The Technology Trap

by Carl Benedikt Frey

  Though this process has been going on for decades, it has intensified in recent years, as new technologies, like multipurpose robots, have come into play. These robots are automatically controlled so that they do not require human operators. And they can be reprogrammed to perform various manufacturing tasks, like welding, assembling, or packaging. Thus, they must be distinguished from single-purpose robots and other computer-controlled machine tools, which are designed for one specific purpose.

  All analyses of the role of technology in rising male joblessness are regrettably limited to multipurpose robots, as systematic data on single-purpose robots remain sparse. But even if this means that we underestimate the pervasiveness of robots in the economy, such data are still informative. Daron Acemoglu and Pascual Restrepo estimate that each multipurpose robot has replaced about 3.3 jobs in the U.S. economy. Blue-collar people in heavily robotized industries—like automobile manufacturing, electronics, metal products, chemicals, and so on—have naturally felt the force of automation most keenly. But where robots have been adopted, people in nearly all occupations suffered both wage and employment losses. Job losses were unsurprisingly more significant among workers without college degrees. And men have been twice as likely as women to find themselves replaced by robots.35

  Those estimates focus on the period 1993–2007. However, statistics on the adoption of multipurpose robots from the International Robotics Federation show that American industry has continued to use robots more extensively in the postrecession period: the number of robots in use grew by almost 50 percent between 2008 and 2016. But needless to say, the worker-replacing effects of robots may have been counterbalanced by other technologies. Computers, as noted above, have also spun off new tasks for labor, in which they augment workers’ skills. All the same, there is compelling evidence that technological change in general has become more worker replacing in recent years. As we saw in figure 10, real wages for non-college-educated men have fallen since the 1980s. Of course, this might also reflect some other permanent factor, like jobs being sent abroad. But there is more direct evidence to suggest that technology has become increasingly labor replacing. In a major study of eighteen Organisation for Economic Co-operation and Development (OECD) countries, published in 2018, the economists David Autor and Anna Salomons found that whether automation is measured by productivity gains, patenting flows, or the implementation of multipurpose robots, it has reduced the share of national income captured by labor. Technological progress, they show, only turned replacing as computers became more pervasive in the 1980s, and its negative effects on the labor share became more pronounced throughout the 2000s.36

  The Return of Engels’s Pause

  The age of automation is not without parallel, however. We saw in chapter 5 how the Industrial Revolution caused a hollowing out of middle-income jobs, put downward pressure on workers’ wages, and prompted an upsurge in inequality. The classic years of industrialization are known as Engels’s pause, when the mechanized factory displaced domestic industry, worsening the economic prospects of many citizens even as the British economy took off. During the classic years, output experienced an unprecedented expansion, yet the gains from growth didn’t trickle down to most people. Output per worker grew more than three times faster than average weekly wages. As middle-income artisan jobs were trimmed, the gains of the Industrial Revolution went to industrialists who saw the rate of profit double. Engels’s assertion that industrialists grew rich on the misery of workers was broadly right for the period he observed. The pause came to an end only around 1840.

  If Friedrich Engels were living today, what would he have written about the computer era? Working conditions in the industrial West clearly do not have much in common with the “dark, satanic mills.” But the trajectories of per capita output and people’s wages look exceedingly similar. In America, labor productivity has grown eight times faster than hourly compensation since 1979.37 Even as the American economy has become much more productive, real wages have been stagnant, and more people are out of work; consequently the labor share of income has fallen. Corporate profits have swept up an ever-greater share of national income while the share going to workers has rarely been smaller. And official measures of labor compensation include the paychecks of CEOs and superstars in music, sports, and media, meaning that the percentage of income going to the average worker has declined even further. As was the case in the classic years of the Industrial Revolution, the gains from growth have shifted from the bottom to the top of the income distribution and from labor to owners of capital. During the postwar years, the labor share hovered around 64 percent, but since the 1980s it has steadily declined to its lowest postwar level after the Great Recession, averaging around 58 percent in recent years.38 This is in accord with the trends depicted in figure 9, where we saw a widening gap between labor productivity and worker compensation emerging in the 1980s. And it is not just a U.S. phenomenon. The economists Loukas Karabarbounis and Brent Neiman, for example, have documented that the share of national income that goes to labor has declined dramatically in most countries since the 1980s, which, they argue, is thanks to cheaper computers.39

  There are good reasons to think that rising profits and the falling labor share is linked to the automation of routine middle-income jobs (such as those of machine operators, bookkeepers, and mortgage underwriters) and the shift of labor into low-income service jobs (for example, those of janitors, waiters, and receptionists). In 2017, the International Monetary Fund (IMF) published a report showing that “technological advancement, measured by the long-term change in the relative price of investment goods, together with the initial exposure to routinization, have been the largest contributors to the decline in labor income shares in advanced economies.”40 Consistent with the hollowing out of labor markets, as computer-controlled machines have taken over the jobs of the middle class, the IMF found that the decline in labor shares has been particularly sharp for middle-skilled workers.

  The changing face of technology is also reflected in long-run trends in the Gini coefficient (figure 14). As Branko Milanovic has noted, “This revolution [the computer revolution], like the Industrial Revolution of the early nineteenth century, widened income disparities.”41 These periods were not only times when the profit share of income reached historical heights and the wages of ordinary citizens were stagnant. As noted above, both were episodes when technology replaced middle-income workers. In the computer era, the increase in inequality happened in large part because new technologies strongly rewarded more highly skilled symbolic analysts while driving up the capital share of national income. At the same time, as middle-income routine jobs were shredded, unskilled labor moved toward low-paying service occupations, causing wage disparities to rise. In similar fashion, technological change during the Industrial Revolution pushed people out of middle-income jobs in the domestic industry, to the detriment of many craftsmen, while creating low-income production jobs in the factories and high-paying skilled jobs for white-collar workers to manage and administer production. Indeed, the economists Lawrence Katz and Robert Margo have pointed out that the effects of computers on the labor market today have been similar to those that accompanied the spread of the mechanized factory in the nineteenth century.42

  FIGURE 14: Income Inequality in England/U.K. and the U.S., 1688–2015

  Sources: See appendix, this volume.

  So far, new computer technologies have not caused widespread unemployment as has been so widely feared. Though industries and occupations have lost jobs due to automation, job losses have been offset by the creation of new tasks, customers and suppliers benefiting from cheaper goods, and increases in overall consumer spending.43 But computer technologies have shrunk the size of the middle class, put downward pressure on unskilled workers’ wages, and reduced labor’s share of income. And, as the experience of the Industrial Revolution illustrates, even when new jobs are being added, it can take a long time for workers to acquire the necessary skills to s
uccessfully move into the newly emerging jobs. In many cases, new or changing job roles require a different breed of worker. Case studies of office automation have shown that computers reduced clerical staff in routine activities, while opening up only a “relatively small number of better paid positions for programming and operating the new systems.”44 Aptitude tests provided a means of selecting staff for the new and substantially more skilled jobs: “Those selected were chiefly men in their late twenties with some college education and some company experience in … related work.”45 Few older workers or employees displaced from the affected units were chosen for the newly created positions. The adverse impacts of automation were much greater for middle-age and older employees.

  When replacing technologies make the skills of existing workers redundant, they reduce the earning capacity of significant parts of the population. Though new tasks may be spun off in the process, new skills take time to learn and are often seen in workers’ wages only years later. As discussed in the context of the Industrial Revolution, the wages of power-loom weavers took off long after the jobs of hand-loom weavers were shredded. A modern equivalent is the case of typographers, which also illustrates this point vividly. One advantage of the computer was that a file could be saved to its memory, eliminating tedious rekeyboarding of typewritten text to correct errors. The effect on typographers’ jobs and wages was significant. James Bessen has calculated that between 1979 and 1989, employment among typesetters and compositors fell from 170,000 to about 74,000, while their median wages declined by 16 percent, adjusted for inflation. “Membership in the International Typographical Union fell sharply, and in 1986, much weakened, it merged with another union.”46

  As computer publishing eliminated tedious rekeyboarding of text and made typesetting less costly and more versatile, desktop publishers and graphic designers took on much of the typographical work, and employment opportunities in these occupations soared. But making this transition required workers to learn graphic design software, such as page layout programs. While we lack the data to determine the extent to which typographers successfully took on jobs as graphic designers, these jobs required very different skills from those used in typography, and few typographers probably made this transition. And even those who managed to become graphic designers didn’t see their wages rise, as average pay for designers remained stagnant. Bessen explains:

  After accounting for inflation, the average hourly pay of graphic designers has been stagnant in recent years; the average pay of all types of designers has actually fallen since the 1970s. While designers are paid, on average, a bit more than typographers, the median designer of 2007 earns only about a dollar more per hour than the median typographer of 1976. Designers seem to have shared little in the benefits of this technology. Why don’t average designers earn more, now that they have acquired substantial new skills and job responsibilities? Because the technology and organization of work for designers seem to be in constant flux. The print designers who replaced the typographers have been partly replaced by web designers, who are partly being replaced by mobile designers. Technology is continually redefining what publishing is and how it is done. Each of these changes requires new, specialized skills—skills learned largely through experience or by sharing knowledge, rather than in school. Each year, designers have to learn new software and new standards in order to keep up. A few years ago they learned Flash; now it is HTML5. Next year, perhaps something else.47

  When occupational skills are replaced by machines, the investment workers have made in building up the human capital associated with that occupation has gone industrially bankrupt. A worker displaced from a steel mill will not be able to begin a new career as a barber the next morning, and he or she is rarely equipped to switch into a professional, managerial, or engineering job. The higher the cost of accumulating new human capital, the longer the transition will take. Even low-skilled service jobs in restaurant, hotels, and gasoline stations require some skills. Experience is valuable in just about every occupation. But unquestionably, the cost of acquiring new human capital to move into well-paying jobs has become much greater in the ever-higher tech economy, leading to a hardening division between those who went to college and whose who did not. What’s more, as we shall see, location matters about as much as education. The most serious adjustment problems have occurred among unskilled workers in declining towns and cities.

  10

  FORGING AHEAD, DRIFTING APART

  As discussed above, Engels’s pause was a time when citizens saw rapid change, not just in the workplace but also in the communities in which they lived. Indeed, in the early nineteenth century, social critics like Peter Gaskell made the impact of mechanized industry on society the center of public debate. Gaskell’s The Manufacturing Population of England: Its Moral, Social, and Physical Conditions, published in 1833, was the first of its kind and a source of inspiration for the essay Friedrich Engels later wrote on the conditions of the English working classes. Gaskell believed that the struggle between human power and steam-powered machinery was approaching a crisis, and he argued that mechanization was changing the “very framework of the social confederacy.”1 In addition to depriving craftsmen of jobs in rural industry, he declared, the factory system was creating a new socially deprived class in Britain’s emerging industrial cities: “The universal application of steam power as an agent for producing motion in machinery, has closely assimilated the condition of all branches, both in their moral and physical relations. In all, it destroys domestic labour; in all, it congregates its victims into towns, or densely peopled neighbourhoods; in all, it separates families.”2

  The computer revolution, as we shall see, has caused to the demise of many of the factory cities that industrialization once gave rise to. And as was the case with the Industrial Revolution, its social consequences for individuals, families, and their communities have been profoundly negative. Since its peak in 1979, more than seven million American manufacturing jobs have disappeared. And America’s industrial towns and smokestack powerhouses, where blue-collar jobs were clustered, have felt the consequences most keenly. Where middle-class jobs withered, whether due to automation or globalization, a range of societal problems have emerged. In the 1990s, the sociologist William Julius Wilson attracted enormous attention with his study of urban ghettos where work had vanished. Using large-scale surveys and ethnographic interviews, he concluded that “the consequences of high neighborhood joblessness are more devastating than those of high neighborhood poverty. A neighborhood in which people are poor but employed is different from a neighborhood in which people are poor and jobless. Many of today’s problems in the inner-city ghettos—crime, family dissolution, welfare, low levels of social organization, and so on—are fundamentally a consequence of the disappearance of work.”3 While the work of Wilson focused on black neighborhoods in inner cities suffering from economic restructuring and suburbanization, many of their troubles are now shared by communities of the white working class.

  When Jobs Disappear

  No town or city could represent all of America, but the closest we can get is probably Port Clinton, New York, on the shores of Lake Erie. In Our Kids, sociologist and political scientist Robert Putnam looks back at his Port Clinton high school class of 1959, in what was then a blue-collar middle-class town, and in almost every respect a remarkable microcosm of America. He recalls that few of his classmates’ parents were educated. A full third had not even graduated from high school. Yet like just about everyone in town, they were reaping the benefits of postwar prosperity: “Some dads worked the assembly lines at the local auto part factories, or in the nearby gypsum mines, or at the local Army base, or on small family farms.”4 But in an era when technological change was of the enabling sort, few families had experienced joblessness or felt economic insecurity. Though few families in Port Clinton were affluent, very few were poverty-stricken. And their children were like everybody else’s. Regardless of social background, they participated in
many extracurricular activities, including sports, music, and drama: “Friday night football games attracted much of the town’s population.”5 Fast-forwarding half a century, Putnam’s classmates had also made a great leap forward relative to their parents. Three-quarters of them were better educated, and the overwhelming majority had moved up the economic ladder. Perhaps most strikingly, many children with less well-off parents had climbed farther up the rungs than those from more privileged, better-educated backgrounds. Upward mobility among the kids from relatively disadvantaged backgrounds was almost as great as among the affluent ones.

  Yet today, the American dream in Port Clinton is a split-screen nightmare. The children of the next generation are facing a very different reality. Those whose parents were symbolic analysts or the like have stayed on track, while those whose modest family fortune depended on factory jobs have founded themselves stuck on the wrong side of the tracks. As Putnam points out, the causes of faltering upward mobility in America are many. But one is surely that the manufacturing foundation upon which Port Clinton’s prosperity rested has dwindled. As well-paying blue-collar jobs were trimmed, births out of wedlock rose sharply, child poverty skyrocketed, and upward mobility went into reverse. Because blue-collar workers tend to shop close to where they live, they also support many other jobs in the service sector. Thus, the disappearance of blue-collar jobs meant a blow to the local service economy, with shops closing down and people leaving town. As work vanished, many people left in search of a better future elsewhere, leaving Port Clinton in despair: “The Port Clinton population, which had jumped 53 percent in the three decades prior to 1970, suddenly stagnated in the 1970s and 1980s, and then fell by 17 percent in the two decades after 1990. Commutes to jobs got longer and longer, as desperate local workers sought employment elsewhere. Most of the downtown shops of my youth stand empty and derelict, driven out of business partly by the Family Dollar and the Walmart on the outskirts of town, and partly by the gradually shrinking paychecks of Port Clinton consumers.”6