Cynthia Estlund is the Catherine A. Rein Professor at the New York University School of Law, and a longtime teacher and scholar of labor and employment law.  This post is the first in a three part series: parts two and three are available here and here. 

The labor world took notice when Andy Stern emerged from a years-long deep dive into the future of work, and concluded that the future will bring a lot less work.  His book, Raising the Floor, helped to spur a debate over the universal basic income (UBI), including on this blog.  But the underlying issue of technology-related job loss has not yet engaged the close attention of labor and employment law scholars.  That should change.  Even more than firms’ flight from direct employment through fissuring, their replacement of human labor with ever more capable and cost-effective technology threatens the foundations of economic and social life, and calls for a reexamination of prevailing approaches to regulation of employment.

I recently posted a paper on SSRN about the problem of automation and job loss.  Here I will take up three of its points in highly condensed form.  This post will briefly review the debate on automation and job loss.  The second post will discuss the role of labor costs, including those attributable to labor and employment law, in motivating firms both to automate and to contract out labor needs through “fissuring.”  The third post will ask what can be done within the boundaries of labor and employment law to address the risk of impending job losses, considering the high degree of uncertainty about that risk.

The Debate over Automation and Jobs

The forecast that machines will supplant human labor is a recurring theme in the history of industrialization.  Time and again, however, the destruction of some jobs has been offset by the creation of other jobs, usually better-paid and less grueling.  The history of automation’s impact on the labor market has been one of “creative destruction,” a mantra to which many economists adhere today.  For those who are confident that past trends will continue, the current wave of anxiety about automation amounts to scare-mongering by “neo-Luddites.”

Those who predict that the coming wave of automation will be different point to the distinct nature of current technological advances.  Hard and soft forms of technology – robots and algorithms, e.g. – are replicating a wider range of human capabilities, and integrating them more seamlessly, than ever before.  The terms “artificial intelligence” and “machine learning” hint at what is new:  Technology is acquiring and refining cognitive capabilities that were thought to be uniquely human, and is outpacing humans at increasingly complex tasks.

For example, most readers know that computers have beat human champions at two notoriously complex board games:  Chess and Go.  Another telling advance occurred recently in the field of translation:  In November 2016, Google launched a new version of Google Translate that instantly and dramatically improved the quality of translations.  Ironically, just four months earlier, then-Chair of Obama’s Council of Economic Advisors had used translation to illustrate the enduring advantages of humans over technology:  “AI today … cannot come close to what a human can do with his or her knowledge of both languages, social and cultural context, and sense of the author’s argument, emotional states, and intentions.”  But it has suddenly come a lot closer – and, unlike humans, it does the job instantaneously and for free.

Such developments have led to a spate of efforts to measure the extent of likely job losses.  One much-cited Oxford University study estimated that nearly half the jobs in the current economy are at risk.  A 2017 report by a team at McKinsey Global Institute (MGI) came to a similar conclusion using more granular analyses of both what current technology can do and what activities humans are currently paid to do.  They estimated that 46 percent of all of the time for which people are now paid in the U.S. economy is spent in activities that could be automated based on currently available technology.  Notably, that estimate excludes the impact of future advances, which are galloping along at a pace that surprises even savvy Silicon Valley observers.

Of course, technical automatability is only a threshold factor in firms’ decisions about automation.  It will take time and managerial skill to disaggregate automatable tasks from those that humans still do better, and to reconfigure jobs and organizations.  It will also require some highly skilled workers, currently in short supply, to implement and work with the new technology.  Given these hurdles, the MGI study estimates that it may take two to four decades to move from technical automatability to large-scale implementation.

The other side of the equation in these debates is new job creation.  The MGI study, for its part, makes an explicit assumption that “human labor displaced by automation would rejoin the workforce and be as productive as it was in 2014, that is, new demand for labor will be created.”  Strikingly, however, the study makes no effort to identify the new demand or new jobs that will absorb displaced workers.  There may well be no way to do so.  But that colossal leap of faith stands in marked contrast to the meticulous analysis behind estimates of likely job losses.

The concern is not only net job loss but growing wage polarization between those who produce or own the new technology, or whose high-end skills are complimented by that technology, and the masses who are stuck competing, and driving down wages, for the jobs that remain.  Labor shortages in some skilled job categories will coexist with a surplus of labor and downward pressure on wages outside those categories.

Rapidly evolving technology helps drive a high-stakes tournament in which firms will lose market share and profits if they continue to employ people to do things that machines can do more efficiently.  People will lose out if they fail to acquire hard-to-automate in-demand skills, or if they lack the resources and opportunities needed to acquire and constantly update those skills, or if they crumple under the pressure of the tournament itself.  To thrive in this economy will require above-average levels of intelligence, psychological resilience, and tolerance for risk and change.  Pity those who are not blessed by both nature and nurture with the makings of those traits.  Pity, too, those who seek a more balanced life – working to live rather than living to work.

If new jobs and sectors do arise to absorb displaced workers, as many economists expect, we will still need much better institutions of education and training and a stronger social safety net to enable individuals to make the necessary transitions.  If the economists are too optimistic in their faith-based assumptions about new job creation, or if job destruction outpaces current estimates due to advances in artificial intelligence and machine learning, then we face a genuinely bleak prospect in a decade or two:  As highly automatable mid-level jobs disappear, many more people will be competing for jobs that are not automatable but require no special human skills.  The more rapidly these transformations play out, the more wrenching the social consequences will be.

The question is how to approach these issues in the face of uncertainty – specifically, in the face of an uncertain but significant probability of serious net loss of jobs and growing disparities in labor market outcomes within a decade or two.  Ideally we should attempt to chart a path that that will lead in the right direction – spreading the benefits and mitigating the costs of automation – whether or not the most alarming predictions are borne out.  But first we should explore the role of labor and employment law in these transformations, as I will do briefly in the next post.