Milton Friedman: The Methodology of Positive Economics 
"The Methodology of Positive Economics, Milton Friedman, Essays in Positive Economics (Chicago: Univ of Chicago Press.) Introduction In his admirable book on the scope and method of Political Economy John Neville Keynes distinguishes among "a positive science ... a body of systematized knowledge about what it is, a policy or regulatory science ... a body of systematized knowledge discussing criteria of what should be ..., an art .. ., a set of rules to achieve a particular purpose, "says that" confusion between them is common and has been the source of many intentional mistakes "and urges the importance of" recognizing a distinct positive science of political economy "[1].
This document is primarily concerned with some methodological problems that arise in the construction of the" distinct positive science "Keynes called for - in particular the problem of how to decide if a suggested hypothesis or theory should be tentatively accepted as part of the body of systematic knowledge about what it is. "But the confusion Keynes laments has been so common and so much of an obstacle to the recognition that the economy may be, and in part, a positive science that looks good in the preface to the main body of the document with some observations on the relationship between economics normative and positive.
I. The relationship between positive and normative economics confusion between positive and normative economics, to some extent inevitable. The issue of economics is considered by almost everyone as vitally important to himself and within range of their own experience and expertise, but that is the source of continuous and extensive controversy and the occasion of the common law. Self-proclaimed "experts" speak with many voices and almost all can be regarded as disinterested in any case, on issues are so important, "expert" opinion could not be accepted solely on faith even if the "experts" were nearly unanimous and clearly disinterested. [2]
The findings of positive economics seems to be, and are of immediate interest to policy issues important to the questions of what to do and how any proposed target can be achieved. Laymen and experts alike are inevitably tempted to shape positive conclusions to fit regulatory entrenched prejudices and reject positive conclusions if their normative implications - or what it says are its normative implications - are unacceptable.
Positive economics is in principle independent of any ethical position or normative judgments. As Keynes said, is "what is," not "What should be." Its mission is to provide a system of generalizations that can be used to make correct predictions about the consequences of any change in circumstances. Their performance should be judged on accuracy. Scope, and conformity with the experience of the performance predictions. In short, positive economics is, or may be, an "objective" science, in precisely the same sense that any of the physical sciences. Of course, the fact that economics deals with the interrelations of human beings, and that the investigator is the same part of the subject under investigation in a more intimate than in the physical sciences, particular difficulties in achieving objectivity, while providing social scientist with a kind of data available for the physical scientist. But neither the one nor the other is, in my opinion, a fundamental distinction between the two groups of sciences. [3]
Normative economics and the art of economics, however, can not be independent of positive economics. Any policy conclusion necessarily based on a prediction about the consequences of doing one thing and not another, a prediction that should be based - explicitly or implicitly - in positive economics. There is, of course, a one to one relationship between the conclusions policies and the findings of positive economics, if any, there would be separate normative science. Two people can agree on the consequences of a particular piece of legislation. One may consider desirable in the balance and to promote legislation, and the other as undesirable and thus oppose the legislation.
I dare the sentence, however, that today in the Western world, especially in the United States, differences about economic policy among disinterested citizens derive predominantly from different predictions about the economic consequences of action - differences that, in principle, can be eliminated by the progress of positive economics - rather than fundamental differences in basic values, differences about which men ultimately can only fight. An obvious example and without significance, no minimum wage legislation. Beneath the tangle of arguments offered for and against such legislation is not a basic consensus on the goal of achieving a "living wage" for everyone to use the ambiguous phrase so common in these debates. The difference of opinion is largely grounded in an implicit or explicit differences in predictions about the effectiveness of this medium in particular in promoting that agreed in the end. The Proponents believe (predict) that legal minimum wages reduce poverty by increasing the wages of those receiving less than minimum wage, and some receive more than minimum wage without any counterbalancing increase in the number of fully unemployed or employed less advantageously than would otherwise be. Opponents believe (predict) that poverty legal minimum wage increases, increasing the number of people who are unemployed or working less advantageous and more than offset any positive effect on wages of employees who remain. Agreement on the economic consequences of the legislation could not produce an agreement all about convenience, may still be differences on political or social consequences, but, given agreement on objectives, it would certainly go a long way toward producing consensus.
Closely related to differences in the basis of a positive test divergent views about the proper role and place of trade unions and the desirability of direct price and wage controls and tariffs. different predictions about the importance of so-called "economies of scale," has greatly divergent views on the desirability or need for detailed government regulation of industry and even of socialism rather than private enterprise. And this list could be extended indefinitely. [4] Of course, my view that the main differences on economic policy in the Western world are of this type is itself a "positive" statement is accepted or rejected on the basis of empirical evidence.
If this statement is valid, means that a consensus on "corrrect" economic policy depends much less on the progress of adequate normative economics in the progress of an economy producing findings that are positive, and deserve to be widely accepted. It also means that an important reason to distinguish sharply positive economics normative economics is precisely the contribution it can make an agreement on policy.
II. Positive Economics The ultimate goal of a positive science is the development of theory "or" hypothesis "that occurs (ie, no truistic) valid and meaningful predictions about phenomena not yet observed. This theory is generally a complex mixture of two elements. It is partly a "language" to promote "systematic and organized methods of reasoning." [5] In part, it is a body of substantive hypotheses designed to abstract the essential features of complex reality.
Viewed as a language, the theory has no substantive content, is a set of tautologies. Its function is to serve as a filing system for organizing empirical material and facilitating the understanding of it and the criteria by which judges are appropriate for a file system. If the categories clearly and precisely defined? Are they comprehensive? Do we know where to file each individual item, or there is ambiguity? Is the system of headings and subheadings so designed that it can quickly find an item you want, or do we hunt for a place to another? If you want to consider items presented together? Does the filing system avoid the development of cross-references?
The answers to these questions depend partly on logical, partly for reasons of fact. The canons of formal logic can only show whether a particular language is complete and consistent, that is, if the propositions in the language are "right" or "wrong." Facts alone can show whether the categories of "analytical file system" have a meaningful empirical counterpart, that is, if they are useful in the analysis of a class of specific problems. [6] The simple example of the "supply" and "demand" illustrates this point and the previous list analog questions. Viewed as elements of the language of economic theory, these are the two major categories in which the factors affecting the relative prices of products or factors of production are classified. The usefulness of the dichotomy depends on the "empirical generalization that an enumeration of the forces affecting demand in any problem and the forces affecting supply two lists will contain some common elements." [7]
However, this generalization applies to markets as the final market for a commodity. In this market there is a clear-cut distinction between economic units that can be considered as demanding the product and may be regarded as providing the same. Rarely is there much doubt whether a particular factor should be classified as affecting the supply, on the one hand, or demand, on the other, and there is rarely much need to consider cross-effects (cross references) between the two categories. In these cases, the simple and obvious step, including the submission of the relevant factors under the headings of "supply" and "demand" effects a great simplification of the problem and effective protection against the fallacies which would otherwise tend occur. But the generalization is not always valid. For example, does not apply to the daily fluctuations of prices in a market essentially speculative. It's a rumor of an increased excess profits tax, for example, to be considered as a factor mainly operating in the current supply of corporate stock in the stock market or the current demand for them? Similarly, almost all factors can with about as much justification be classified under "supply" under the title These concepts can still be used and can not be completely useless "demand." Are still "right" but clearly less useful than in the first instance, because they have no meaningful empirical counterpart.
Seen as a body of substantive hypotheses, theory must be judged by its predictive power for the class of phenomena that are intended to "explain." Only objective evidence can show whether it is "right" or "bad" or rather, tentatively "accepted" as valid or "rejected." As I will argue in more detail below, the only relevant test of the validity of a hypothesis is comparison of its predictions with experience. The hypothesis is rejected if its predictions are contradicted ("frequently" or more often than predictions of an alternative hypothesis) is accepted if its predictions are not contradictory great confidence is attached to if it has survived many opportunities for contradiction. factual evidence can not "prove" a hypothesis, but can only leave to challenge, which is what usually mean when we say, somewhat inaccurately, that the hypothesis was "confirmed" by experience.
To avoid confusion, perhaps it should be noted explicitly that the "predictions" by which the validity of a hypothesis test is to be about phenomena that have not yet occurred, ie need not be the forecasts of future events, but may be about phenomena that have occurred, but the comments which have not yet been done or not known to the person making the prediction. For example, a hypothesis may imply that such and must have happened in 1906, taking into account certain known circumstances. If a search of the records reveal that such and such was the case, the prediction is confirmed, and if it shows that such and such did not happen, the prediction is contradicted.
The validity of a hypothesis in this sense is not itself a sufficient criterion for choosing among alternative hypotheses. Observed facts are necessarily finite in number, the possible hypotheses, infinite. If there is a hypothesis that is consistent with the available evidence, there is always an infinite number are. [8]
For example, suppose a specific excise duty on a product causes an increase in price equal to the amount of tax. This is consistent with the conditions of competition, a stable demand curve and a horizontal supply curve and stable. But it is also consistent with the conditions of competition and a supply curve positively or negatively inclined to the change required to compensate for the demand curve or supply curve, with monopolistic conditions, constant marginal costs, and the curve stable demand, the particular form required to produce this result, and so on indefinitely. further evidence that the hypothesis has to be consistent can rule out some of these possibilities can never be reduced to only one possibility could be consistent with the evidence finite. The choice among alternative hypotheses equally consistent with the available evidence is somewhat arbitrary, although there is general agreement that relevant considerations are suggested by the criteria of "simplicity" and "fertility", the same concepts that defy completely objective specification. One theory is "simpler", less the initial knowledge needed to make a prediction in a given field of phenomena, but more "fruitful" the more accurate the prediction is, the larger the area in which yields predictions of the theory, and most of the additional lines of research that suggests. logical integrity and consistency are important, but play a subsidiary role, its role is to ensure that the hypothesis says that what is intended and does the same for all users - who play the same role here as checks for arithmetic accuracy to statistical calculations.
Unfortunately, rarely can test particular predictions in the social sciences by experiments explicitly designed to eliminate what are considered the most important disturbing influences. In general, we must rely on evidence from the "experiments" that happen to occur. The inability to carry out the so-called "controlled experiments" not, in my opinion, reflect a basic difference between the social and physical sciences, both because it is not proper social sciences - astronomy witness - and because the distinction between a controlled experiment and uncontrolled experience is the best of grade. No experiment can be fully controlled, and every experience is partly controlled, in the sense that some disturbing influences are relatively constant in the course of it.
The evidence from the experience is abundant and often as conclusive as that from artificial experiments, so the inability to conduct experiments is not a fundamental obstacle to test hypotheses the success of their predictions. But such evidence is much more difficult to interpret. It is often complex and always indirect and incomplete. Its collection is often difficult, and their interpretation generally requires a subtle analysis and involved chains of reasoning, which rarely lead to true belief. The negative to the economy of the dramatic and the evidence of the "crucial" experiment is an obstacle to adequate testing of hypotheses, but this is much less important than the difficulty that stands in the way of achieving a reasonably rapid and broad consensus on the conclusions justified by the evidence. This makes the weeding-out hypothesis unsuccessfully slow and difficult. They are seldom downed for ever and ever emerging again.
There are, of course, considerable variation in these aspects. Sometimes the experience is a test that is so direct, dramatic and compelling as any other that may be provided by controlled experiments. Perhaps the most important example is obviously evidence of inflation on the assumption that a substantial increase in the amount of money in a relatively short time accompanied by a substantial increase in prices. Here the evidence is dramatic, and the chain of reasoning required to interpret that is relatively short. However, despite numerous cases of substantial increases in prices, its essence, one to one correspondence with the substantial increase in the amount of money, and a variety of other circumstances that may seem relevant, every new experience of inflation gives birth to strong allegations and not only by the public, that the increase in the money stock is either a side effect of increased prices caused by factors other purely incidental or accidental and unnecessary price increases.
One effect of the difficulty of confirmatory testing economic hypotheses has been to encourage a retreat in the purely formal or tautological analysis. [9] As noted, the tautologies have an important place in the economy and other sciences as a specialized language or "File system analysis." Beyond this, formal logic and mathematics are tautologies, are an essential aid to verify the correctness of the reasoning to discover the implications of the hypothesis, and whether supposedly different hypotheses can not actually be equivalent or where the differences lie.
But economic theory must be more than a structure of tautologies if you will be able to predict and not merely describe the consequences of the action;. If you want to be something different from disguised mathematics [10] And their usefulness ultimately depends tautologies, as noted above, on the acceptability of the substantive hypotheses suggested by the different categories that are organized refractory empirical phenomena.
A critical effect of the difficulty of testing economic hypotheses by their predictions is encouraging a lack of understanding of the role of empirical evidence on theoretical work. The empirical evidence is vital in two different but closely related phases: the construction of hypotheses and test their validity. complete and comprehensive test of the phenomena to be generalized or "explained" by a hypothesis, besides its obvious value to suggest new hypotheses, it is necessary to ensure that a hypothesis explains what is proposed to explain - that the consequences of such phenomena are not contradicted in advance by the experience that has already been observed. [11]
Since the hypothesis is consistent with the evidence available, the more evidence it is inferred new facts capable of being observed but not previously known to check these facts and additional inferred from empirical evidence. For this test to be relevant, the facts must be deducted from the class of phenomena the hypothesis is designed to explain, and must be sufficiently well defined to observation may show an error.
The two stages of the construction of hypotheses and evidence of their validity are related in two different senses. First, the facts that come in each stage are partly an accident of data collection and knowledge of the researcher in particular. The facts are evidence of the consequences of a hypothesis could equally well have been one of the raw materials used for construction, and vice versa. Second, the process does not start from zero, the "initial phase" itself always involves the comparison of the consequences of the previous set of hypotheses through observation, the contradiction these consequences is the stimulus to the construction of new hypotheses or revising old ones. So the two methodologically different stages are always proceed jointly.
Misunderstandings about this seemingly simple process centers on the phrase "class of phenomena the hypothesis is designed to explain." The difficulty in the social sciences to obtain new evidence for this kind of phenomena and of judging its conformity with the implications of the hypothesis makes the temptation to assume that other, more readily available, the evidence is equally relevant to the validity of the hypothesis - to suppose that the assumptions are not only "implications" but also "assumptions" and that the conformity of these "supposed" to "reality" is proof of the validity of different or additional hypothesis to the test for the consequences. This common view is fundamentally flawed and productive of much harm. Far from providing an easier means for sifting valid from invalid hypotheses, only confuses the issue, promotes confusion about the meaning of the empirical evidence of economic theory, a deviation from a great intellectual effort devoted to the development of positive economics, and impedes the attainment of consensus on tentative hypotheses in positive economics.
To the extent that a theory can be said to have "assumptions" at all, and to the extent that his "realism" can be judged independently of the validity of the predictions, the relationship between the meaning of a theory and "realism" of its "assumptions" is almost the opposite of the view suggested by the critics. Truly important and significant hypotheses will be found that "assumptions" that are wildly inaccurate descriptive representations of reality, and in general, the more significant the theory, the more unrealistic the assumptions (in this sense.) [12] The reason is simple. One hypothesis is important if it "explains" much by little, that is, if it abstracts the common and essential elements of the mass of complex and detailed circumstances surrounding the phenomenon that explains and allows valid predictions based on them alone. To be important, therefore, a hypothesis must be descriptively false in its assumptions, but takes into account and is responsible for any of the circumstances of the case for many others, since its very success shows that are irrelevant to the phenomena to explain.
To put this point less paradoxically, the relevant question to ask about the "hypothesis" theory is not whether they are descriptively "realistic" because they never are, but if they are sufficiently close good for the purpose at hand. And this question can be answered only by seeing whether the theory works, which means that if yields sufficiently accurate predictions. The two supposedly independent tests, thus reducing to a test.
The theory of monopolistic and imperfect competition is an example of the neglect of the economic theory of these propositions. The development of this analysis was explicitly motivated, and its wide acceptance and approval largely explained by the belief that the assumptions of "perfect competition" or "perfect monopoly" said to the core of neoclassical economic theory is an image false reality. Y This belief has been based almost exclusively on the directly perceived descriptive inaccuracy of the assumptions rather than any recognized contradiction of predictions derived from neoclassical economic theory. The long debate on marginal analysis in the American Economic Review a few years ago is an even lighter, but much less important, example. Articles on both sides of the controversy largely neglect what seems to me clearly the main issue - the conformity with the experience of the consequences of marginal analysis - and concentrate on the question irrelevant whether businessmen do or not do, in fact reach their decisions by consulting schedules, or curves, or multivariable functions showing marginal cost and marginal revenue. [13]
Perhaps these two examples, and many others suggest easily serve to justify a broader discussion of the methodological principles that would otherwise seem appropriate.
Not sure there, but if it appears that there is interest, I can send more of this essay later. In particular, the "Can a hypothesis to be tested by the realism of its assumptions" section that follows. Bibliography
* I incorporated body of this article without mentioning my most special report "Review" in a survey of contemporary economics, vol. II (BF Haley, ed.) (Chicago: Richard D. Irwin, Inc., 1952), pp 455-57.
I am indebted to Dorothy S. Brady, Arthur F. Burns and George J. Stigler for helpful comments and criticisms.
1. (London: Macmillan & Co., 1891), pp 34-35 and 46.
2. Social sciences or economics is not peculiar in this regard - witness the importance of personal beliefs and "home" remedies of medicine where, of course, convincing evidence "expert" opinion is missing. The current status and acceptance of the views of physicists in their fields of expertise - And all too often in other fields - is derived not by faith alone, but from the evidence of his works, the success of their predictions, and the spectacular achievements of the implementation of its results. When the economy seemed to provide such evidence of its value, in Britain in the first half of the nineteenth century, the prestige and acceptance of "scientific economics" rivaled the current status of the physical sciences.
3. The interaction between the observer and the observed process is as important a feature of the social sciences, besides its obvious parallel in the physical sciences, has a counterpart in the more subtle uncertainty principle arising from the interaction between the measurement process and the phenomena being measured. And both have a counterpart in pure logic, Gödel's theorem, the assertion of the impossibility of complete self-contained logic. It is an open question whether all three can be considered as different formulations of a more general principle.
4. A rather more complex example is the stabilization policy. Superficially, the divergent views on this issue appear to reflect differences in the goals, but I think this impression is misleading and that basically different views largely reflect judgments different about the origin of fluctuations in economic activity and the effect of cyclical action alternative. For a significant positive account that explains much of the divergence seen "The effects of a policy of full employment, economic stability: a formal analysis, below, pp 117-32. For a summary of the current state of professional views on this subject see "The problem of economic instability," a report by a subcommittee of the Committee on Public Affairs of the American Economic Association, American Economic Review, XL (September 1950), 501-38.
5. final sentence quotation from Alfred Marshall, "The Current Situation of the Economy "(1885), reprinted in Memorials of Alfred Marshall, ed. AC Pigou (London: Macmillan & Co., 1925), p. 164. See also" The Marshallian Demand Curve, "infra, pp 56 - 57, 90-91.
6. See "Lange on price flexibility and employment: a methodological critique", below, pp 282-89.
7. "Marshallian demand curve, below, p . 57.
8. The qualification is necessary because the "evidence" may be internally contradictory, so there may be no hypothesis consistent with it. See also "Lange on price flexibility and employment, infra, pp 82 -83 2.
9. See "Lange on price flexibility and employment, infra, passim.
10. See also Milton Friedman and LJ Savage, "The expected utility hypothesis and the measurability of utility", Journal of Political Economy, LX (December 1952), 463-74, esp. pp 465-67.
11. In recent years, some economists, in particular, a group connected with the Cowles Commission for Research in Economics at the University of Chicago, have placed great emphasis on a division of this stage of selecting a hypothesis consistent with known evidence into two sub-stages: first, the selection of a class of all admissible hypotheses case scenario (the election of a "model" in his terminology), secondly, the selection of a hypothesis of this kind (the choice of a "structure"). This subdivision may be heuristically useful in some types of work, particularly in promoting the systematic use of available statistical evidence and theory. From the methodological point of view, however, is a completely arbitrary division of the process of deciding on a particular scenario that is on par with many other subdivisions that may be suitable for one purpose or another or to fit the psychological needs of researchers in particular.
One consequence of this branch in particular has been to give rise to the "identification" problem. As noted above, if a hypothesis is consistent with the available evidence, an infinite number are. But while this is true for the class of hypotheses as a whole can not be true of the subclass obtained in the first of two steps above - the. "Model" may be that the test used to select the hypothesis end of the subclass can be compatible with the majority in a scenario it, in which case we say that the "model" to be "identified", otherwise it is said is clear from this way of describing the "no identified. "term" identification "is essentially a special case of the more general problem of selecting among alternative hypotheses equally consistent with the evidence - a problem that should be decided by an arbitrary principle as Occam's razor. Introducing two sub-stages in the selection of a hypothesis means that this problem occurs in two stages and provides for a special cast. While the class of all hypotheses is always unidentified subclass a "model" has to be , so the problem arises from the conditions that a "model" must meet to be identified. However useful the two sub-steps may in some contexts, its introduction poses the danger that the different criteria used inadvertently to make the same kind of choice between alternative hypotheses in two phases.
The general methodological approach discussed in this note Tryvge see Haavelmo, The probability approach in econometrics ", Econometrica, vol. XII (1944), Supplement, Jacob Marschak, "Economic Structure, Path, politics, and prediction", American Economic Review, XXXVII (May 1947), 81-84, and "StatisticaI Inference in Economics: An Introduction," in TC Koopmans (ed), Statistical inference in dynamic economic models (New York: John Wiley & Sons, 1950), TC Koopmans, "Statistical Estimation of Simultaneous Economic Relations," Journal of the American Statistical Association, XL (December 1945), 448-66; Gershon Cooper, "The role of economic theory models econometrics, "Journal of Agricultural Economics, XXX (February 1948), 101 to 16. In the identification problem to see Koopmans, "Identification of Problems in the construction of the econometric model", Econometrica, XVII (April 1949), 125-44; Leonid Hurwicz, "The generalization of the concept of identification", in Koopmans (ed. ), Statistical Inference
12. The opposite of the proposition no, of course, wait: unrealistic assumptions (in this sense) do not guarantee an important theory.
13. See RA Lester, "Shortcomings of marginal analysis for wage employment problems," American Economic Review, XXXVI (March 1946), 62-82; Fritz Machlup, "Marginal Analysis and Empirical Investigation," American Economic Review, XXXVI ( September, 1946), 519-54; Lester RA, "marginal, minimum wages, and labor markets", American Economic Review, XXXVII (March 1947), 135-48; Fritz Machlup, "Rejoinder to Antimarginalist" American Economic Review, XXXVII (March 1947), 148-54; GJ Stigler, "Professor Lester and Marginallsts ", American Economic Review, XXXVII (March 1947), 154-57; Oliver HM, Jr.," The marginal theory and business behavior, "American Economic Review, XXXVII (June 1947), 375 - 83. Gordon RA, "Determination of short rates in Theory and Practice", American Economic Review, XXXVIII (June 1948), 265-88.
should be noted that, along with much material that allegedly affect the validity the "hypothesis" marginal theory, Lester referred to compliance testing experience with the implications of the theory, citing the reactions of employment in Germany in the Papen plan and in the United States to changes in legislation minimum wage as examples of non-compliance. However, a brief comment Stigler is the only other papers referred to these tests. It is also noted that Machlup careful and complete exposition of the logical structure and meaning of marginal analysis is requested by the misunderstandings in this respect that the role of Lester markets and almost hide the evidence presented is relevant to the key question poses. But Machlup emphasis on logical structure, which is dangerously close to the presentation of the theory as a pure tautology, though it is evident in a number of points he is aware of this danger and anxious to avoid. Jobs Oliver and Gordon are the most extreme exclusive focus on the compliance of the conduct of business with the "hypothesis" theory.
