Categories
Harvard

E.B. Wilson Lecture USDA Graduate School 1930

UNITED STATES DEPARTMENT OF AGRICULTURE
GRADUATE SCHOOL
SPECIAL LECTURES ON ECONOMICS
DELIVERED BEFORE THE GRADUATE SCHOOL
FEBRUARY – MARCH 1930
WASHINGTON, D. C.
1930

 

Contents: The following lectures were delivered before the students of the Graduate School in February and March 1930, and are issued in this form for present and former students of the school.

 

Scientific Method in Economic Research                                                                1

by Dr. E. B. Wilson, President, Social Science Research Council.

o[p. 1]

SCIENTIFIC -METHOD IN ECONOMIC RESEARCH

By Dr. E. B. Wilson, President, Social Science Research Council.

It is a real embarrassment to me to come here announced to speak upon a topic, Economic Research, of which I have never known much, and today know less than I have at times in the past. The awkwardness is not alleviated by my appreciation of the fact that there are several in the audience who could talk about the subject better than I, and with that increased assurance and authority which comes of a wider experience and greater personal accomplishment in the field. Furthermore, the title of my lecture contains the cryptic words “Scientific Method” which slide smoothly from the tongues of many persons dealing with relatively inchoate research fields, but are apt to stick in the crop of those whose major experience is in well developed branches of science. I could, for instance, hardly imagine a group at the Bureau of Standards asking me or any other for a lecture on scientific method in physical research. Indeed, I should not be astonished to hear physicists say there was no such thing as scientific method, that there was an infinite variety of techniques, theoretical and experimental, available to the physicist from which each individual must select for a particular problem those which may be applicable to that problem, but that as to scientific method in general there was little which could be stated other than something very general with respect to the scientific attitude of mind, patient and unbiassed searching for facts, earnestness in seeking for corroboration by others wherever personal bias may have influenced the array of facts or the conclusions drawn from them, and a persistent effort to improve techniques so that facts may become more objective and precise. There would probably be some who boldly would affirm that method was for the other fellow and not for. themselves, meaning by this that each individual investigator, and especially those of any real genius, pursued his own studies without conscious dominance by any notion of scientific method until he came to the somewhat unwelcome task of writing up his work for communication to others, when perforce he must consider what method of presentation would most effectively convey his findings and their justification or validity to that particular group of fellow scientists whose good opinion he would most value.

I might state as a generalization that the interest in scientific method on the part of investigators in any scientific field varies inversely as their interest in that field of science. But this statement is much too striking to be scientific; in the first place there are always exceptions and it would have to be understood as applying on the average, and in the second place there is no way known to me to evaluate quantitatively either the interest in method or the interest in the field which would justify so precise a statement as that the one interest varied inversely as the other.

I do not, however, propose entirely to cheat you of all discussion of the topic of my lecture. I will enumerate some methods.

[p. 2]

(l) The method of definition. The choice of a suitable definition if often the touchstone of scientific advance. Scientifically definitions need not be explicit, they may be implicit; they need not be quantitative, they may be first qualitative. There are numerous illustrations which could be given from the economic field, but we may all get forward faster if we take an example from physics. Let us define force. And let us go back to Newton. In his first law of motion he stated that a body at rest remained at rest and a body in notion remained in motion in a straight line at uniform velocity unless acted on by some force. Now this cleared the way. It eliminated once for all the notion that bodies stopped of themselves – they had to be stopped or started, accelerated or decelerated by applied force. There is here a qualitative and implicit notion of force which can furnish a sound basis for the quantitative definition of force as equal (or proportional) to the rate of change of motion. As a matter of fact a single definition cannot well exist by itself, it can only be part of a system of definitions – force, mass and acceleration are tied together. At any particular time in the advance of a science some new term may seem to be defined absolutely alone and the complex of definitions to which it belongs may not be perceived because this complex lies in the general background of our accepted thinking whereas the newly defined element emerges.

From such experience as I have in teaching and in research, particularly in emergent fields, I should say that most pupils and colleagues were so thoroughly oblivious to the necessity for and significance of definitions in science as to be almost impatient of them. In this they are far from scientific. Science is a congeries of definitions. We are seeking agreement in science and one of the most effective ways of seeking it is by agreement on the meaning of our words. Take such words as “wealth”, “income”, and others which are basic in economic research. How can one expect agreement between two persons who are tacitly using different though perhaps unformulated definitions for these or similar words? Of course, I do not wish to over-emphasize the necessity for precision in definitions – there is no such thing as absolute precision apart from the construction of complete logical systems such as mathematicians have constructed for arithmetic, geometry and algebra of several types. We have at any stage of science to deal with a relative precision of definition. And it may not be amiss for me to point out that the absolute precision of the completely defined mathematical discipline exists only for that intellectual discipline and breaks down immediately when that system of thought is applied to nature, by virtue of the fact that there do not exist in nature any objects precisely corresponding to the exact intellectual concepts precisely defined.

Let us consider the question of “level”, “rolling”, or “hilly” agricultural land. It would doubtless be possible to construct mathematical definitions of level, rolling, and hilly in such form that a surveyor equipped with the proper instruments could classify land under these three categories with great scientific precision, by which we mean that specifications for the classification could be so drawn as to make certain that different surveyors equipped with similar instruments would make the same classifications with only the fewest disagreements in certain very doubtful intermediate cases. Such a precise definition with its accompanying exact classification would be costly in time and in money and thus of

[p. 3]

limited practical utility. Very little land could be so classified. On the other hand, it would be possible to set up rough and ready descriptive definitions of level, rolling, and hilly intended to be used for classifications by mere inspection which would be utilizable at comparatively slight expense in money and time. Now different observers would undoubtedly differ considerably in their allotments of parcels of an intermediate character to one or the other of the adjacent categories, and in this allotment there might be bias, so that even on the average the classifications by different observers would have different percentages of the three sorts. The reliability of the method would certainly be less, its utility, even its scientific utility, might be far greater because of its greater feasibility.

The essential element in precision is not so much absolute precision as a sound estimate of what degree of precision has been attained. It is really reliability in the sense of reproducibility or agreement which is the scientific item and if the reliability is sufficient for the purposes in hand and if the conclusions drawn from the work limit themselves to what is justified by the reliability attained, that is all we can really ask. Thus two treatises on economics neither of which explicitly defines wealth or some other term but both of which develop their analysis upon tolerably similar implicit definitions may well not come into actual conflict even though if each pushed its analysis far enough they could hardly fail to come into some conflict resoluble only through rendering the definitions more precise. Much medical work, particularly of the clinical variety, could not proceed at all if a high degree of precision were required. So to define broncho-pneumonia or dementia precox as to result in a high degree of reliability in diagnosis may actually be impossible in the present state of our technique. I do not mean that definitions could not be given but that such definitions would not be those of what should be called broncho-pneumonia or dementia precox – they would be artificial and illusory definitions. What we need at the moment is not such definitions so much as a recognition by scientific clinicians of their need for the concept of reliability, their need to have a feeling for what the true precision is; because it is only upon the basis of such recognition that one is likely fruitfully to improve the definitions. The situation may well be the same in respect to a large number of economic concepts.

(2) The method of mathematical analysis. Beginning rather clearly with Cournot, there has been a great development of mathematical analysis in economics. I do not refer to arithmetically quantitative analysis but to the type of mathematics which is logical and qualitative in the sense in which J. C. Cobb has discriminated between qualitative and quantitative (Econ. Jour. England, March 1928, p. 72). Now, as we were speaking of definitions, it may be well for me to point out that the adjective quantitative is perhaps used in a different sense by some economists from that implied here. Some might consider Pareto’s work almost wholly quantitative because of his insistence on sharply defined concepts and his use of mathematical equations and transformations, including differential equations and integrals, etc. The notion of marginal utility is in a sense quantitative, and so is the concept of the curve of indifference which Pareto was inclined to make basic. Yet this must be said: That his whole treatise goes very little way, and the whole spirit of it is not conceived in the manner to go far toward the determination of the numerical values which should be inserted in the place of the variables in his equations to arrive at a check on the analysis. Indeed, as I conceive it, he was not interested greatly in that sort of check: his interest lay in the logical analysis of the sorts of situations which could arise. In a sense that is qualitative rather than quantitative, and at any rate I want to make it clear that I used qualitative to cover what others might prefer to call quantitative.

Mathematical analysis depends on conventions. If the convention is adopted that the price of a commodity depends on the amount of that commodity one has the system of Walras, and the mathematics is relatively simple. If one allows that the price of each commodity may depend on the amounts of all commodities one has a more complicated convention leading to more complicated analysis. This latter form of analysis is surely so highly involved that one could scarcely carry it on without the mathematical method. If then commodities are really so interlocked that many important economic events can only be understood by taking into consideration the possibilities that arise through the interlocking and would not arise if there were no interlocking, there is no escape from a consideration of the complex situation and no escape from the use of a good deal of mathematical thinking with no small amount of mathematical symbolism. It is perhaps true that many serious students of economics are today being developed without any considerably training in mathematics. Maybe they can so choose their work throughout their life as to avoid problems which require a knowledge of mathematics, and maybe they will learn their mathematics in later life as W. G. Sumner did, according to his personal testimony to me.

I may say that it has been maintained that man’s essential characteristic lies in his effort to cover not space alone but duration of time (Korzybski). It may be that prices do not depend merely on quantities of commodities in existence but on those and quantities that may come into existence. In recent months we have been told that what was important in the price of a stock was what it was going to sell for. Very likely. There is certainly speculative adjustment of prices. If that complex of phenomena which arise from man’s time-binding propensities is of relatively little economic significance, it may be neglected, but if it be of considerable significance, economists will arise to treat it and they will probably be forced to use much more complicated mathematical thinking than Pareto used – there may be integral equations and integro-differential equations. It is not impossible that they cannot even state adequately the essential logical interconnections of the conceptual situation which arises in time-binding economics and cannot formulate the proper conceptual complex without appeal to a type of mathematics so advanced as to make it necessary for some mathematicians themselves to become economists. Indeed, Pareto started as an engineer, I believe a railroad engineer; Irving Fisher as a mathematician, and so more recently Roos. The mathematical method is not yet through in economics; indeed, it may be just beginning, but it is likely largely to remain the work of a small fraction of students of economics.

(3) The statistical method. Although the statistical method is highly valuable in dealing with classifications of data not yet reduced to quantitative form, presumably the chief interest of the economist in

[p. 5]

statistics is quantitative. There is of course no such thing as the statistical method. Statistics is a complex of techniques. Take, for example, the sampling technique, the sampling error, and the question of the fairness of a sample – this whole group of ideas which is of great importance in social studies and in experimental biology may be of little interest to the economist working in some such line as the theory of the business cycle. If we allow an average duration of 40 months, we have in 40 years some twelve cycles. Of what can those twelve be regarded as a sample? Of course they may be regarded as a sample of the universe defined by themselves, which does not get us ahead very far. Indeed 40 years takes us back to 1890, before Bryan, before the U. S. Steel Corporation, before the war, and before the “New Era” of 1929. In many ways the economic situation has so changed that there may be no sense in asking the question as to what universe the sample comes from. Yet how talk of sampling errors without samples or of samples without their appropriate universe? I do not wish to imply that there are not opportunities to use sampling methods in economics; there are many such, but. on the other hand we have many problems where the opportunity is lacking.

Then there is the regression technique. What a regression equation does is to give us the mean value of the dependent variable in terms of assumed values of the independent variables. A regression equation may not be solved for one of the independent variables to get the new regression equation. This is a limitation on the use of regression equations for mathematical analysis. Indeed the mathematical theory of economics assumes that the variables occurring in the equations are like those occurring in thermodynamics or other fields to which analysis is applied in that they have values, not that all the independent variables have values and the dependent variables only a mean value. This is a real difficulty in coordinating the mathematical and the statistical methods so that the latter furnishes concrete and practical illustrations of the theoretical developments of the former. It is a difficulty which must persist so long as inherent variability due to lack of complete control is present. Thus one of the chief techniques of the mathematical economics is that of counting the number of variables and the number of equations to verify that these numbers are equal and that the problem is therefore determinate, whereas statistical procedures are applied in just those cases in which there still remains indetermination, that is, where the number of equations is fewer than the number of variables and it becomes necessary to treat the dependent variable as an assemblage of values and take the mean or median or some other special value as representing the group of undetermined values.

The analysis of time series is a technique which has remained rather specially that of economic statistics. It is employed very slightly in other applications of the statistical method. Time series are treacherous, but I do not see how their treatment can be ignored. Also it is not easy to see how the treatment can be made scientifically satisfactory; in the nature of the case we may have to be content with exploratory methods. Of course we can apply the method of the periodogram and the resolution into trignometric series. Such analyses work well on observations of the magnitudes of some classes of variable stars for which there are presumptively true periodicities in the background of the observed time series. When however the method is applied to meteorological or economic phenomena somewhat bizarre results are often obtained, results which do not always,

[p. 6]

perhaps do not usually, serve as a satisfactory basis either for forecasting the future or for representing the past except for the limited portion of it which has been used in the analysis. Indeed at the present time I should say that the evidence that there are cycles does not justify one in assuming that there are periods in the sense required for sound
periodogram analysis.
One of the chief aids to the advance of scientific knowledge is the existence of artificial or natural repetition in the sequence of phenomena. The extreme regularity of the real motion of the planets and the moon enabled the ancients very considerably to unsnarl the obvious irregularities of the apparent motions and to construct calendars of a high degree of prevision, to forecast eclipses with considerable success, and in general to develop the science of astronomy. The periods involved were not all short compared with human life, but the basic regularity allowed observations to accumulate. In the laboratory we may repeat experiments as often as we please without serious danger that conditions have seriously changed, – indeed the first and basic type of control is repetition. In the economic field the situation is quite different; so far as is yet known we do not have highly regular basic periods nor do we have conditions that may be repeated without serious changes, and we are very much handicapped thereby in scientific advance. It is true that the rapidity of change in conditions varies greatly at different times. The younger generation of economists have seen tremendous apparent changes within their own relatively short lives. It may be that a prolonged period of relative stagnation might help toward the discovery of some economic laws by making it easier to ignore the trees and concentrate on the forest; such a period would however have compensating disadvantages through the temptation to a feeling of security with its inevitable resulting dogmatism.

After all, it is perhaps remarkable not that so little but that so much has been accomplished in selecting from the mixed, complex, and variable economic field some items which have been reduced to method, by definition, by mathematics and by statistics sufficiently to merit the application of the term economic science. Whether the proportion of economic instruction in our universities which can be rated as scientific is today larger than it was a generation ago may be difficult to determine – there is much more economics taught, much more that is descriptive, illusory and unscientific, also much more that is sound and scientific. In the governmental, banking and business activities of our country there is also an increase in the amount of economics in use and very likely an increase in the proportion of scientific economics. The scientific economist may look at the future with caution as to what economic events may occur, but he may look at it also with confidence that the scientific method will continue to conquer new worlds for him as for his scientific confreres in other scientific fields.

Source: United States Department of Agriculture Graduate School. Special Lectures on Economics, Delivered Before the Graduate School February-March 1930, pp. 1-6.

 

One reply on “E.B. Wilson Lecture USDA Graduate School 1930”

The polymath E. B. Wilson was a mentor of Paul Samuelson at Harvard. This lecture was held in a series that included Frank Knight and John D. Black of Harvard.

Comments are closed.