The Long-Term Jobs and Wage Picture

“We are being afflicted with a new disease of which some readers may not yet have heard the name, but of which they will hear a great deal in the years to come – namely, technological unemployment. This means unemployment due to our discovery of means of economizing the use of labour outrunning the pace at which we can find new uses for labor.”

– John Maynard Keynes, Economic Possibilities for Our Grandchildren, 1930

There is a useful but neglected way of distinguishing a recession from a depression or major economic downturn. After a recession the economy goes back to the “normal” of the previous expansion, whereas after a severe downturn the economy is reconfigured in some significant respect. Nineteenth century American capitalism was in recession or depression almost as often as not, featuring three major depressions and a continuous series of bankruptcies in the period’s major industries, railroads and steel. After the dust had settled, capital had learned to preclude these kinds of ongoing crisis by consolidating, centralizing and mechanizing on an unprecedented scale and virtually banishing cutthroat competition. What emerged was what’s now called organized or oligopoly or monopoly capitalism. The Great Depression was followed by the next new configuration, Keynesian capitalism with its sundry forms of civilian and military government contribution to economic growth and income distribution. With respect to economic security, these were the best years ever for US workers. Alas, then came, in the 1970s, the elite revolt against the “welfare state”, with declining labor income, a corporate revenue squeeze, accelerating credit inflation, rising inequality and financialization.

That prelude to crisis eventually gave rise to the ensuing meltdown of 2007-2008 and the attempt by business and government to reconfigure the old Keynesian settlement into the shape of Microeconomics 101. The world after the current depression will neither look nor feel as it did in the lifetimes of those of us who grew up in the 1950s, ‘60s and ‘70s. Persistent inequality and lowered living standards will be among the highest costs of neoliberalism, a return to a peculiar fusion of key features of the capitalism of the 1920s (unregulated, free markets and dramatic inequality) and 1930s (high unemployment and lowered living standards).

Long-Term Unemployment, Job Polarization and the Disappearance of Middle-Skill Jobs

Neoliberalism intensifies a labor-market trend under way since the beginning of the postwar period, during which the short-term unemployed have been a shrinking percentage of all unemployed. Since the late 1960s long-term unemployment has been steadily rising. Looking at the business cycle over the last forty years, an ominous trend is evident: in each business-cyclical expansion, the long-term unemployment rate remains either at or above the level of the previous expansion. In a word, for the last forty years the short-term unemployed have been a declining, and the long-term unemployed an increasing, percentage of all unemployed. Most importantly for the purposes of this article, the persistence of unemployment is closely related to the disappearance of middle-pay jobs. The result has been that low paying jobs comprise an increasingly large percentage of all jobs.

Is there a structural (1) explanation of the disproportionate proliferation of low skill, low paying jobs? A key to an illuminating explanation is the remark, in the New York Times last summer that “The disappearance of midwage, midskill jobs is part of a longer-term trend that some refer to as a hollowing out of the work force…” (“Majority of Jobs Pay Low Wages”, Catherine Rampell, Aug. 30, 2012) A “hollowing out” implies a polarization, an emerging structure of inequality within the labor market. The job market is bifurcating into high skill,  high paying, advanced-education jobs at one extreme, and low skill, low paying, low education jobs at the other. Disappearing are occupations in the middle of the skill and pay distribution. Much research in recent years (2) throws light on this phenomenon and implicitly calls into question common explanations, that this job shedding is due largely to offshoring and outsourcing, that it is concentrated in manufacturing and that is the result of a mismatch between skills required by employers and the skill-level of job seekers. The citation from Keynes at the head of this article is closer to the truth.

In both the natural and the social sciences new insights are often the fruit of perspicuous categorization. It’s a certain type of job that is disappearing but the categories low skill, high skill, manual , cognitive, high paying, low paying fail to uncover the systemic mechanisms generating increasing labor market polarization. What is important is that it is routine jobs that are vanishing. These are jobs involving tasks consisting of a specific set of activities accomplished by workers following well defined instructions and procedures. These are not merely manual or “blue collar” jobs in production and maintenance like mechanics, machinery diagnostics, machine operators and tenders, meat processors, cement masons, dress makers, fabricators and assemblers. Routine occupations also involve “cognitive activities” in sales and “office and administrative support” such as secretaries, retail salespersons, some workers in law offices, bank tellers, travel agents, mail clerks and data entry keyers.

Vanishing Jobs, Automation, Robotization and Computerization

It is in these types of routine occupation that automation, robotics and the use of computers facilitate the replacement of human labor with machines. In some cases labor is entirely eliminated and all the work is done by machines or computers, but in the typical case technology reduces the demand for a portion of mid-skilled labor. Most of us are familiar with the replacement of bank tellers by ATMs, secretarial work replaced by personal computers and/or SIRI, Apple’s “intelligent personal assistant” integrated with the iPhone, self checkouts in grocery stores, self-service terminals in airports, video stores replaced by web-ordered DVD shipping and cable access and telephone customer service replaced by voice menus and web-based FAQs. An office in the 1980s employing 40 people working without computers may require, in the early 1990s, only 4 workers using 4 computers. The productivity  -output per unit of labor input-  of the office can be further enhanced not by adding skilled workers nor by replacing less productive workers with more productive computers, but by replacing less powerful computers and software with more powerful ones. In the initial case, actual workers were replaced by computers. In the latter case potential workers were kept out of the workplace by better computers. Thus the notable reduction in the demand for office workers.

Automation and robotics have had similar effects. The aggregate effects of all kinds of mechanization is felt across the entire labor market. In manufacturing the demand for machinists and machine operators is trending downward. Routine work in transportation and warehousing is also disappearing. All this is a portent of the broad range of jobs that are liable to permanent loss due to increasingly labor-saving advances in the development of the means of production. Right now this is most evident in the computer electronics industry. Robots do nearly all the work in making the most valuable part of computers, the motherboard, housing microprocessors and memory. Workers slip in the batteries and snap on the screen. A long-time analyst of the industry predicts that “[Robots} will replace most of the workers, though you will need a few people to manage the robots.” (Catherine Rampell, “When Cheap Foreign Labor Gets Less Cheap”, The New York Times, Dec. 7, 2012.)

Welding is virtually ubiquitous in widget production. Job loss in this occupation is rampant, due exclusively to the widening use of robotic arc welding to replace manual welding. Robots do the work in half the time it takes workers. The loading and unloading of machines has been made much more efficient by robots. A machine that is manually loaded “waits”, i.e., is unproductive, longer than one that is robotically loaded. A robot is faster than a human operator because it does not have to wait for a cutter or a part to stop moving, or for a door to open. Instead, the robot accesses parts through the top of a machine and unloads them immediately upon their completion. In one study, robotically loaded machines turned out 545,000 parts annually, while operators produced only 445,000. And robots don’t take lunch or other breaks, so the employer gets 8 hours of work out of them daily, not the 7.5 demanded by workers. And not a single robot has gone on strike.

Until very recently, most commentators have concurred that since human beings are not machines, there is a limit, set by factors such as pattern recognition and complex human communication, to how much human labor can be automated. There is surely such a limit, but it is not as insurmountable as many of us have thought. Producers of capital goods seek ever greater possibilities of reducing the human contribution to productivity, and successful experiments in the last two years reveal the spectacular potential of digital technologies and the rapidity with which advances are achieved.

In 2004, a driverless vehicle was built designed to navigate a 150-mile route through the Mohave Desert. No people, no structures, just sand and space. The vehicle failed before 8 miles and took hours to traverse that short distance. In 2010 Google succeeded in fully automating a fleet of Toyota Priuses. These cars navigated over 1,000 miles of road with no human intervention at all, and over 140,000 miles with only minor human involvement. Driving in traffic had long been cited as a paradigm case of a task requiring complex pattern-recognition skills not amenable to digitalization.

A classic case of complex communication is translating from one human language to another, requiring as it does a degree of emotional sensitivity and the capacity to deal with ambiguity. However, language used for common business services is less reliant on these abilities. The translation services company Geofluent in partnership with IBM sought to provide automatic translation good enough for business purposes. They developed techology capable of translating online chat messages sent by Spanish and Chinese customers to English-speaking employees. 90 percent of the senders found the translations useful.

Was it possible to pose even more demanding challeges to the machine, combining the previous accomplishments in pattern recognition and complex human communication? A supercomputer called Watson was designed to provide the questions yielding the answers posed to the computer. I.e., Watson was asked to play the popular game Jeopardy. The machine executes a remarkably complex series of searches and comes up with candidate answers which are in turn subject to a series of answer-scoring analyses. This is done with such speed and accuracy that the two most accomplished contestants in the television show’s history were defeated by the machine in a televised contest.

These examples illustrate not merely how easy it is to replace routine workers with machines, but also how non-routine work hitherto relatively unaffected by computers, automation and robotics may now be reduced or done away with through technological advance. These are the causal factors accounting for the job polarization in evidence since the early 1980s. A recent study by the Federal Reserve Bank of St. Louis (3) reveals that employment has been increasingly concentrated in the highest- and lowest-paying occupations, with middle-skill and middle-pay jobs steadily disappearing. Just about all of these vanishing jobs involve routine mid-skill tasks, many of which are increasingly done by machines.

Since the 1970s the percentage of workers performing routine manual and cognitive tasks declined, not only in the US but in Europe as well, and the proportion doing nonroutine jobs rose. These shifts in labor input were not evident in the precomputer decade of the 1960s. But in each subsequent decade the shift accelerated. In 1984 routine workk accounted for 54.6 percent of all employment, in 2011 for 44.0%. As a share of the total labor force, it has fallen from 50.4 percent in 2000 to 44.6% in 2011.

Jobless Recoveries and the Disappearance Of Routine Jobs

We have in recent years been introduced to the cynical notion of the “jobless recovery.” For most of US economic history this term would have been dismissed as self-contradictory. That it is now part of common economic discourse is testimony to a major conceptual revision in the discourse of propaganda: that the economy is recovering is no reason to expect unemployed workers to find work. Economic recovery is now treated as consistent with declining standards of living. Lowered expectations and acquiescence in long term working-class hardship are now built into what we are told to regard as recovery. This political-economic innovation demands closer scrutiny. We want to know why recoveries since 1990 have been jobless, and what it is that makes them jobless. This will give us a clear picture of exactly what is happening in the “jobless recovery” that distinguishes it from the normal postwar cyclical recovery.

The key lies in the greatly heightened importance of a particular kind of unemployment, referred to by Keynes in the citation above as “technological unemployment”, and correlative to the advance of mechanization described above. This is not the kind of unemployment that attends a garden-variety recession, which disappears as the economy recovers. Peter S. Goodman correctly projected in The New York Times that the recovery following the 2009 recession would not bring sufficient jobs to absorb the record-setting ranks of the long-term unemployed. (“The New Poor: Millions of Unemployed Face Years Without Jobs”, February 21, 2010) He describes the new poor as “people long accustomed to the comforts of middle-class life who are now relying on public assistance for the first time in their lives – potentially for years to come.” What is distinctive about the jobless recovery?

Let’s look at the last 6 recoveries  -after the recessions of 1970, 1975, 1982, 1991, 2001 and 2009- and compare the jobs lost during the downturns with those restored in the subsequent recovery. After the recessions of 1970, 1975 and 1982 both production and employment recovered. The jobs lost during the recession, including routine jobs, were regained in the recovery. Routine jobs were the largest single category of work in this period. It is the disappearance of precisely these jobs that distinguish the recessions of 1991, 2001 and 2009 from previous recessions. By the time of these recessions, routine jobs were more than 50 percent of all jobs and accounted for virtually all the job loss. Most importantly, this type of employment never recovers beyond its trough peak, nor does it approach its pre-recession peak. The permanent decline of middle-skill employment as a proportion of all employment has occurred nearly every year since 1984. The 1991, 2001 and 2009 recessions were the first to exhibit jobless recovery. The jobless recovery, then, is due to the disappearance of routine work, or, alternatively, to the polarization of the job market during these years.

What has accounted for the loss of these jobs? There seems to be an erroneous consensus on the Left that offshoring/outsourcing explains this phenomenon. About a third of all manufacturing work, some 6 million jobs, has been lost since 2000. But the exporting of jobs fails to explain most of this. “[W]hile many of these jobs were lost to competition with low-wage countries, even more vanished because of computer-driven machinery that can do the work of 10, or in some cases, 100 workers.” (Adam Davidson, “Skills Don’t Pay the Bills”, The New York Times, Nov. 20, 2012) This is permanent job loss, and contributes to the inequality endemic to labor-market polarization: “Those jobs are not coming back, but many believe that the industry’s future (and, to some extent, the future of the American economy) lies in training a new generation for highly skilled manufacturing jobs – the ones that require people who know how to run the computer that runs the machine.”

The Times article does not ask what will happen to the millions of workers left over after the far fewer new skilled workers have been found. If the “jobs are not coming back”, there will be workers, lots of workers, whose only recourse will be long-term unemployment or low-skill, low-pay work. There you have it – neoliberal austerity for the masses. That’s the long-run prospect.

Investment, The Business Cycle And The Loss Of Routine Work

Some researchers have noted that the accelerating loss of routine jobs occurs exclusively during the downward phase of the business cycle, concluding that these job losses are essentially, but not exclusively, a business-cyclical phenomenon. The tendency of mainstream economists to associate unemployment trends with the cycle is probably what accounts for this odd observation. The loss of routine work is ongoing and is accounted for by causal factors independent of the economy’s cyclical physiology. Unemployment does indeed increase in a downturn, but this is a tautological observation. If a cyclical grid is superimposed on any factor that is trending downward, a ratchet-like pattern will of course be observed: you will see bursts alternating with stabilization. But what happens when you remove the grid? You see a secular downward trend in routine employment not at all peculiar to recessions.

This is not to say that downturns are irrelevant to understanding the vanishing of routine jobs. But we need to look not at the alternating pace of job loss, but at the pattern of investment that is associated with economic contraction. There are two types of investment, capital-widening  and capital-deepening. The former consists of additions to the stock of existing equipment in order to expand production, and tends to be concentrated in periods of expansion. The latter involves investing in new, more efficient equipment in order sometimes to expand production but always to enhance productivity, and occurs mainly in economic downturns. During downturns sales revenues decline so that maintaining profits becomes primarily a matter of cutting unit costs. That’s precisely the point of introducing more productive equipment (and speeding up the labor process) like better computers and other types of labor-displacing or labor-complementing machinery. Silicon based employees are substituted for carbon based employees. This is why we expect an acceleration of routine-job loss during contractions.

Dispelling Myths About Offshoring and Outsourcing

I’ve noted above that offshoring and outsourcing is not the major cause of job loss in recent years. In fact, a growing number of US manufacturing companies are “re-shoring” jobs they had previously sent to lower-wage countries. It appears that US companies had overlooked some of the significant costs of overseas production. The shipping process has been found to be especially problematic. The widget must be shipped to the Chinese port, loaded, unloaded in the US and shipped to its final destination. This can take 4 to 6 weeks. This time-cost can be considerably increased, the companies learned, by things like the 2002 West Coast dock strike. Homeland security complications have further lengthened shipping schedules.

Companies discover too frequently that the product, once it is on the ocean, fails to meet standards and needs to be re-worked. (Chinese workers are often paid based on the number of units completed, so a finished unit is a good unit.) These products cannot just be shipped back. Fees must be paid at both ports. There is the additional cost of exporting raw materials not available overseas from the US to the point of production. Travel by representatives of US compoanies can be lengthy and expensive. And there are intangible costs: counterfeiting of intellectual property and unpredictable currency fluctuations. The whole business has amounted to an apparently unanticipated and ungainly cost. One study found that these tangible and intangible costs can amount to as much as 24 percent of total product cost.

Combine these considerations with the narrowing of the wage gap between e.g. China and the US (Inflation-adjusted average wages in China have almost tripled since 2000, whereas median household income in the US has declined over the same period), and the incentive to produce overseas diminishes. Further reducing the disadvantage to employers of higher US wages is that a high proportion of the work that has been brought back to the US is the kind of high-value-added work tied to automated production. Relatively higher wages don’t matter much if workers aren’t required anyway.

The complaint that China has “stolen jobs from the US” is highly misleading. Over the period 1995-2002, China lost 15 million manufacturing jobs, the US lost 2 million and the whole world lost 22 million manufacturing jobs. The great majority of these jobs were lost to automation and other productivity-enhancing innovations. As one leading researcher points out, “Manufacturing will go to the countries whose companies win the race to automate…” (Rick Schneider, “Robotic Automation Can Cut Costs”, Manufacturing Engineering, November, 2005) International automation competition is set to replace wage competition. As I write, the developing countries are automating at a hasty pace as wages rise there.

It is clear that the displacement of labor by machines is a long-term tendency of capitalist development which will be apparent on a global scale. Keynes argued that the disappearance of private sources of employment should be offset by the expansion of public employment in public works programs. That’s much of what Keynes’s insistence upon the “socialization of investment” is about. It is remarkable that J-B Say, the arch-target of Keynes’s critique of the conceit that capitalist markets tend toward equilibrium (Say’s Law: “Supply creates its own demand”), should have advocated public works as the rational response to technology-related unemployment. In A Treatise on Political Economy (1832) Say wrote that “… a benevolent administration can make prevision for the employment of supplanted or inactive labor in the construction of works of public utility at the public expense as of canals, roads, churches or the like…” How about that? Yes, a “benevolent administration” could do just that, but…

The Unfolding Logic of Capital

At bottom, we are looking at the long-term logic of capital. Under capitalism means of production, capital goods, count as costs of production. There is system-endogenous pressure to produce capital goods that are both cheaper to purchase and more efficient. Commentators tend to ignore the former feature of capital innovations. But this feature is no less important than greater efficiency in explaining employers’ attraction to digital technologies and robotics. The price of ever-more-efficient computers has fallen precipitously over the years. And robots are made by robots, a cost-reducing leap forward. Capital’s motivation to depend more on computers and robots and less on human labor is irresistible. It’s no news that workers are the major pain in capital’s ass. They strike, slow down, get sick, demand a greater share of what would otherwise belong to capital, stubbornly insist on lunch breaks – it never ends. Capital’s wettest dream has been to be able to make money by making money, without that bothersome business of production, scrambling to sell the product and dealing with recalcitrant workers. Financialization has made part of that dream come true. Robots and computers could take care of the rest.

The historical tendency of capital to abhor the human contribution to production and productivity goes way, way back. There is a continuity from craftsmen making an elegant cabinet with hand-scrolled legs to Ikea. The replacement of the spindle and the distaff by the spinning wheel long predates the Industrial Revolution. Increased mechanization may have the same effect on industry and services that farming technology had on agriculture. In 1900 agricultural workers comprised over 38 percent of US employment. Today, they make up 2 percent of the work force and produce more of the world’s food than ever.

We have reached the point at which labor is becoming increasingly obsolete for the purposes of reaping profit. The production of both goods and services is rapidly becoming more capital intensive. Paul Krugman has pointed out, in a recent column on the “notable shift in income away from labor”, that automation intensifies the tendency to inequality inherent in neoliberal capitalism: ‘It’s “capital-biased technological change” which tends to shift the distribution of income away from workers to the owners of capital.’ (“Rise of the Robots”, The New York Times. Dec. 8, 2012) Krugman confesses that he had until recently overlooked the inequality between capital and labor by having focused instead on “major changes in income distribution… among workers (when you include hedge fund managers and CEOs among the workers)” You read that right: “when you include hedge fund managers and CEOs among the workers”! These are the lengths to which permissible thinking goes in order to head off talk of class conflict. It’s a form of intellectual self-censorship; that’s how political self-deception works.

Krugman further ‘fesses up that his current view “has echoes of old-fashioned Marxism – which shouldn’t be a reason to ignore facts, but too often is.” He should know. Shouldn’t Krugman conclude that Marxism isn’t so “old-fashioned” after all?

Marxian analysis is especially well equipped to size up the developments discussed in this article in a particularly fresh and creative way. The continuously astonishing way in which capitalism increases the productivity of labor, and so makes labor decreasingly necessary in order to satisfy the material requirements of life, is an adumbration of what human beings can become. Technological unemployment needn’t be catastrophic – but it will be under capitalism. In a different economic-systemic context, the obsolescence of labor can be emancipatory, the creation of free time to enable us to do what only the human species can – develop a broad range of capacities and pursue a range of satisfactions available only to humans: to be a sculptor in the morning, a philosopher in the afternoon and a musician in the evening (to paraphrase the Old Man before he was old). How odd that we should have these capacities only to live under social arrangements that preclude their realization. Under present political-economic circumstances the capacities that partially define human nature itself must lie dormant. It’s downright unnatural. Both Aristotle and Kant would have concurred with Marx that this kind of repression must damage our souls. Socialist aspirations, then, are by no means utopian, a mere “better idea”. Capitalism itself puts them on history’s agenda and reveals them to be necessary if we are to be in the end fully human.

Alan Nasser is Professor emeritus of Political Economy and Philosophy at The Evergreen State College. This article is adapted from his book, The “New Normal”: Persistent Austerity, Declining Democracy and the Globalization of Resistance will be published by Pluto Press in 2013. If you would like to be notified when the book is released, please send a request to Follow me on Twitter:, and Like me on Facebook:


(1) I do not use ‘structural’ here in the currently most common use of the term in connection with unemployment, connoting a mismatch between needed skills and those actually offered by job seekers. I use the term to refer to capitalism’s endogenous systemic dynamics.

(2) See Jaimovich, Nir and Siu, Henry E., “The Trend is the Cycle: Job Polarization and Jobless Recoveries”, National Bureau of Economic Research Working Paper No. 1, August 14, 2012; Autor D.H. and Dorn, D., “The Growth of Low Skill Service Jobs and the Polarization of the US Labor Market”, American Economic Review, forthcoming; Autor, D.H., Levy, F., and Murnane, R.J., “The Skill Content of Recent Technological Change: An Empirical Exploration”, Quarterly Journal of Economics  118 (4); Autor D.H., Katz, L.F., Kearney, M.S., “The Polarization of the US Labor Market”, American Economic Review: Papers & Proceedings, 2006, (96)2; National Employment Law Project, “The Low Wage Recovery and Growing Inequality”, Data Brief, August 2012; Brynjolfsson, Erik and McAfee, Andrew, Race Against the Machine, Digital Frontier Press, 2011


Alan Nasser is professor emeritus of Political Economy and Philosophy at The Evergreen State College. His website is:  His latest book is Overripe Economy: American Capitalism and the Crisis of Democracy. He can be reached at: