The Computer Did It? Technology and Inequality

The Computer Did It? Technology and Inequality

A renovated Chrysler assembly plant in Sterling Heights, Michigan (Chrysler Group/Flickr)

Does technological change breed inequality? This notion, in some version or other, has become a staple explanation for our growing economic divide. It features prominently in the “meritocracy” defense of the 1 percent, which assumes (as Harvard economist Greg Mankiw put it recently) that “changes in technology have allowed a small number of highly educated and exceptionally talented individuals to command superstar incomes in ways that were not possible a generation ago.” It peppers the breathless prose of columnist Thomas Friedman, who makes a living mixing metaphors about our flat, smart, hyperconnected, seamless supply chain of a world.

And it inevitably accompanies the annual handwringing over U.S. rankings in the international PISA (Program of International Student Assessment) educational achievement scores. The conviction that American workers are unprepared to meet the challenges of the new economy has underwritten a generation of educational reform animated in large part by concerns about national competitiveness. “Knowledge, learning, information, and skilled intelligence are the new raw materials of international commerce,” the Department of Education worried in 1983, concluding that—on this score—the United States was falling dangerously behind its peers.

Technology and inequality are linked by two apparent mechanisms. The first of these is the capital bias of recent technological change. Rapid automation, in this view, displaces labor entirely and delivers more and more of the returns on productivity directly to capital. Variations on this view (from writers such as Paul Krugman, Mother Jones’s Kevin Drum, and Jacobin’s Peter Frase) generally point to the slow but steady decline in labor’s share of national income since the early 1970s and look to a frightening future in which robots build robots while the rest of us look on.

The second mechanism is the skills bias of recent technological change. In this view, most prominently developed in the work of Lawrence Katz and Claudia Goldin, rapid technological change has outpaced educational achievement, and the demand for skilled labor has outstripped its supply. This has bid up the wages of workers (in the United States and beyond) with the right skills and left the unskilled and uneducated in the dust.

These explanations come together in the work of David Autor and his colleagues, which shifts our attention from educational gaps to occupational change. In this view, technology is hollowing out the labor market. Robots and computers are displacing middle-wage jobs (in administration, production, and fabrication) and reserving job growth for tech-savvy jobs at the top and low-wage service occupations (which are harder to displace with technology) at the bottom. Skill-biased technological change remains the culprit, but the mechanism by which it generates inequality (polarized job growth) results from changes in the demand for skilled workers rather than our failure to sustain the supply of skilled workers.

All of this, unfortunately, brings us to something of a political dead end. Technological innovation or change is not something you can readily control, so there seems little recourse but to occasionally lament the quality of American education or the unfortunate prevalence (in the recent recovery, or in the long term) of McJobs. Those who design or program the next Roomba will get rich; those displaced by it will not.

Recent work from Larry Mishel and Heidi Shierholz (Economic Policy Institute) and John Schmitt (Center for Economic and Policy Research [CEPR]) has recast this debate, calling into question the underlying pattern of polarized job growth and—more important—its utility in explaining American inequality. The significance of this intervention lies both in its convincing demolition of the “blame the robots” argument and in its political and policy implications.

The first insight is that there is nothing novel about the recent impact of technology and technological change on labor markets. A series of industrial and technological revolutions—the rise of the factory system in the early nineteenth century, the emergence of steam power (in production and infrastructure) in the later nineteenth century, the succession of electricity and “Fordist” mass production early in the twentieth century—all shared the same motive and logic: the search for productivity gains through the transformation or displacement of wage labor. Over this long haul, technological change has tended to push job growth to the margins—displacing middle-wage production and clerical occupations, increasing the productivity and skills of high-wage workers, and leaving many low-wage service jobs untouched.

But there is little evidence that this is now happening at an accelerated or game-changing pace. While we are commonly warned, as Kevin Drum put it recently, that “smart machines are going to put lots of people out of work over the next few decades, and this is going to substantially increase income inequality,” neither recent unemployment nor inequality trends suggest that this dystopian nightmare has begun. As CEPR’s John Schmitt and Kris Warner point out, the share of displaced workers reporting unemployment as a result of “plant closing” or “position abolished” has been pretty constant since such surveys were launched in the late 1980s; those reporting jobs lost due to “insufficient work,” by contrast, nearly doubled (from 22 to 43 percent) between 2006 and 2010. Increased unemployment, in other words, is cyclical (a reflection of insufficient demand in a recessionary economy) and not structural (the result of a skills mismatch or technological displacement).

More to the point, patterns of occupational change do not line up in any credible, chronological manner with concurrent patterns of wage inequality. As the graph below (drawing on Mishel, Shierholz, and Schmitt) summarizes, occupational shares across the last generation shifted fairly steadily—especially from middle-wage to high-wage occupations. But the corresponding trends in wage inequality are starkly discontinuous. There is a spike in the gap between low-wage and median-wage workers in the 1980s (especially stark for women, and starting a bit earlier for men), but this levels off after 1990. By the same token, trends in the wage gap at the top (between median and high-wage workers) also bear little relation to changing occupation shares—sometimes growing as the shift in occupational shares slowed, sometimes narrowing as the shift in shares picked up. Slow and steady shifts in the demand for low-, middle-, and high-wage workers, in short, seem to have little to do with the wage gaps between them.

Another way to look at this is to consider the actual returns on education. If the gap between the winners and losers was really about technological innovation or skills, we would expect to see a close correspondence between wages and educational achievement. Yet while those with a college education or better pulled away from the pack in the 1980s and 1990s (see figure below), that advantage has slowed dramatically. Wage growth has been flat since the late 1990s for all educational cohorts. And workers with a college degree have lost nearly as much ground since 2007 as everyone else.

Indeed, as John Schmitt and Janelle Jones have documented, workers with some college education or a college degree are now more likely to remain mired in low-wage work than workers with the same educational background a generation ago. The graphic below, based on their work, shows the distribution of low-wage workers, by educational attainment, for 1979 and 2012. In 1979 the educational returns are pretty clear: most low-wage workers are those with a high-school education or less. While the share of low-wage workers has fallen by 2012, their educational achievement has risen sharply: now very nearly half of the low-wage workforce (as compared to about a quarter in 1979) has some college education.

While K–16 education and workforce development are undoubtedly important (and underperforming), it is not clear that even dramatic improvements on that front would close the gap. The last fifteen years have seen significantly slower growth in high-skill, high-wage jobs than in the economy at large. Even those with the right skills are pounding the pavement. Education and training may not be the golden ticket but instead, as James K. Galbraith suggests, “a kind of lottery, whose winners and losers are determined, ex post, by the behavior of the economy.”

In short, the notion that inequality is generated by rapid technological change and skill shortages is not sustained by the recent American experience. If demand for certain workers or certain skills were reflected in wages, we would expect to see wage gains where that demand was highest and wage stagnation where it was weakest. But this is not the case. Since 1969 labor’s share of income has fallen most rapidly in those sectors where union presence withered, not where computers displaced labor. Across our last two business cycles, income concentrated not in sectors or regions where skills were most in demand but where speculative bubbles (dot-com, housing, finance) bloomed and burst. During our most recent recession and recovery, the notion of a “skills shortage” was belied by the fact that job openings and available workers were distributed fairly evenly across the economy, and that skilled workers saw no “bidding up” of their wages or increase in their work hours. Indeed, most of the growth in wage inequality across the last generation can be found within occupations, and not in their relative share of the labor market.

Finally, whatever causal importance we assign to technological change, it is hard to see it as a credible account of the different trajectories of inequality across countries. Technological change is a challenge faced by all national economies, and the secular decline in labor’s share of national income is common to most advanced economies. And yet on key measures of inequality, differences across national settings (and especially the outlying status of the United States) remain profound. To illustrate this, Graph 4 charts the Gini index of inequality against a simple metric of educational achievement (the share of the population with a postsecondary education) for the United Statesand its peers (other wealthy democracies). In both 1985 and 2007, the United Statesleads the pack in both educational attainment and inequality.

Technological change, in short, falls flat as either a causal or cross-national explanation for American inequality. Indeed, as Thomas Piketty and Emmanuel Saez concluded recently, the very fact that wealthy nations with essentially similar histories of technological change show such divergent patterns of inequality suggest that institutional and policy differences—and not the underlying technological change—are the key. In this sense, the magnitude or pace of labor’s displacement is less important than the fact that, as Peter Frase notes, such displacement is now occurring “without many of the countervailing protections that labor enjoyed in the heyday of the postwar Keynesian compromise.”

In the United States, those differences—or those lost protections—are now pretty familiar: a long decline in union density and bargaining power, the retreat of basic labor standards and their enforcement, a fiscal policy that accommodates spells of high unemployment out of an irrational fear of inflation, and the evaporation of fiscal or regulatory restraint on incomes at the very top. The robots and computers had little to do with it.

Colin Gordon is a professor of history at the University of Iowa. He writes widely on the history of American public policy and is the author, most recently, of Growing Apart: A Political History of American Inequality.

Blind Injustice [Advertisement]