Technology's Revenge

SYNOPSIS: Krugman discusses why increasing income inequality might be due to the nature of new production technology

In his science-fiction novel of 1952, Player Piano, Kurt Vonnegut imagined a future in which the ingenuity of engineers has allowed machines to eliminate virtually all manual labor. The social consequences of this technological creativity, in his vision, are disastrous: Most people, instead of finding gainful employment, live on the dole or are employed in pointless government make-work programs. Only the most creative and talented can find meaningful work, and their numbers steadily shrink as more and more jobs are automated out of existence.

For the first 20 years after Player Piano appeared, it seemed that Vonnegut could not have been more wrong. Between World War II and the early 1970s, the world's advanced economies were spectacularly successful at creating precisely the kind of employment that he imagined automation would destroy: well-paying jobs for workers of average skills and education. Social observers waxed eloquent over the unprecedented prosperity of the working class. Thanks to the 30-year "Go-Getter Bourgeois business boom," writer Tom Wolfe announced, "the word proletarian can no longer be used in this country with a straight face." Economists, who had always regarded most fears about automation as nonsense, felt confirmed in their dismissal of the issue.

But the past 20 years have not been good ones for ordinary workers. Even as the earnings of many college-educated workers soared in the United States, young men without college degrees have seen their real wages drop by 20 percent or more--this in spite of productivity growth which, while disappointing, nonetheless allowed the average American worker to produce about 25 percent more in 1993 than in 1973. In Europe, the growth of wage inequality has been less dramatic, but there has been a steady, seemingly inexorable rise in unemployment, from less than three percent in 1973 to more than 11 percent today (versus six percent in the United States).

Many economists believe that the American and European experiences are two sides of the same coin. For whatever reason, employers have been increasingly reluctant to pay for the services of those who do not offer something exceptional. In the United States, where unemployment benefits are relatively skimpy and of relatively short duration (26 weeks), and where the unemployed often find themselves without health insurance, workers have little choice but to accept jobs not matter how low the pay. Thus, U.S. labor markets have been, in the fine euphemism of official documents, "flexible." In Europe, much more generous social benefits make it easier for workers to turn down jobs they find unacceptable, and various government regulations and restrictions make employers less willing and able to offer low-wage jobs in any case. Thus, the same forces that lead to less pay for the less skilled in the United States lead to rising unemployment for the same group in Europe. The larger outcome is the same on both sides of the Atlantic: The broad equality of economic outcomes that the postwar West had come to take for granted seems to be receding into memory.

Most people who read intellectual magazines or watch public television know why this is happening. Growing international competition, especially from low-wage countries, is destroying the good manufacturing jobs that used to be the backbone of the working class. Unfortunately, what these people "know" happens to be flatly untrue. The real reason for rising wage inequality is subtler: Technological change since 1970 has increased the premium paid to highly skilled workers, from data processing specialists to physicians. The big question, of course, is whether this trend will continue.

Before we can get to that question, however, it is necessary to clear away some of the underbrush. Much public discussion of jobs--even among people who consider themselves sophisticated and well-informed--has been marked by basic misunderstandings of the facts. Consider this statement: "Modern technologies of transportation and communication make it possible to produce anything anywhere. This technological shrinking of the world has only been reinforced by the fall of communism, which has made the Third World safe for multinational corporations. As a result, a massive redeployment of capital and technology from the high-wage countries of the West to low-wage developing nations is now occurring. This redeployment of capital along with the flood of low-cost imports is destroying the well-paying manufacturing jobs that used to support a large middle class in the United States and Europe. In short, globalization favors Western capital, but it is devastating to Western labor."

Convincing as this may sound, the statement is specious. In fact, I made it up to illustrate a view of the world that passes for sophistication among many policy intellectuals but is almost completely refuted by the available evidence.(FN*)

At the basic level, this conventional view suggests that capital and technology are in fixed supply, and that growth in new countries necessarily comes at the expense of the more established countries. The reality is that the diffusion of technology, while it increases competition faced by the leaders' exports, also expands their markets and reduces the price of their imports. For example, the United States must buy virtually all of its laptop computers from foreign producers, but the growth of overseas production has enlarged markets for U.S.-made microprocessors and cut the price of laptops. In principle, the net result of the diffusion of technology could be either to raise or to lower First World income. In practice, there is little discernible effect.

Nor is the world supply of capital a fixed quantity. As countries grow, they also save--in the case of rapidly growing Asian nations, they save at astonishing rates. Third World growth may thus add to the world supply of capital as fast as or faster than it increases the demand.

Moreover, the amount of imports arriving from newly industrializing countries and the size of capital flows going to them fall far short of what is suggested in alarmist rhetoric. If there is a single piece of knowledge that separates serious international economists from fashionable popularizers, it is a sense how big the world economy really is. We have all heard enough stories of particular factories that have moved to Mexico or Indonesia to form the impression that a massive global trend is underway. But even a billion-dollar investment is insignificant amid the sheer immensity of the economies of the industrialized nations. Their combined gross domestic products in 1990 exceeded $19 trillion, and their combined domestic investment exceeded $4 trillion. The total movement of capital to newly industrializing countries in 1993--a record year, unlikely to be surpassed in 1994--was roughly $100 billion. That is, less than 2.5 percent of the investment of the First World actually flowed south. While it is true that tens or even hundreds of thousands of workers in advanced countries have lost their jobs to low-wage imports, the total labor force in the industrialized world is more than 400 million strong; almost every effort to quantify the reasons why more than 30 million of these workers do not have jobs finds that Third World competition plays little if any role. That is not to say that international trade and capital mobility could not have a more important impact in the future. But declining wages and rising unemployment are not things that might happen once globalization really gets going; they are trends that have been in progress for 20 years. What is causing them?

Economists use the word "technology" somewhat differently from normal people. Webster's defines technology as "applied science," which is pretty much the normal usage. When economists speak of technological change, however, they mean any kind of change in the relationship between inputs and outputs. If, for example, a manufacturer discovers that "empowering" workers by giving them a voice in how the factory is run improves quality--and allows the plant to employ fewer supervisors--then in the economic sense this would be an improvement in the technology, one that is biased against employment of managers. If, however, a manufacturer discovers that workers will produce more when there are many supervisors constantly checking on them, this is also a technological improvement, albeit one biased toward employment of managers.

In this economist's sense, it seems undeniable that over the past 20 years the advanced nations have experienced technological change that is strongly biased in favor of skilled workers. The evidence is straightforward. The wages of skilled workers, from technicians to corporate executives, have risen sharply relative to the wages of the less skilled. In 1979, a young man with a college degree and five years on the job earned only 30 percent more than one with similar experience and a high school degree; by 1989, the premium had jumped to 74 percent. If the technology of the economy had not changed, this sharp increase in the relative cost of skilled workers would have given employers a strong incentive to cut back and substitute less-skilled workers where they could. In fact, exactly the opposite happened: Across the board, employers raised the average skill level of their work forces.

It is hard not to conclude that this technologically driven shift in demand has been a key cause of the growth of earnings inequality in the United States as well as much of the rise in unemployment in Europe. It is not the only possible explanation. It could have been the case that rising demand for skilled workers was not so much the result of greater demand for skill within each industry as of a shift in the mix of industries toward those sectors that employ a high ratio of skilled to unskilled workers. That sort of shift could, for example, be the result of increased trade with labor-abundant Third World countries. But in fact the overwhelming evidence is that the demand for unskilled workers has fallen not because of a change in what we produce but because of a change in how we produce.

Is it really possible for technological progress to harm large numbers of people? It is and it has been. Economic historians confirm what readers of Charles Dickens already knew, that the unprecedented technological progress of the Industrial Revolution took a long time to be reflected in higher real wages for most workers. Why? A likely answer is that early industrial technology was not only labor saving but strongly capital using--that is, the new technology encouraged industrialists to use less labor and to invest more capital to produce a given amount of output. The result was a fall in the demand for labor that kept real wages stagnant for perhaps 50 years, even as the incomes of England's propertied classes soared.

Economists more or less agree that the same thing is happening to the Western world today, except that the benefits of biased technological change are flowing not to capital but to the highly skilled.

It is easy to understand why the Industrial Revolution was capital using and labor saving. Just think of a factory full of power looms replacing thousands of hand weavers--the development that gave rise to the Luddite rebellion in early-19th-century Britain. Can we come up with comparable images that relate recent technological change in the economist's sense to its more normal usage? That is, what is changing in the way that we produce goods and service that has apparently devalued less-skilled workers?

The short answer is that we do not know. There are, however, several interesting stories and pieces of evidence.

Probably the simplest story about how modern technology may promote inequality is that the rapid spread of computers favors those who possess the knowledge needed to use them effectively. Anecdotes are easy to offer. Economist Jagdish Bhagwati cites the "computer with a single skilled operator that replaces half a dozen unskilled typists." Anecdotes are no substitute for real quantitative evidence, but for what it is worth, serious studies by labor economists do suggest that growing computer use can explain as much as one-half of the increase in the earnings edge enjoyed by college graduates during the 1980s.

Yet there is probably more to the story. The professions that have seen the largest increases in incomes since the 1970s have been in fields whose practitioners are not obviously placed in greater demand by computers: lawyers, doctors, and, above all, corporate executives. And the growth of inequality in the United States has a striking "fractal" quality: Widening gaps between education levels and professions are mirrored by increased inequality of earning within professions. Lawyers make much more compared with janitors than they did 15 years ago, but the best-paid lawyers also make much more compared with the average lawyer. Again, this is hard to reconcile with a simple story in which new computes require people who know how to use them.

One intriguing hypothesis about the relationship between technology and income distribution, a hypothesis that can explain why people who do not operate computers or fax machines can nonetheless be enriched by them at the expense of others, is the "superstar" hypothesis of Sherwin Rosen, an economist at the University of Chicago. Almost 15 years ago, before the explosion of inequality had become apparent, Rosen argued in the Journal of Political Economy that communication and information technology extend an individual's span of influence and control. A performance by a stage actor can be watched by only a few hundred people, while one by a television star can be watched by tens of millions. Less obviously, an executive, a lawyer, or even an entrepreneurial academic can use computers, faxes, and electronic mail to keep a finger in far more pies than used to be possible. As a result, Rosen predicted, the wage structure would increasingly come to have a "tournament" quality: A few people, those judged by whatever criteria to be the best, would receive huge financial rewards, while those who were merely competent would receive little. The point of Rosen's analysis was that technology may not so much directly substitute for workers as multiply the power of particular individuals, allowing these lucky tournament winners to substitute for large numbers of the less fortunate. Television does not take the place of hundreds of struggling standup nightclub comedians; it allows Jay Leno to take their place instead.

Will technology continue to favor a few lucky people over the rest, or will the last quarter of the 20th century turn out to have been a transitory bad patch for the common man? At first sight, it seems obvious that the progress of technology must lead to an ever-growing premium on skill. How could it be otherwise in an era when sophisticated computers and information systems are becoming ever more crucial to our economy? Isn't it obvious that the only good jobs will be for those who possess exceptional intellectual talent and skills--those who, in the phrase of Secretary of Labor Robert Reich, are able to work as "symbolic analysts"?

History teaches us, however, that merely assuming a continuation of recent trends is often very misleading. Technology is less like a railroad track than a spiral staircase, with many reversals of direction along its upward path. The long-term effect of the Industrial Revolution is a case in point. To Victorian futurists, it seemed obvious that the capital-using bias of industrial technology would continue indefinitely, bringing with it an ever-greater gulf between the owners of capital and the working class. In The Time Machine (1895), H. G. Wells forecast a future in which workers have been reduced to subhuman status. These Victorians were wrong--indeed, if Wells had possessed the kind of data available today, he would have known that wages had begun to rise again long before he wrote his novel. During the 20th century, capital has claimed a declining share of the national income and labor has taken a growing share.

Technological advance, moreover, does not always increase the need for skilled labor. On the contrary, in the past one of the main effects of mechanization was to reduce the special skills required to carry out many tasks. It took considerable skill and experience to weave cloth on a hand loom, but just about anybody could learn to tend a power loom. What is true is that, to date, technological progress has consistently tended to increase the demand for a particular kind of skill, the kind that is taught in formal education and is most easily acquired by the kind of reason who does well in formal education. Two centuries ago, only a minority of jobs required literacy; one century ago, only a few jobs required anything like a modern college education. Nowadays higher education is not a luxury for the wealthy but something intensely practical, a virtual necessity for the career minded.

But it is not all clear that this trend will continue indefinitely. There is no inherent reason why technology cannot be "college-education saving" rather than college-education using. It is possible to see examples of how this might occur even today. This essay, for example, was written using a newly acquired word processor. I did not bother to read the manual; the graphical interface, with its menus of icons, usually makes it obvious how to do what I want, and I can easily call up on-screen help with the push of a button if I get lost. Whenever we use the term "user-friendly," we are implying that we have a production technique that requires less skill than it used to.

But isn't this kind of reversal always going to be the exception rather than the rule? Not necessarily. In fact, I would make a speculative argument that in the long run technology will tend to devalue the work of "symbolic analysts" and favor the talents that are common to all human beings. After all, even the most brilliant specialists are actually rather poor at formal reasoning, while even the most ordinary person can carry out feats of informal information processing that remain far beyond the reach of the most powerful computers. As the artificial intelligence pioneer Marvin Minsky points out, "A 1956 program solved hard problems in mathematical logic, and a 1961 program solved on college-level problems in calculus. Yet not until the 1970s could we construct robot programs that could see and more well enough to arrange children's building blocks into simple towers.... What people vaguely call common sense is actually more intricate than most of the technical expertise we admire." Chess-playing programs are not yet quite good enough to beat the world's greatest players, but they are getting there; a program that can recognize faces as well as a two-year-old can remains a distant dream.

Rereading Player Piano recently, I found the totally automated factories. Vonnegut imagined more than 40 years ago completely credible, but found myself wondering who cleans them (or for that matter the houses of his industrial elite)? It is no accident that no description is given of how these mundane tasks are automated--because as Vonnegut must have sensed, it will be a very long time before we know how to build a machine equipped with the ordinary human common sense to do what we usually regard as simple tasks.

So here is a speculation: The time may come when most tax lawyers are replaced by expert systems software, but human beings are still needed--and well paid--for such truly difficult occupations as gardening, house cleaning, and the thousands of other services that will receive an ever-growing share of our expenditure as mere consumer goods become steadily cheaper. The high-skill professions whose members have done so well during the last 20 years may turn out to be the modern counterpart of early-19th-century weavers, whose incomes soared after the mechanization of spinning, only to crash when the technological revolution reached their own craft.

I suspect, then, that the current era of growing inequality and the devaluation of ordinary work will turn out to be only a temporary phase. In some sufficiently long run the tables will be turned: Those uncommon skills that are rare because they are so unnatural will be largely taken over or made easy by computers, while machines will still be unable to do what every person can. In other words, I predict that the current age of inequality will give way to a golden age of equality. In the very long run, of course, the machines will be able to do everything we can. By that time, however, it will be their responsibility to take care of the problem.

America's Fastest Growing Occupations, 1992-2005 (In parentheses: the number of projected new jobs, in thousands)

Percent change

Home health aides (479) 138

Human services workers (256) 136

Personal and home care aides (166) 130

Computer engineers and scientists (236) 112

Systems analysts (501) 110

Physical and Corrective therapy assistants and aides (57) 93

Physical therapists (79) 88

Paralegals (81) 86

Teachers, special education (267) 74

Medical assistants (128) 71

Detectives, private (41) 70

Correction officers (197) 70

Child care workers (450) 66

Travel agents (76) 66

Radiologic technologists and technicians (102) 63

Nursery workers (44) 62

Medical records technicians (47) 61

Operations research analysts (27) 61

Occupational therapists (24) 60

Legal secretaries (160) 57

Teachers, preschool and kindergarten (236) 54

Manicurists (19) 54

Producers, directors, actors, and entertainers (69) 54

Speech-language pathologists and audiologists (37) 51

Flight attendants (47) 51

Guards (408)

Source: U.S. Bureau of Labor Statistics

Maschinenmänner (1930), by Heinrich Hoerle

American industry is producing more with fewer workers: Two million manufacturing jobs disappeared between 1988 and '93.


* For a fuller discussion of this point, see my article in the Harvard Business Review (Summer 1994). In a comprehensive survey of the literature on job creation, High and Persistent Unemployment: Assessment of the Problem and its Causes (1993), economist Jørgen Elmeskov flatly concludes that "trade seems an unlikely prime candidate for explaining increased unemployment."

The fastest growing occupations in percentage terms are not necessarily those that will produce the largest number of new jobs. The most growth in absolute terms will occur in the retail sales clerk category, which will grow by 786,000 jobs (21 percent) between 1992 and 2005.

Originally published, Fall.94