W. Edwards Deming (1900—1993)
Nancy R. Mann
William Edwards Deming was born in Sioux City, Iowa, on the 14th of October, 1900. Although he has been honored throughout the world as a “quality-management guru,” he insisted upon being known as a “Consultant in Statistical Studies.” His path to the eminence that he attained as a statistician was circuitous and full of serendipity.
After Ed Deming’s graduation from the University of Wyoming in 1921 as an engineer, he remained there another year to study mathematics. It was during that time that, as he once told me, he received a letter from the Colorado School of Mines. The letter informed him that he was known to be a good flute player and that the professor of physics wanted to have a band and therefore would like him to come to teach. He accepted the invitation and, after a year, decided to get a master’s degree in mathematics and physics from the University of Colorado. Just before he completed his degree, one of his professors who had studied at Yale with Willard Gibbs, a famous mathematician and physicist, recommended him to his alma mater. Yale subsequently offered him free tuition and a job as a part-time instructor, both of which were eagerly accepted.
Upon finishing the requirements for his Ph.D. at Yale in 1928, Ed Deming began his career in government as a mathematical physicist at the Fixed Nitrogen Research Laboratory of the U.S. Department of Agriculture (USDA), and he remained in that position until 1939. His 38 publications during the period had to do principally with the physical properties of matter, but there were several that reflected his interest in statistical methodology. I once asked him why he, a mathematical physicist, became a statistician.
His answer was quite involved: “Courses in engineering and surveying led me to the theory of errors, and in studying physics and mathematics, I learned a lot of probability. Kinetic theory of gases is a theory of probability. So are thermodynamics and astronomy. And so is geodesy, involving measurement of the earth’s surface for the purpose of figuring the curvature or other characteristics of the earth. It makes use of least squares. And I had very good teachers in least squares.
“When people had problems with experimental data, I just worked on them and found myself able to make a contribution, of thought anyway. And I suppose that’s the way I eased into it.”
Analysis of results of experimental work in bacteriology and chemistry gave him a chance to learn about the statistical adjustment of data. There were three papers on “The Application of Least Squares,” published in the Philosophical Magazine. In his book, Statistical Adjustment of Data, published in 1943, he brought together, in readily usable form, the substance of these papers and of the earlier literature and his own studies on the subject. This text is still frequently consulted for guidance on the application of the method of least squares in various situations.
From 1930 through 1946, Ed Deming was a special lecturer on mathematics and statistics in the Graduate School of the National Bureau of Standards. His courses later inspired many lectures and articles by his students. These paved the way for the establishment in 1947 of the Statistical Engineering Laboratory within the Bureau of Standards. During an overlapping period that extended from 1933 through 1953, he was head of the Department of Mathematics and Statistics of the Graduate School of the USDA and made major contributions to the mathematical and statistical education of a whole generation. In 1936, he went to London to study the theory of statistics with Ronald Fisher at University College, the University of London.
While at University College, Ed Deming met and attended lectures by Jerzy Neyman, who had been Head of the Biometrics Laboratory of the Neeki Institute in Warsaw, Poland. Neyman read, at a meeting of the Royal Statistical Society, a revolutionary paper: “On the Two Different Aspects of the Representative Method: The Method of Stratified Sampling and the Method of Purposive Selection.” As a result of the lectures and particularly this paper, which marked the beginning of a new era in sampling, arrangements were made for Neyman to visit the USDA Graduate School in 1937 and lecture there.
Ed Deming took pains to ensure that Neyman’s lectures in Washington were well attended by U.S. Government statisticians, and he worked an entire year to produce the book Lectures and Conferences on Mathematical Statistics. The lectures and the book together had a tremendous impact on sampling theory.
The staff of the Bureau of the Census was already planning in the late 1930s for the 1940 Population Census. Users of census data have always wanted more information than can possibly be provided with a normal budget. Many of them were willing to accept sample results but some of the old timers at the Bureau were opposed to the idea of sampling. “Sampling was abhorred,” Ed Deming told me, “because the census had always been complete. It couldn’t be anything other than complete. But sampling was in the air.”
The final decision rested with Secretary of Commerce Harry Hopkins. After listening to the arguments pro and con, Hopkins decided that sampling procedure would be used in the 1940 population census. “Well,” Ed told me, “one day in 1939 the telephone rang, and it was Dr. Philip Hauser, the Assistant Director of the Census Bureau, wanting to talk with me about a job. I said ‘Right away!’ and joined the Bureau of the Census as Head Mathematician and Advisor in Sampling.”
After leaving the Census Bureau in 1946, Ed Deming began his practice as a Consultant in Statistical Studies from an office in the basement of his home in Washington, DC. For the remainder of his life, he conducted his consulting from this office, aided for many years before her death in 1986 by his wife Lola, a distinguished mathematician in her own right. For almost 40 years he was also assisted by his extraordinary secretary, consultant, and confidant Cecelia Kilian, known to hundreds of people throughout the world as “Ceil.”
At the same time that he began his consulting practice, Ed Deming joined the Graduate School of Business Administration at New York University as a full professor. Before he “retired” from NYU in 1975 to become Professor Emeritus, he regularly taught two courses in survey sampling and one in quality control; and he served as advisor to about 100 students who earned their master’s and doctoral degrees. The fact is that until a few months before his death, Ed Deming continued to teach at NYU every Monday during the academic year and to direct the studies of graduate students. I asked him on one occasion if NYU didn’t have some sort of policy concerning retirement of academic and other personnel at age 65 or 70. His response was, “Well, if they did have, they didn’t tell me about it.” He also taught Monday mornings during the last few years of his life as a “Distinguished Lecturer” at Columbia University, where a Deming Center has been established.
Ed Deming’s entrance into the world of quality improvement was inspired by the 1931 book Economic Control of Quality of Manufactured Product, written by his friend and mentor Walter Shewhart, the father of statistical process control. In 1938, he arranged for Shewhart to deliver a series of four lectures entitled “Statistical Method from the Viewpoint of Quality Control” at the USDA Graduate School. These lectures were published by the Graduate School in 1939 “with the editorial assistance of W. Edwards Deming.”
The crusade that Ed Deming subsequently undertook for the improvement of quality resulted, as we know, in the economic renaissance of Japan and eventually in his own worldwide prominence as a “prophet of quality” and philosopher of management.
Ed Deming’s extensive contributions to statistical thinking are too voluminous to suit the present purpose. It suffices to say that, throughout his life, he championed the belief that statistical theory shows how mathematics, judgment, and substantive knowledge work together to the best advantage. Thus he himself was a master as logician and architect of statistical studies.
Ed Deming died quietly in his sleep on December 20, 1993, surrounded by family.