PRODUCTION/OPERATIONS MANAGEMENT

Michael J. Showalter, Feature Editor, Florida State University

Making Operations Management a Science

by John G. Wacker, Iowa State University

This article proposes that operations management become scientific in developing the field through the use of theory. Although that statement seems fairly innocuous and non-threatening, its implementation requires an extreme amount of discipline not yet exhibited in the operations management's literature. In order to be scientific and theoretical, operations management must avoid non-scientific methods. This short article outlines some problem areas that concern many academics in operations management. The position taken is derived from the philosophy of scienceþthe core academic area which studies theory and science. Naturally, due to the article's brevity, it cannot fully develop the logical arguments from the philosophy of science used to support the position.

The two non-scientific pure methods referred to in this article are called: mathematics methods and statistics methods. The first method is the mathematics model.

Is Mathematics a Science?

Although many of the top journals in the operations management field carry the word science in their titles (MANAGEMENT SCIENCE, DECISION SCIENCES, etc.), are the articles that appear in them really scientific? Many operations management academics, if asked this question, would immediately say, ``Yes, they are unquestionably scientific.'' The philosophy of science does not readily accept mathematics as a science. According to the philosophy of science, one requirement for an academic field to be scientific is that results should be applicable to the corporeal world. For many of the articles in these journals, it takes a stretch of imagination to see how the article could be applied to a real world situation. This lack of application has concerned many operations management academics when discussing what they call the þquant typeþ of research. These academics are expressing exactly the same concern that philosophers of science have been expressing for some time. For mathematical articles to be scientific, many operations management academics believe these articles must have some actual immediate value to the real world. They conclude that these articles are not scientific since they believe these articles have no real world application. (Some operations management cynics state that the only value in these articles is for future publications not applications.) Currently, in the philosophy of science, the most accepted view is that mathematics is a science since it is a basic tool used for developing theory that will be used in empirical investigation. Although real world application requirement is an interesting issue, the issue cannot be resolved by the philosophy of science since its resolution depends on the degree of real world application required for a field to be a science.

Can Statistics Directly Develop Theory?

Currently, in operations management there is an increased emphasis to publish empirical articles. From a philosophy of science perspective, many of these articles are not theoretical since they violate many of the basic tenets of theory. These articles use what will be called the statistics-theory method for developing theory which is derived from behavioralist psychology: (1) The articles develop an interesting issue. (2) They define the variables from statistical constructs (using Cronbach's alpha). (3) They use numerous statistical methods to find statistically significant results. (4) They then explain the results using simplistic logic while not explaining all the other possible relationships. In the philosophy of science literature, this method leads to what is called the ``patchwork quilt fallacy.'' The underlying reason for the fallacy comes from two purposes of theory, prediction and explanation. Although the statistics can ``backcast'' prediction, they do not completely explain the relationships. Therefore, as each new article is published, a new explanation adds another ``patch'' to the quilt of the theory. Under these conditions, the theory never can be fully developed since there is no underlying agreement on what the theory is. Because of this methodology, behavioralist psychology is known as an incomplete (pseudo) science [Gale, 1976]. The future of operations management theory and operations management as a science depends upon avoiding the use of statistics-theory methods for theory development. (There are other reasons for avoiding the statistics-theory methods [see Hunt, 1991].)

Although statistics must be used to confirm, test, and give new directions to theory, there are many severe restrictions placed upon proper use of statistical estimation. These restrictions limit how statistics are used. Due to the brevity of the article, a complete list of all the theoretical problems in empirical research cannot be presented here. Only four will be addressed here: definitions, models, statistics, and explanations.

The first major problem facing empirical research is the definition of concepts. Without clearly defined concepts, any empirical research cannot show meaningful relationships between concepts. To define the concepts, conceptual definitions must be developed: by using denotative terms, by clarifying terms, and by giving measurement guidance [Lachenmeyer, 1971]. The denotative requirement means that the concept must eliminate as much of the connotative aspects (terms that carry emotional appeal) as possible of the definition. The clarifying requirement means that the definition must state what it is and, more importantly, what it is not. The measurement requirement means that the definition should help limit the choice of measures. Many of operations management's "programs" (i.e., TQM, JIT, World Class Manufacturing, etc.) are not conceptually defined since they include only what is "good" (connotative violation), fail to specifically exclude other concepts (clarifying violation) , and do not give any guidance to their measurement (measurement violation).

The second major problem facing empirical research is that of relationship building. This problem is best exemplified by the failure to develop the complete mathematical model before performing statistical estimation. For empirical estimation to be theoretically important, it must incorporate all the relevant relationships from the literature into a complete mathematical model before the statistical estimate is made. The mathematical models that are developed from the literature give the variables' expected signs BEFORE estimation. When the empirical tests are performed, the estimated signs are compared to expected signs developed in the model. As the theory and model are extended, the theory develops and the field systematically develops. From an operations management theory perspective, the failure to include all the relevant literature into the mathematical models causes the estimate to stand alone and neither be theoretical nor scientific. This failure to create the a priori mathematical model causes the statistical estimates to be internally inconsistent and the results become theoretically meaningless [Bollen, 1989]. To be sure, the development of the mathematical model is extremely tedious especially if the model is large. For that reason, other fields (such as marketing) present complete mathematical models for future empirical testing (Moorthy, 1993).

The third major problem is the use of statistics. After the relationships are built, statistical tests are used to determine if the model adequately reflects the real world. If the model is carefully built using specific types of relationships, then there is one best statistical method to test those relationships. Many times, researchers use many different types of statistical tests (such as non-linear or non-parametric methods) and the researcher picks the one that is most statistically significant to confirm their beliefs or to find a ``new theory (?)''. These statistical tests are neither theoretical nor scientific since they violate the very basic tenet of refutability needed for theory development. In short, these types of studies are non-scientific and neither build nor develop theory.

The fourth major problem is the explanation of the results. Many times empirical researchers find unexpected results during empirical estimation. Frequently, the researcher explains the results and infers that the explanation has been tested. In the philosophy of science literature, this explanation is ``calling in the conventionalist stratagem.'' Basically, the explanation was not part of the a priori mathematical model and, therefore, there was no a priori reason to believe these results would occur before the theoretical model was tested. Consequently, the post hoc explanations have not been tested and are not a new theory until the theoretical model is logically developed and then that model is tested.

In conclusion, there are two major approaches to theory development: the mathematics approach and the statistics approach. For many years (probably from the 1950's), operations management's major focus was on the mathematics approach. By the late 1970's, many operations management academics considered the mathematics approach not close enough to real world application to be scientific. In recent years, operations management is becoming empirical and it must follow good scientific procedures if it is to develop into a science: It must clearly define concepts in denotative terms for measurement. It tediously must build logical/mathematical models for empirical testing. It must properly use statistics for the results to be meaningful. It must carefully develop logical arguments for adequate explanation of results. Currently, many operations management empirical studies are not scientific since they do not strictly follow scientific procedures. If operations management does not follow scientific procedures, then it is pseudo-scientific and cannot develop into a legitimate field of scientific discovery.

This very brief overview may be extended at the 1994 National Decision Science Institute Meeting in Hawaii.

References

Gale, George. Theory of Science: An Introduction to the History, Logic, and Philosophy of Science. McGraw-Hill: New York. 1976.

Hunt, Shelby D. Modern Marketing Theory: Critical Issues in the Philosophy of Marketing Science. Southwestern Publishing Co.: Cincinnati, OH. 1991.

Lachenmeyer, C.W. The Language of Sociology. Columbia University Press. 1971.

Moorthy, K. Sridhar. Theoretical Modeling in Marketing. Journal of Marketing Research, Vol. 57, April, 1993. pp. 92-106.

JOHN G. WACKER, Ph.D. is a Professor of Production and Operations Management at Iowa State University. He is a frequent international scholar and has taught in the Czech Republics, Holland, Hong Kong, Italy, People's Republic of China, Slovakia, Spain, and Taiwan.