification for much of the contemporary approach to economics research. Nevertheless, the general points raised in that essay were not new even at the time, but were an ingenious adaptation of some of the positivist arguments of the 1930s, and the somewhat revisionist work of Sir Karl Popper.² Today, thoroughgoing positivism is clearly in retreat, if not already defeated, in philosophical circles, but a variant of it remains quite vibrant in many of the social sciences, particularly economics. It is the task of this essay to present a critique of “positive economics” and, at least, some indications of a viable alternative.

I.

PREDICTION AS THE GOAL

From the positivist epistemological viewpoint, is the opening quotation to be taken as an a priori or an empirical statement?

If a priori, then it is a statement about how we shall use the term “positive science” and is merely a linguistic stipulation. As such, one might equally well choose to stipulate some other meaning.

If empirical (i.e., a statement about what people have in fact considered positive science), then, of course, it does not express a necessary truth and could be otherwise. But then for a long time the Darwinian theory of evolution yielded no predictions and yet was considered scientifically acceptable.³

Furthermore, Friedman makes no attempt to survey what has been considered economic science to find out whether “prediction” has indeed been the defining characteristic. In fact, there are many theoretical frameworks which generate no testable predictions but are, nonetheless, considered part of economics. For example, it is frequently unclear what (predictive) relevance discussions on the existence and stability of equilibrium under many special assumptions (the empirical significance of which is unknown) have for a world which is never actually in equilibrium. Of course, one might claim that this is bad economics, and so the demarcation is really between “good” and “bad” science. There is, however, no escape here, for it merely leaves unanswered the question: Why is nonpredictive economics bad science?

Another possible escape might be to claim that, while nonpredictive theories may be scientific, they do not qualify as positive science. To this we are justified in merely replying: “So what?” What advantage is being claimed for positive science except that its ultimate goal is prediction? In that case, we are back where we started: Why must prediction be our goal?

The goal of prediction might well obscure what has in fact been considered a worthwhile aim of science: the explication and apprehension of necessary connections. Purely predictive “theory” is little more than a mnemonic device designed to relate x to y. But the nature of that relation may be unknown. The Babylonian astronomical forecasting techniques, which were merely trial-and-error arithmetic calculations, are an example of this kind of “black box” framework.⁴ The principle of explanation remains unknown in the sense that the connection between the initial and marginal conditions (x₁, x₂, x₃, etc.) and the consequence (y) is not apprehended as necessary. The relation is characterized by an arbitrary givenness.

But while it may be true that prediction cannot be considered a sufficient attribute for “scientific” theory, it still might be a necessary one. However, we have already implicitly refuted this assertion by showing that within a positivist epistemological framework such necessity can be derived only from an essentially arbitrary prior stipulation.

II.

FALSIFIABILITY AS THE CRITERION OF MEANING

The emphasis on prediction as the aim of science has its roots in a positivist criterion for the meaningfulness of a statement. To be meaningful, it has been said, a statement must be in a form such that it is in principle falsifiable by any observer.⁵

For example, let us take “the hypothesis that a substantial increase in the quantity of money within a relatively short period is accompanied by a substantial increase in prices.”⁶ Aside from problems concerning data availability and the skills of the particular investigator, is this hypothesis falsifiable? For now, let us say it is. Hence, the positivist would claim that this is a genuinely scientific statement. In fact, the meaning of a hypothesis is identified with the relevant test of its veracity. As Moritz Schlick tells us, “the meaning of a statement can be given only by indicating the way in which the truth of the statement is to be tested.”⁷ Of course, this cannot be literally true. If meaning is identified with the test, then what is being tested? But, if there is a meaning independent of the test, then the positivist criterion falls in on itself, and unfalsifiable statements can be meaningful. If we are not to take Schlick’s statement literally, then it seems difficult to find any coherent interpretation of it.

But, of course, the whole concept of a unitary criterion for meaning is somewhat strange when viewed from within a positivist framework. Once again: Is the proposition a priori or empirical? A stipulated definition of “meaning” to include falsifiability is not in itself impressive: One could have stipulated otherwise. Viewed empirically, the criterion is immediately refuted by two thousand years of Western philosophy which claims that metaphysics and ontology are meaningful pursuits.

Aside from these issues, the falsifiability criterion loses much of its initial plausibility when the contradictory of a falsifiable statement is examined.⁸ If we admit as falsifiable that all inflations are caused by increases in the money supply, then the contradictory,⁹ some inflations are not caused by increases in the money supply, is not falsifiable. If the latter hypothesis is meant to apply to the future as well as to the past, one could always claim that the inflation not caused by money supply increases will appear if you just search long and hard enough. No example of money-supply-induced inflation refutes the proposition, and with a future, as well as a past, time horizon one has an infinite pool of inflations within which to search for the complete absence of nonmonetary inflations.

Consequently, the falsifiability criterion involves a major transformation in our system of logic: Although a given statement may be meaningful (or scientific), the negation of that statement is meaningless (or unscientific).¹⁰

A possible route of escape from this argument might appear to be the claim that while, strictly speaking, the statement that some inflations are not caused by increases in the money supply is not falsifiable, evidence could be accumulated which would render it more or less “probable.” Alas, this is no escape either. The truth or falsity of any statement is not a random variable like tosses of a coin, and hence a frequential interpretation of the “probability” concept is impossible here. So the meaning of the term “probable” can only involve a subjective degree of belief. This amounts to a radical transformation of the whole positivist framework. The criterion now becomes: Any statement which could be rendered more or less “probable” by reference to empirical evidence is a meaningful statement. But then this is a psychological—rather than a logical—criterion. Any proposition for which our subjective degree of belief could be increased or decreased by “evidence” is meaningful. Worse still, what kinds of statements does this criterion exclude? Probably none. It would seem that human beings are not imaginative enough to conceive of propositions that have no relationship at all to the world. Hence, for any nontautologous (in the narrowest sense) statement, it is possible to find empirical “evidence” that has some bearing on its truth or falsity. Hence, all statements are meaningful. If this is so, then the original intent of the positivist criterion crumbles.

Any statement of degree-of-belief probability does not fit comfortably within the positivist framework. Statements such as “that some inflations are not monetarily induced is ‘probable’” are, of course, neither verifiable nor falsifiable in principle. More importantly, they do not carry with them any element of intersubjective testability (which was such an important goal). A stipulation that certain kinds of evidence will be interpreted as making a statement “probable” is no real solution. This makes the criterion of meaningfulness (or the demarcation between science and non-science) purely conventional.

III.

CRITIQUE OF ECONOMETRICS¹¹

Ceteris paribus prediction is prediction of “stylized facts”: x leads to y if other factors are held constant. But since, in general, they aren’t, we are not predicting a “real-world” event. Rather, we are predicting a hypothetical consequence.

To subject the hypothesis to potential falsification, we must control for the other relevant variables. Suppose we try to do this by using multiple regression analysis. Then:

• 1 How do we know when we have adequately controlled for extra-economic factors? (There is no a priori assurance that economic factors are the only ones that matter in a given situation.) This would require a theory of the interaction between economic and non-economic variables. How do we go about subjecting this to falsifying tests?
• 2 How do we test the theory which enables us to determine the other economic factors that must be held constant in order to isolate the effect of x?

From the positivist framework the problem is crucial. How could we ever know that the (auxiliary) hypothesis, i.e., all other relevant factors have been held constant, has been falsified? We obviously cannot claim that it has been refuted if x does not result in y because it is that very relationship which is undergoing testing in the first place. It is clear that, unless we have additional hypotheses about the effects of each of the to-be-held-constant variables on y, we shall not be able to subject the crucial ceteris paribus clause to refutation. Furthermore, these auxiliary hypotheses (or perhaps a single hypothesis since it is their total effect with which we are concerned) must be independent of the central one in the sense that the falsification of the former must be independent of the falsification of the latter. Now, if we claim that we really don’t care if the ceteris paribus clause is “true” because all that counts is the predictive ability of the central hypothesis, then we have gotten ourselves into a new quagmire. First, why have ceteris paribus clauses at all? Second, what are we falsifying if, in fact, x does not result in y? Certainly not the hypothesis as stated. Suppose the “evidence” fails to refute our hypothesis; then what have we corroborated? Again, not the original hypothesis because the apparent consistency of the data with the framework may be illusory, being entirely due to the “proper” variation of the factors which were supposed to be constant. Third, this whole viewpoint reinstates the “black box” approach to science and hence vitiates the aim of rational explanation.

It is quite possible to claim that, although the central hypothesis must be falsifiable in order to be meaningful or scientific, the ceteris paribus clause need not be. All that is needed in the latter case—it might be asserted—is a kind of educated judgment or verstehen. While this might be permissible within other epistemological frameworks, it will not be adequate to support the claims of positivism. If we can say that “all other relevant factors have been held constant” without falsifiability and still can be making a meaningful empirical statement, why can’t we do the same in the case of “x causes y,” the central hypothesis? If we can (which seems likely given the initial admission), then once again the criterion of positive science crumbles.

IV.

MAXIMIZATION

Under the influence of the “marginalist revolution,” economics has become a discipline devoted in major part to the finding of functional maxima and minima. The individual consumer or producer is assumed to maximize or minimize something and, from this postulated behavior, testable implications are drawn. It is important to keep in mind that the maximization behavior itself is not subject to falsification, because it serves not as a substantive hypothesis but as a superstructure which gives rational coherence to the falsifiable implications.

Any particular instance of concrete behavior may be “explained” or rationalized in terms of maximization (or minimization) of some appropriate quantity (e.g., utility, wealth, etc.). Since maximization is fundamentally a characteristic of intention (this the positivists won’t admit), any concrete behavior may be viewed as if it were the maximization of something. This has serious implications.

Suppose we wish to test not the applicability of a specific economic hypothesis to a given area of human behavior (say, marriage), but, rather, the validity of viewing this kind of behavior as an instance of economic or maximizing activity per se. In other words, we don’t care whether a particular maximizing model is appropriate, but we ask whether this is an example of maximizing behavior at all.

It might be claimed that this formulation of the problem makes no sense. After all, we are never testing economics or maximizing behavior as such, but only specific hypotheses of whatever kind. This, of course, misses the crucial point of the need to decide upon a research framework in advance of specific cases.

Is the statement “this is an example or instance of maximizing behavior” a meaningful and scientific one? Clearly not. Since the set of possible falsifiers is empty, any behavior can be “explained” in terms of maximizing something.¹² But the hypothesis, “this is an instance of maximizing sales,” can be refuted by appropriate behavior, and so is a meaningful statement. This produces a curious paradox. The more general statement about maximization is meaningless (or unscientific), but the more particularized version of it constitutes a positive scientific hypothesis.

Some authors have tried to escape this problem by claiming that the (maximizing) framework can be refuted by comparison to an empirically richer and more general alternative framework. Indeed, Lakatos has gone so far as to say, “There is no falsification before the emergence of a better theory.”¹³ This means, in effect, that if two hypotheses—one maximizing and the other non-maximizing—both equally well “explain” a particular case of economic behavior, then the one which is part of a more general approach, the specific applications of which have been corroborated in other cases, is to be preferred. This, however, introduces a subtle and important change in the falsifiability criterion. No longer is a statement meaningful or scientific by virtue of its empirical content but, rather, by the overall corroborated empirical content of other statements to which it is in some sense related. It is hard to recognize this as an epistemological criterion rather than as an aesthetic one.¹⁴ Nevertheless, by some inexplicable train of thought, a statement becomes meaningful because of its relation to other similar statements which, having been corroborated, are themselves meaningful by virtue of their relation to, say, the former hypothesis. (Apparently, there is some kind of “simultaneous determination of meaning” argument underlying all of this.)

Let us look at this problem in a slightly different manner. The maximizing framework “proves” its worth, we might say, by predicting everything that the alternative framework does, plus a little more.¹⁵ Hence, it acts, in a sense, as a falsifier of the alternative perspective.

This formulation does not seem very convincing. In economics, at least, it would be surprising if, say, the maximizing framework predicted literally all of the facts predicted by the alternative. Normally, I suspect, the “better” framework would predict some of these facts, and some additional ones. Furthermore, competing frameworks frequently do not even ask the same questions. Why, then, should they be judged on whether they give the same answers (plus a little more)?

All this aside, it is hard to see why, from a purely positivist epistemological perspective, considerations of the framework’s success in other particular instances should affect the meaningfulness or scientific character of a hypothesis in any given specific case.

V.

EVIDENCE

Until this point, we have implicitly considered as self-evident the answer to the question: “What shall count as evidence for and against a hypothesis?” How do we recognize a falsifying or corroborating result? The answer is, indeed, far from self-evident. In fact, this issue poses some crucial problems for the positivist approach, which, we shall contend, it is incapable of handling.

A hypothesis relates a variable x to a variable y, ceteris paribus. Let us assume that the ceteris paribus clause has been corroborated adequately; then what would amount to falsification of the hypothesis? To be more specific, hypothesize “that a substantial increase in the quantity of money ... is accompanied by a substantial increase in prices.”¹⁶ In order to test this statement, we must have some criteria by which we can relate the theoretical terms “money” and “prices” with their empirical counterparts. This is the crux of the problem.¹⁷

Something must point the way from theory to the relevant “facts”; we need what shall be called “referential statements.” In our illustration, examples of referential statements might be: “The empirical counterpart of theoretical ‘money’ is M₁“; or, “by ‘prices’ is meant the consumer price index.” The need for referential statements in applied economics is not restricted to the positivist variant of the science. What is peculiar to positivist economics, however, is a problem arising out of the epistemological status of such statements. If they are to be considered a priori, then (from a positivist viewpoint) we are merely talking about how we use words, and no link between the theoretical constructs and “empirical reality” is established. Then it must be established via falsifiable hypotheses. Yet this is an impossibility. (Referential statements make no predictions; they do not say, for example, that an increase in x results in an increase in y. Hence no predictions can be falsified.)

Now it is possible to recast the referential statements in such a way that they will be refutable: “If the criteria of applying the theoretical construct ‘money’ are, in fact, applied, then M₁ will be found to be the appropriate empirical counterpart.” Clearly, this won’t work because it requires that we know the criteria prior to the testing procedure which was to establish (or at least corroborate) these criteria in the first place.

Testing the referential statements is impossible unless we already know the criteria of applying the theoretical terms. If we already know these (in any meaningful way), then testing is unnecessary. But, from a positivist perspective, it is clearly impossible to have any meaningful knowledge about the real world which is given a priori.

One might attempt to obviate these difficulties by choosing empirical variables so as to present the particular hypothesis in its best light. (Choosing a definition of the money supply so as to best predict GNP is an example of this.) Unless one is attempting to insulate a hypothesis from refutation, there seems to be no clear reason for doing this. If empirical variables were chosen so as to present the hypothesis in its worst light, and it still remained unrefuted, would we not then have more fully corroborated it? In any event, the outcome of a potential test should not be the determining factor in whether it is performed.

VI.

THE LOGICAL CHARACTER OF PRAXEOLOGY

The epistemological status of praxeology (which is identical to economics very broadly conceived) is a subject of considerable misunderstanding and confusion. Within a positivist framework the claims of praxeology make no sense. Knowledge is either a priori and certain but not pertaining to “reality,” or it is empirical and uncertain but clearly embedded in the “real” world. An examination of the logical character of praxeology reveals these categories to be totally inappropriate. Praxeology claims to present knowledge which is at once both absolutely certain and empirical. This is the paradox which we shall have to explain.

Praxeological theorems or deductions are based upon the fundamental self-evident axiom, i.e., man acts or, what is the same, engages in purposeful behavior. The question at issue, then, is: In what precise sense is this axiom “self-evident,” and what does it say about the world?

The action axiom is empirical in the sense that it is derived from inner experience or immediate introspection. It is scientifically empirical because it passes the intersubjectivity test: The experience is universal and hence, in principle, can be assented to by the observers and the observed alike. Hence, the fact that the axiom is based on introspection cannot open the praxeologist to the charge that his deductions are of a purely personal and unscientific character. We are dealing here with “universal inner experience.”¹⁸

An attempt to deny the action axiom involves us in blatant self-contradiction. Denial consists of the use of means (arguments) to achieve ends (conclusions) and, hence, purposeful behavior. In addition, the assumption that men act is a necessary prerequisite for the existence of a scientific community. Arguments, attempts to convince other researchers of a different view, etc., are all fundamentally based on a conception of scientists themselves as engaging in purposeful behavior. To separate out the scientists, and say that while the observers engage in action and the observed do not, would seem to be an artificiality for which no support could be adduced.

While the action axiom is empirical and self-evident, it is, in a sense, also a priori.¹⁹ That man acts is logically prior to any concrete manifestation of action. In fact, one must have a concept of action before one can even recognize action in the so-called real world. The action axiom is derived from absolutely certain inner experience but is a priori to historical phenomena. History, as a complex of human behavior, is analyzed and interpreted by use of praxeological theorems which are, in turn, derived from relatively simple experience.

Praxeology concerns the form of action qua action. Just because it is not about this specific action or that specific action does not mean that it concerns itself only with words. The category of action is about every action that has and will take place emptied of its specific means-ends content. As such, it is no less about “reality” than any generally recognized empirical statement. All statements about the world involve some degree of abstraction, so it is not the abstraction of praxeological deductions which is at issue. What may be of concern is that they are incapable of falsification. In principle the statement “man acts” cannot be falsified since we cannot conceive of the contrary. This is not because we are simply dealing with an arbitrary stipulated definition of “man” as an acting being. Rather, it is because our acquaintance with empirical man as acting is both so intimate and necessary that a purely reacting being would not be human in the only sense we can conceive. The concepts of purposeful behavior and man are linked so tightly not because of arbitrary definition, but because they are necessarily linked in empirical reality. Our language reflects something real, yet necessary.

Praxeology as applied to history (broadly viewed as to include current history) does not depend merely on deductions from the action axiom. It requires subsidiary assumptions derived empirically in order to delimit the scope of a praxeological system.²⁰ For example, we do not want to develop monetary theory in a world without money. Now, the subsidiary empirical assumptions are not self-evident or necessarily true like the action axiom. These assumptions could conceivably be otherwise, although they may be virtually certain (e.g., the existence of indirect exchange). Insofar as they are uncertain, so too is the applicability of the praxeological statements we can make using them.

To increase the quantitative definiteness of relationships in applied praxeology (economic history), we require increasing specificity of the subsidiary assumptions: These assumptions must become both more numerous and more precise. This, of course, results in conclusions which are no longer apodictically certain. In our terminology, we refer to applied praxeological theory as hypotheses (to indicate their tentative nature). Hence, while economic theory is immutable and necessary, economic hypotheses are changeable and could be otherwise. The view that economic theory is a body of tentative statements about the world (subject to refutation) is implicitly the position that knowledge of social reality is confined solely to historical knowledge.

VII.

THE ROLE OF ECONOMETRICS

While it might appear as if econometrics has no role in the advancement of economic theory (defined as deductions from the action axiom), this is not quite accurate (although it may serve as a tolerable first approximation of the truth). Statistical regularities can be the starting point for a purely theoretical investigation, insofar as they raise questions to which the praxeologist addresses himself. But the connection here is more suggestive than logical.

The central role of econometrics is in the application of economic theory to the complex phenomena of history (current or past). There are two questions on which econometric work can shed light:

• 1 To what extent is a given (historical) instance of human behavior explicable by reference to purposeful activity, i.e., how much does a praxeological hypothesis explain?
• 2 What is the magnitude of the effect of x on the whole complex phenomenon, y, at some specific point in time?

With regard to the first question, it is important to understand that while man necessarily acts, it does not follow that he always acts, i.e., that he never engages in automatic response to stimuli or some other kind of nonpurposive behavior. To what extent is a given historical phenomenon the result of some blind emotion aiming at nothing? The answer to this cannot be given a priori.²¹

On the second question, it is important to keep in mind that praxeological reasoning per se cannot reveal quantitative relations (or even qualitative ones, when many conflicting forces are operative) in economic history. For this, statistical investigations are our only recourse. However, it is important not to interpret econometrically derived relations as great constants applicable to all situations at all times. These relations are not theoretical but merely historical. To extrapolate the latter to the former requires an inductive leap that we are not prepared to take.

In answering both of these questions, econometric evidence cannot, of course, give us the same certainty as praxeological reasoning. Answers in economic history must always be uncertain. Nevertheless, this is not the uncertainty of economic theory; rather, it is the uncertainty inherent in the application of a structure (involving the form of action) upon historico-temporal actions with specific content. The application of theory to history is not an exercise in deduction; it necessitates the use of judgment or understanding (verstehen) in defining the relevant variables and the appropriate means of measuring them.

A caveat is, however, in order. Econometrics ought to be only one tool in the apprehension of historical phenomena. Clearly, not all issues of interest are quantifiable. If we try to explain complex phenomena only by reference to quantifiable variables, then we are likely to be throwing away some information that we do, indeed, have. Another danger is that we shall begin to identify reality with statistical data when, in fact, it is just one aspect of reality, a particular transformation of more elementary experience. There is no reason whatever why a specific way of viewing history ought to be identified with history itself or, what is worse, with the whole of social reality.

VIII.

CONCLUSIONS AND UNRESOLVED QUESTIONS

The purpose of this paper is primarily to present a critical analysis of “positive economics” and only secondarily to examine the praxeologic alternative.It is in the latter area that a great deal of work needs to be done.At this point, however, a number of concluding observations might be made:

• 1 A thoroughgoing positivist approach to economics cannot be consistently pursued. The positivist framework creates certain problems that are insoluble from within that framework.
• 2 Although praxeology is concerned with action qua action, i.e., a historical and emptied of specific means-ends content, it is still about reality. The form of action is no less real than any of the other abstractions necessary in making generally recognized empirical judgments.
• 3 A crucial problem in praxeology is the epistemological nature of applied praxeology (economic history). How is the transition from theoretical constructs to empirical counterparts to be made? Verstehen is too vague an answer.
• 4 Does a praxeologist do economic history differently from a positive economist? If so, in what way?

In discussing some of the more philosophical issues of economics, it has been our intention to show that the day-to-day issues of explanation, hypothesizing, and testing do not go on in a philosophical vacuum. We do not have a choice as to whether we shall make methodological decisions. Our choice, rather, is whether we shall make them explicitly, examining the various implications and subtleties of meaning, or whether we shall make them implicitly, blind to everything but technique.

REFERENCES