When Workers Lose to AI

Early last summer, buried deep in an obscure budget document, the Trump administration effectively pronounced dead the primary government program for workers who lose their jobs because of trade. The U.S. government had created Trade Adjustment Assistance, as the program is known, in 1962 under the Kennedy administration. Over six decades, it provided Americans in industries negatively affected by imports with cash assistance, job training, and other services. But President Donald Trump had long made clear that his only remedy for workers threatened by trade was new tariffs. By the time of its demise, the TAA program had few enthusiastic champions on either the left or the right.

But the TAA’s history is worth a closer look as the United States confronts another potential upheaval in the economy—the prospect of workers up and down the income ladder displaced by artificial intelligence. Similar to reduced trade barriers in previous eras, AI presents the possibility of faster economic growth that could benefit many while harming a few. The TAA’s creation reflected the fact that, at the dawn of globalization, lawmakers and experts took seriously the danger that free trade might hurt certain workers and communities, and they acknowledged both an economic and a political imperative to support those who lost out.

The TAA had an impeccable pedigree: it was first proposed by the United Steelworkers, drew on the ideas of leading economists, and was championed by John F. Kennedy as both a senator and president. And yet it fell short of its promise. Understanding why may be crucial to responding better to future displacement, especially as voices as varied as U.S. President Joe Biden’s Council of Economic Advisers, members of Congress on both sides of the aisle, and the AI giant Anthropic have cited the TAA as a potential model for dealing with the ramifications of AI adoption.

There is no shortage of dire predictions as to how AI could displace workers: business leaders have warned of unemployment as high as 20 percent within five years. But even if the most extreme fears prove wrong, AI could still upend the livelihoods of millions in ways that would be deeply painful for those affected and destabilizing for American politics. Job losses during the manufacturing decline of the 2000s, after all, affected only a small subset of the labor force, yet they carried profound economic and political consequences. Part of what makes AI’s rise so unnerving is that the range of its outcomes is so wide; the new technology could expose rideshare drivers, software engineers, and just about everyone else to potential displacement. Governments cannot afford to wait to act until the shape of that displacement becomes clear, because the institutions that help workers take time to build.

The TAA’s record demonstrates both how challenging it is to help workers displaced by economic change and why getting the response right the next time around is crucial. The United States will need programs that provide a financial cushion for those facing a sudden loss of income and help workers find new jobs. But such programs must be paired with efforts to prevent unnecessary displacement in the first place and to create good jobs for workers to move into when their old ones disappear. And unlike the TAA, they must be built to match the scale of any displacement, avoiding narrow eligibility criteria that would leave many workers in the cold. They must also operate with speed and give workers real agency in shaping their own economic future.

Getting the right labor policies in place now will help determine whether AI proves a blessing or a curse. At stake are not only the economic fortunes of millions of workers, but also the resilience of the United States’ social fabric in an age of rapid technological change.

A MIXED RECORD

The case for the TAA gained political momentum in the early 1960s because such an initiative seemed to answer several challenges at once. The Cold War strategy put forward by Kennedy and others called for new trade deals that would deepen the United States’ economic ties to the rest of the world, building a stronger front against the Soviet Union. American corporate leaders were eager to open new markets for their products. But advocates for free trade knew that workers would marshal fierce opposition if they stood to lose out when tariffs fell.

At the time, little targeted assistance existed for people and places caught up in difficult economic transitions. Federal aid for displaced workers was mostly limited to unemployment insurance; modern programs for job training and economic development were still in their infancy. The TAA was a seemingly elegant solution. It helped smooth the passage of major trade legislation that lowered tariff barriers, what became the landmark Trade Expansion Act of 1962, by allaying the concerns of workers, unions, and their champions in Congress. And it reflected a growing interest on the part of the Kennedy administration to support distressed communities and struggling workers.

And yet, it is hard to argue that the TAA achieved its mission. In the 1960s, the program’s opponents argued that it would inadvertently help the “wrong” workers—anyone who had lost a job for reasons other than import competition. The safeguards built to prevent that possibility ended up obstructing the very workers the TAA was intended to reach. For the program’s first seven years, the U.S. Tariff Commission, the body initially tasked with vetting claims, did not approve a single application. The statute establishing the TAA was overly strict, requiring workers to prove that rising imports from new trade agreements had caused their job loss, and the commission read it even more strictly.

Even after the TAA became more widely available beginning in the 1970s, its approval process remained needlessly complicated, typically requiring adjudicators to examine a firm’s records and even survey its customers to confirm that increased imports had caused the layoffs. If a worker’s job loss had multiple culprits—such as a firm losing market share to a domestic competitor even as it faced rising imports—the worker’s petition might be rejected. Many workers, especially those without a union, never even knew assistance was available. And the lengthy application process meant that even those the program ultimately helped often waited months or years between losing a job and seeing any aid.

Despite later attempts by several administrations to strengthen it, the TAA never shed its early reputation as a dud or expanded enough to matter when imports surged. Congress usually considered improving the program only when lawmakers sought a bargaining chip that would secure votes for deals that further lowered barriers to trade. The so-called China shock cost as many as two million manufacturing jobs between 2000 and 2007, but the TAA did little to help workers rebound. From 2004 to 2006, an average of only about 130,000 workers a year were eligible for assistance, and fewer than 50,000 a year enrolled in training.

Experts routinely panned the TAA as a failure, but the actual record is more complicated. Recent research suggests that workers who received assistance did benefit. One 2018 study estimated that those who went through the TAA-related training earned $50,000 more over the following decade than workers who did not. A wage insurance program added in 2002 to help older workers was found to pay for itself by hastening their return to the workforce; it covered part of the gap between their former earnings and the lower pay of a new job. And in places where the TAA reached workers, other research shows, it blunted the political backlash against trade; people in these areas were less inclined to support protectionist policies and take anti-incumbent positions.

But whatever benefits the program offered were sharply limited by its narrow scope: only a minority of trade-affected workers were ever certified as eligible, and only a minority of those participated. A program that helps the few who enroll while leaving out most of those it is meant to serve is not fulfilling its aims. And simply redistributing income to those who bear the brunt of change, whether from globalization or AI, is not enough if workers cannot find new high-quality jobs. At its worst, the TAA bred complacency, letting policymakers imagine that displaced workers had a cushion that wasn’t really there.

The TAA’s mixed record illustrates how hard it is to build and sustain such a backstop. The best alternative, of course, is to avoid painful job cuts to begin with. Companies can find new roles for employees rather than simply letting them go; one priority should be to invest in research on how AI can be better deployed by firms to augment workers rather than replace them. Greater bargaining power for workers, including a strongly protected right to organize, would push firms to seek alternatives to layoffs and guard against labor abuses that AI might enable.

MATCH THE SCALE

But if AI proves as transformative as both its proponents and its critics suggest, some workers will inevitably be displaced. Here, the TAA’s history offers lessons. It points to the value of a program strong enough to match the scale of disruption. Rather than limiting support to only those workers who can demonstrate exactly why their position disappeared, as the TAA did, a better approach would err on the side of being overly inclusive, extending comprehensive assistance to all displaced workers, no matter the cause of their job loss. Even setting aside the risks AI presents, stronger state support for all displaced workers, including more inclusive and comprehensive unemployment insurance, is long overdue. The case for intervention will only grow if AI proves to be a “general-purpose technology” capable of reshaping a broad array of occupations. Beyond just helping people who are struggling, such measures would have economic and fiscal benefits as well: helping displaced workers transition to new jobs reduces their need for long-term support and increases tax revenue.

Any AI displacement program must also move quickly. A large body of research shows that extended periods of unemployment leave lasting effects on earnings, health, and well-being, even after a worker is reemployed. Future adjustment efforts should operate against the clock, treating every additional day out of work as an institutional failure and offering support for those in at-risk industries or occupations before layoffs even happen. To that end, programs could grant “presumptive eligibility” for certain services even before layoffs occur, making it possible for significant numbers of workers to qualify with minimal paperwork. For example, if accountants or programmers appeared to be facing high levels of displacement, retraining could be made available to them even before they lost their jobs.

Finally, any future program must respect workers’ agency in shaping their own path forward. The TAA earned a reputation as a consolation prize forced on people who had no say in their economic fate. (Labor leaders often disparaged it as “burial insurance.”) The drawn-out, convoluted process left recipients with little sense of control. Giving workers a real voice, both in the workplace and in the design of efforts to support the displaced, would make such programs far more likely to succeed.

One model could be the “automation funds” that emerged through collective bargaining in the 1960s to cushion an earlier wave of technological change. These funds—pursued by meatpackers, longshoremen, steelworkers, and others—typically involved a partnership between labor and management to address the risks that automation posed both to workers who stayed on the job and to those who might be displaced. Some funds supported research into new approaches to job training. Others boosted severance pay or gave workers a share of the profits that automation generated.

These funds were only a little raft amid swirling economic currents, and their champions emphasized that they could succeed only if complemented by stronger public workforce programs such as effective training initiatives and job-search assistance, as well as macroeconomic policies supporting low unemployment. But they offer a template for how workers can be given a seat at the table during a turbulent economic transition. Companies so often deploy AI in ways that feel opaque and disempowering; addressing the technology’s consequences will require treating workers as humans, even when their previous jobs are no longer performed by people.

SHARING THE GAINS

The past few generations of trade and globalization should serve as a cautionary tale: even if only a small minority of workers lose out in a major economic shift, their losses can ripple outward, devastating not just those workers but their families, their communities, and the economy at large.

At a moment of heightened anxiety about what AI means for workers, the government should invest in funding financial support, retraining, and job-search assistance for those at risk of displacement. Such an effort would amount to a commitment to sharing the gains of technological progress. Alongside investments in good jobs and new labor protections for a transformed workplace, it would signal a vision of an AI future that does not consign people to a dimmer fate simply for having chosen the wrong profession at the dawn of a revolutionary technological breakthrough.

Repeating the errors of the TAA—or worse, doing nothing—would risk the same economic and political fallout that came with globalization, potentially on an even broader scale. Whether AI displacement follows the most worrisome predictions or the more modest ones, millions of workers could face diminished incomes, disconnection from the workforce, and lost status. The past quarter century has shown that sudden job loss—whether concentrated, as with the China shock, or widespread, as with the Great Recession—can have lasting repercussions for the country, including by fueling a backlash against democratic institutions and widening the audience for right-wing populism.

To get ahead of the problem, political leaders need to act ambitiously now, both because effective programs take time to build and because the public rightly senses that a sufficient safety net does not yet exist. The fact that earlier efforts largely failed should temper any expectation that the task ahead will be easy, but it should also spur policymakers to do better this time.

Loading...