In discussing automation, it may be useful to keep in mind an old Jewish saying: For example is no proof. Often the argument about automation is carried on by citing spectacular examples—such as new control equipment, sensing devices and the like—rather than focusing where it should be focussed, on the aggregate effects of automation on the economy as a whole. In other words, even though many jobs have been eliminated in particular industries (e. g., printing or coal mining) the question remains whether the present phase of automation really poses new and wholly unprecedented problems for the economy.
In the past year there has arisen a new “school,” the Ad Hoc Committee on the Triple Revolution, which has argued three propositions: 1) That automation—or, to use Donald Michael’s phrase, Cybernation—represents a radical break from previous kinds of mechanization in its economic effects; 2) That the pace of technological change in recent years has been accelerating; and 3) That the immediate, visible consequences of the first two propositions are demonstrated by the continuing high level of unemployment which, this year for the first time since 1957, has dipped below the 5 per cent mark.
In this essay, I shall argue that the first proposition is false; that the second proposition is unprovable and, if one narrows the idea of technological change to the specific measure of an increase in the rate of productivity, probably false as well; and that unemployment is and remains a serious problem, but not because of automation.
The Ad Hoc Committee has made its claims in sweeping terms:
A new era of production has begun…Cybernation is already reorganizing to meet its own needs…As machines take over production from men, they absorb an increasing proportion of resources, while the men who are displaced become dependent on minimal and unrelated government measures—unemployment insurance, social security, welfare payments.
Donald Michael, in his book The Next Generation, predicts that “in the next ten to twenty years cybernation will disrupt the whole labor market, from executives to menials.” The basis for these conclusions is the idea of a revolution “brought about by the combination of the computer and the automated self-regulating machine.” In short, what is “new” is not simply additional mechanization but the threat that computers will take over much of the work of production throughout the economy.
The difficulty with this reasoning is that it is based on speculation, not fact. As Charles Silberman pointed out in the January 1965 Fortune, “ten years after computers started coming into use, no fully automated process exists for any major industry in the U.S. Nor is there any prospect for the immediate future” (his italics). Furthermore, when the Department of Labor recently made a study of the probable effects of automation during the coming decade, it found that there was no likelihood of “computerization” in sixteen of the country’s thirty-six largest industries. Of the remaining twenty, six would be able to use computers, but only in their offices. Only fourteen of the thirty-six industries could possibly use computers to control production itself; and of these, eight are industries like petroleum refining and electric power, which are already highly mechanized. In all, the thirty-six industries studied account for almost half of civilian employment. Thus the prospect that the entire manufacturing economy will be run by computers employing only a handful of workers remains in the realm of science fiction.
The same can be said, I’m afraid, about the view expressed in much recent writing to the effect that we live in a new era of “accelerated pace of technological change” (a theme of John Diebold’s book); for the startling thing is that not only do we lack any precise measurement of technological change, we lack even a definition of the phenomenon that is presumably being described. What do we mean by “technology”? Is it machinery and equipment? If so, how do we establish an aggregate measure of such things as computers, tools and dies, drop forges, and the like? One measure could be the monetary value of a piece of capital equipment, but a moment’s reflection would show how useless this is for measuring technological change. The cost of a large road-building bulldozer, for example, may rise because of oligopolistic market power, thus causing artificial inflation, or new equipment may be of a “capital-savings” type, which has largely been the case since 1957. Thus any effort to measure technological change by the growth in the monetary value of a nation’s stock of capital goods would be misleading.
Is the narrowing of time intervals between the discovery of a new invention and its industrial application a measure of technological change? We are told by Gerard Piel and others that such time intervals are shrinking, and the transistor is often cited as a case in point. But one often forgets the inventions which have failed, so far, to pay off: In 1946-47 it was widely predicted that cheap atomic energy would transform our energy output; but now, almost twenty years later, atomic energy is barely becoming competitive with other energy sources like coal.
The most useful view is the one taken by economists, which maintains that technological change is change that increases the efficiency of a given set of productive forces; it is, to put it crudely, a change that gives you more production for less work. But how is efficiency to be measured in the economy, or in a particular industry? The most common measure we have is the year-by-year changes in the output per man-hour. This is the measure of what economists call “productivity.”1 It is arrived at by dividing the market value of the goods and services produced during a given year (in the economy as a whole or in a particular industry) by the numbers of man-hours it took to produce them. Productivity so defined is obviously a gross measure. It in no way identifies whether increased efficiency, has been brought about by new machinery or by a better-skilled labor force, or even by a speed-up of work done on the job. Still, if we are to learn whether technological change has been vastly accelerated in recent years, this is the only consistent measurement we have.
What, then, do the productivity figures show? In the period between 1909-47, output per man-hour in the total private economy increased an average 2 per cent a year. Since the end of the war, the rise in productivity has averaged 3 per cent a year, an increase of 50 per cent. But a closer look gives us a far different picture, for the most spectacular gain in productivity—5.7 per cent from 1947 to 1959—was not in industry, but in agriculture. It was caused not by “automation” but by the fact that “excess” workers or disguised unemployed were drawn off the farms by higher paying factory jobs during and after the war.
In manufacturing the figures fail to show such startling change. Here output per man-hour in the postwar period has increased on the average by only 2.5 per cent a year. Therefore, if we go by the only measure we have, gross as it is, it would appear that there is little evidence for the proposition that the rate of technological change is accelerating rapidly. There seems to be some slow upward movement in the rate of change, the result of a healthier and better-educated labor force, new organizational techniques, and new technology. But the growth of technology today does not appear to create problems very different from those that have consistently faced the economy over the last sixty years.
What, then, of unemployment? A high rate has persisted for eight years now, and is decreasing only at a snail’s pace. If “automation” is not the cause, what is?
The general explanation should be fairly obvious: the economy has failed to grow as rapidly as the national rate of productivity, plus the increase in the labor force. If the output per man hour of a labor force of 70 million workers rises by 3 per cent each year, the economy must add two million jobs a year merely to keep pace with the increase in efficiency. Furthermore, we may expect that between now and 1970 the labor force itself will grow by about one and a half million people a year owing to the World War II baby boom. It is now estimated that the national product must increase by 4.5 per cent annually during the next five years simply to prevent unemployment from increasing. Yet the economy has been able to attain this rate of growth in only three of the last eight years.
The underlying reasons for this failure are not hard to find. We have inherited a staggering deficit from the deflationary economic policies put into effect during the Eisenhower years. Between 1947 and 1957, the gross national product increased at a rate of 3.8 per cent a year. From 1957 to 1962, however, it sagged to 3.1 per cent while the labor force swelled at an annual rate of 800,000. During these five years unemployment increased by an average of 200,000 annually. The problem was not the elimination of jobs by automation but the failure of the economy to grow sufficiently during the Eisenhower administration.
Since then, during the Kennedy-Johnson years, there has been some improvement: the number of jobs increased by 1.5 million in 1963 and 1964. But the labor force itself grew by 1.2 million, and thus total unemployment declined by only 290,000. Recent economic growth has not been rapid enough to make up for the deficits of the previous period.
We have been talking, it should be emphasized, about the economy as a whole. The problem of economic expansion, however, provides the basic clue to the differing effects of change on those industries which have become outmoded for technological or social reasons. In the 1950s, jobs declined in the textile industry, not because of automation but because young people, marrying earlier, spent more of their incomes on homes and furnishings than on clothes. In the coal industry, hundreds of thousands of jobs were lost as the competition of natural gas and oil reduced the demand for coal. Union economists claim that in 1963 the automobile industry produced as many cars as it did in 1955, with 17 per cent fewer workers. But this would only prove that the productivity in the auto industry was increasing at the same rate as in American industry generally. (In fact productivity in the auto industry is considerably higher than the average but this is offset by the tremendous profits made by the major companies.) The important fact is that until two years ago the output of the auto industry was not growing—given the state of the American highways we were better off for that—and under these conditions high productivity could only result in a shrinking labor force. In the last year the number of jobs in Detroit has grown and unemployment is lower than the national average. Clearly an industry has to expand in order to absorb increases in productivity if employment in that industry is to remain stable or to grow.
Some examples may help to make this point clear. During the past few years no industry has made greater use of automation than the federal government, which has had to handle an extraordinary amount of “data processing.” But when the Civil Service Commission undertook a survey of 1,325,000 employees in ten federal agencies it found that over a period of three years only 1,628 had been displaced from their jobs by computers; and of these, 77 per cent were reassigned at the same level, while no more than 2 per cent actually lost their jobs. Automation was relatively harmless in this case because the federal government is a rapidly expanding industry. Similarly, the recent, and long overdue, introduction of electronic equipment in banks probably will not throw many employees out of work, because banking—like the telephone industry, another example of large-scale automation—is still growing. True, there will now be fewer additional jobs in these industries than before automation. But this after all is how the standard of living rises—higher productivity allows industry to handle more work at less cost. If it were not for the dial system, the telephone industry would now require the services of every female in the United States.
Still, we all know that the effects of increased productivity in some industries can be cruel. In the printing industry, an entire craft is faced with extinction as typewriter-linked computers replace linotype machines. But where there is effective collective bargaining, the rate of technological displacement can be adjusted to the retirement pattern of the labor force in that industry, and a man’s job can be safeguarded as long as he is willing and able to work. (In the New York newspaper dispute, the publishers were willing to grant this attrition principle, but the union insisted that the present level of jobs be permanently maintained.)
It is important to locate the problem accurately, for some of the popular arguments about “automation” are dangerously misleading. It is frequently claimed, for example, that the causes of unemployment are “structural”: changes in technology are allegedly becoming so drastic that the unskilled and poorly educated will never be able to find jobs of any kind. According to this view, unemployment can be eliminated by retraining workers and launching a wholesale educational effort to upgrade the incompetent and the poor, or simply by giving them an income without work. The argument is that technology is making man redundant.
We find curious alignments among those who make “structuralist” arguments of this kind. On the one hand, there are conservative writers and officials of the Federal Reserve Board who oppose government “interference” in the economy and vast public spending. They fasten on the automation issue in order to attack Keynesian fiscal and monetary policies. For if the problem of unemployment can be traced simply to a lack of manpower training, rather than a failure of demand, there is no need to engage in deficit financing to solve. Just retrain the unemployed.
On the other hand, the contingent represented by such writers as Robert Theobold of the Ad Hoc Committee on the Triple Revolution, the economist Ben Seligman, and others further to the left do not believe that the American economy can solve the unemployment problem by Keynesian methods. They find in “automation” proof that a crisis is brewing which can only be solved by drastic changes in the system.
It should be obvious that efforts to improve manpower training and the education of the poor are worthwhile, to say the least. But few of the leading academic economists such as Paul Samuelson and Robert Solow of M.I.T. and James Tobin of Yale believe that these measures can bring about full employment nor do they agree that structural changes have made full employment an unattainable goal. They argue that unemployment is high not because automation has put the unskilled and the badly educated out of work, but because the aggregate demand for labor has been weak—again, the economy has not grown fast enough. As one economist recently put it: Does anyone believe that if the United States got into a war like the Korean War we would still have unemployment? The answer is quite clear: more spending would create jobs for all who wanted them. And much of the current concern about manpower retraining might vanish, too; the experiences of World War II and the Korean War showed that when there is a demand for labor, industry goes to great lengths to adapt work to existing skills; or it develops rapid training programs of its own. And this is what has happened in Detroit today where the new demand for autos has reduced “structural” unemployment.2
The question, then, is how to stimulate expansion in the economy. The administration, of course, has argued that a sufficiently large tax cut would increase aggregate demand and employment. And in fact the experience of the past year indicates that the tax cut has had a positive economic effect. But as Leon Keyserling, among others, has cogently argued, the tax cut is socially ineffective because it often puts money in the hands of groups who spend it in ways that are of little benefit to the economy as a whole. It has, for example, sometimes made extra sums available to corporations which have used the money for investment abroad—thus compounding the balance-of-payments problem—rather than increasing the purchasing power of low-income groups.
Keyserling, and other liberal economists have insisted that economic expansion will come about more readily, and various social needs would be served best, by direct government spending in areas such as housing, urban transport, medical care, and education. This argument seems to me convincing.
But it has been extraordinarily difficult to get the government to undertake programs of this kind. Those of the Johnson administration are a step in the right direction but are hardly more than a beginning. It has recently been estimated, for example, that to achieve even the modest goals set by the Eisenhower Commission for adequate national health, education, urban development, transport etc. by 1975 will cost more than $1,130 billion3 . This is about $150 billion more than the estimated total gross national product itself if we grow at the rate of 4 per cent a year during the coming decade. In fact it will require a net growth rate of economy as it is now organized. These facts make all the more absurd the claims of the Ad Hoc Committee that technology is making manpower redundant. With so many needs to be filled, we must have a rise in productivity that will require not only increased automation but the full employment of the entire labor force. Meanwhile, the specters of automation can only distract us from the really difficult political and economic problems that lie ahead.
RECENT BOOKS ON AUTOMATION
Manpower Report of the President 1965 (U. S. Government Printing Office, $1.75) This is the third of the annual reports now required by Congress. The three reports together provide a comprehensive picture of the manpower resources of the country. The 1963 report is still the most useful, in that it supplies basic data on productivity and future manpower needs. The 1965 report concentrates on retraining and development.
Unemployment in the United States by Robert M. Solow. (Almquist & Wiksell, Stockholm) A 51-page essay which is the most meticulous analysis of the “structuralist” hypothesis about unemployment that I know of.
Men Without Work: The Economics of Unemployment edited by Stanley Lebergott. (Prentice-Hall, 1964, paper $1.95, cloth $4.95) An excellent compendium that brings together the conflicting views on the automation issue, plus material on characteristics of the unemployed, and training programs here and abroad.
Automation and Technological Change edited by John T. Dunlop. (Prentice-Hall, 1962, paper $1.95, cloth 3.95) An American Assembly volume which usefully parallels the Lebergott collection in contentrating on the consequences of automation for businesses, for collective bargaining, and for changes in society.
The New Improved American by Bernard Asbell. (McGraw-Hill, 1965, $5) Hailing the potential of automation, Asbell argues that coming changes will require a complete overhaul of our educational system. His descriptions of the inadequacy of American education are telling, and his proposals for necessary reforms are procovative. Yet as Asbell asserts, it is doubtful whether the reasons for such reforms lie in the needs of the new technology.
Beyond Automation: Managerial Problems of an Exploding Technology by John Diebold. (McGraw-Hill, $7.50) A hortatory volume, collecting Mr. Diebold’s legislative statements and public addresses over the past several years; with such statements (under the heading of “A Program for Action”) as: “The State of New York should conduct and sponsor detailed studies of the human and economic consequences of automation within the state,” 216 overblown pages for $7.50? A disgrace.
The Automation Hysteria by George Terborgh. (Machinery and Allied Products Institute, $4) Despite the flamboyant title and somewhat strident tone, this book by an economist who, a generation ago, challenged the Alvin Hansen thesis of economic stagnation, is a useful compilation of evidence on the economic effects of automation so far. By concentrating however on refuting the “alarmists” one is left with the bewildering sense that automation presents no problem at all.
The Next Generation by Donald N. Michael. (Random House, $4.95) An effort to identify the kinds of social changes and problems which the country may face in the next twenty years, the book suffers from a skimpiness (e.g., twelve significant technological areas are covered in twenty pages) which reduces its use for any serious student of the problem.
August 26, 1965
Productivity is the measure of the efficiency of a given set of economic factors, and is not to be confused with the gross national product which is the total monetary value of goods and services. The increase in productivity in the economy is different from the increase in national product, and the two are not always linked. One can have increases in productivity in the economy while the gross national product is decreasing—usually because firms will strive to reduce costs and increase efficiency when business takes a down-turn. ↩
This is, of course, an ad hominem way of dismissing the argument, but it is supported by a careful analysis of the structuralist hypothesis by Robert M. Solow of M’I.T. in his 1964 Wicksell lectures, The Nature and Source of Unemployment in the United States. Professor Solow has meticulously demonstrated, for example, that there was no increase in regional variation of unemployment between 1950 and 1960. Nor have the proportions of unemployment between occupations (from managerial to unskilled) and between industries varied over the past ten years. Yet according to the structuralist hypothesis, some regions should have been more affected than others; and the unskilled should have borne a more than proportional increase in the rates of unemployment. The evidence that, over the decade, there has been no increase in regional variation, or in the proportions of joblessness between groups, would indicate support for the “aggregate demand” thesis I am arguing here. ↩
In the Dollar Costs of National Goals. The National Planning Association, Washington, D.C. ↩