Key takeaways

• Anticipating that trade and globalization could displace workers, lawmakers and experts established Trade Adjustment Assistance more than 60 years ago in an effort to address that potential harm. Yet the program did not live up to its intentions, failing to reach many of the workers it was designed to assist—with significant economic and political consequences.
• The history of the program offers lessons for policymakers when addressing future job displacement, particularly job loss that might be caused by artificial intelligence. Along with taking steps to limit the magnitude and consequences of displacement in the first place, these lessons include ensuring policy solutions are deployed that match the scale of any disruption, making speed paramount in the delivery of any assistance, and supporting the ability of individual workers affected by AI displacement to chart their own futures amid job dislocations.
• Policymakers should also focus on developing high-quality training models that prepare workers for in-demand jobs, compensation and adjustment efforts that go hand-in-hand, better data tools to help policymakers and analysts understand where displacement is occurring and how the job market might evolve, and appropriate place-based policies even if job loss is geographically dispersed.

Overview

More than 60 years ago, U.S. policymakers, economists, and labor leaders alike agreed that the potential for new economic forces to strengthen U.S. prosperity and security was just over the horizon. But the risk, as they understood it, was that even amid the possibility of higher growth that might benefit the nation as a whole, some workers might face severe harm. Whether out of moral obligation or to prevent the risk of political backlash, the question then was how to help these workers successfully adjust to a new economic normal. 

This question was a central one in the debate over how to expand U.S. global trade in the 1950s and 1960s. That debate over how to balance the gains and losses that come from global trade rages on today. But squint, and a similar specter looms over new advances in artificial intelligence, which pose both hopes for massive gains in productivity and fears about the potential to replace a broad array of blue- and white-collar jobs. While emerging research has offered inconclusive—and, at times, contradictory—assessments of the potential extent of labor market dislocation from AI, even the possibility of significant displacement has sparked new debates about how it might reshape U.S. politics and how policymakers might respond.

Often lost in today’s discussion, however, is that even though politicians, experts, and the general public over the past century did not fully anticipate the nature or extent of previous economic dislocations, they were nonetheless engaged in conversations similar to those happening today. Those conversations acknowledged the risks that forces such as increased global trade, technological advancement, or the energy transition could lead to greater overall prosperity but also result in significant harm for a sizable minority of people and places. The question is why, despite this awareness, the efforts that came out of those conversations turned out to be largely unsuccessful—especially with respect to trade and globalization.

Among workers and in communities most affected by an upsurge in imports earlier this century, the economic and political consequences are still being felt today, as economists David Autor at the Massachusetts Institute of Technology, David Dorn at Zurich University, and Gordon Hanson at Harvard University spell out in their contribution to Equitable Growth’s economic populism series.¹ As noted elsewhere in that series, an emerging research base provides evidence that a failure to effectively respond to economic disruption—and perceived threats of displacement—fueled the rise of right-wing populism both in the United States and in other advanced economies.

This paper explores one notable effort to address displacement: Trade Adjustment Assistance. While policymakers have pursued efforts to help workers facing targeted displacement due to a variety of causes—from agricultural dislocation during the Great Depression of the 1930s to the impact of environmental regulations on coal miners and loggers since the 1990s—Trade Adjustment Assistance is the policy with the longest, and arguably most studied, history.

After a discussion around why “adjustment” programs exist, this paper first describes Trade Adjustment Assistance’s history and the research base around its impacts. It explains how the program’s implementation did not live up to its intentions, failing to reach many of the workers it was designed to assist. While recent studies find that the program may have had greater success than previously appreciated in aiding the minority of workers who did receive assistance, its inability to reach the workers and communities it was meant to support sharply limited its effectiveness.

Drawing from Trade Adjustment Assistance’s history, the paper then identifies a set of lessons for policymakers seeking to address future displacement, particularly as a result of AI. It notes that a first task is trying to limit harm in the first place, creating the conditions—by, for example, putting a thumb on the scale for AI investments that augment rather than replace workers, strengthening worker bargaining power, and ensuring a more robust safety net—that reduce both the likelihood and the consequences of job losses. But given that some displacement will likely still occur, the paper lays out three key principles in designing future programs to aid those who are harmed. These principles include:

• Ensuring policymakers fully commit to deploying a policy solution that matches the scale of any disruption, without viewing worker assistance as a bargaining chip for other policy aims
• Making speed paramount in the delivery of any assistance, avoiding complex approval processes that delay workers from receiving aid at the moment it is most necessary
• Supporting the ability of individual workers affected by AI displacement to have agency in charting their future, even amid the dislocation job loss can cause

In addition, the paper identifies specific design considerations that should be applied to future adjustment assistance programs, including:

• Using and updating high-quality training models that prepare workers for in-demand jobs
• Implementing compensation and adjustment efforts that go hand in hand
• Developing better data tools that help policymakers and analysts understand where displacement is occurring and how the job market might evolve
• Exploring place-based solutions even if dislocation due to AI is less geographically concentrated than in the past

The potential scale and scope of future labor market displacements as a result of AI remains highly uncertain, but the possibility of major disruption argues for taking steps now to develop measures to assist those who are harmed. Doing so effectively will require learning from past efforts, including those that did not live up to their promises. This paper—in advance of further policy recommendations that Equitable Growth intends to propose in the coming months—offers some of those lessons.

What are policymakers trying to solve with efforts to help workers adjust to economic displacement?

Policies meant to help displaced workers adjust to economic change are typically taken from the broader toolkit of programs intended to support prosperity, mobility, and growth. What distinguishes them is that they are designed to address trade-offs that come with bigger economic, technological, and societal changes—usually changes that policymakers otherwise believe have broad-based benefits or, at a minimum, cannot be stopped without causing much greater damage.

Efforts to support displaced workers have sometimes been denigrated as attempting to “compensate the losers,” but a key element of these policies is that they are, in fact, responding to an experience of loss. They reflect an intuitive idea borne out in both economic and political science research: There is a unique harm attached to the experience of having attained something—a job, a certain level of income, a particular status in the community—and then losing it.² In this sense, policies addressing economic displacement can be distinguished from other critically important policies designed to support those historically shut out from economic opportunity—those who, in other words, were denied the chance to have as much to lose in the first place.

Moreover, these adjustment policies are most frequently intended to address losses that can be described as significant, concentrated, and persistent. Specifically, these losses are:

• Significant, meaning the harm they cause is substantial, often coming in the form of unemployment, sharply reduced incomes, or business closures, and often resulting in negative externalities that extend to others in local communities
• Concentrated, meaning the loss is experienced by a specific, discrete population that can be identified, whether that population is defined by industry, occupation, or place
• Persistent, meaning that the economic displacement in question is not merely a temporary phenomenon overcome after a brief transition but, instead, is a more lasting experience

These factors set economic adjustment programs apart from either ongoing poverty-fighting policies or responses to cyclical unemployment caused by economic downturns.

In addition, economic adjustment policies are not designed merely to compensate workers or communities for what they have lost, but also to help them—as the name suggests—adapt to a changed economy. These adjustments can be quite difficult for workers not only due to the challenges inherent in switching to new occupations or industries, but also because the previous status quo may have been relatively favorable prior to dislocation.³ With that in mind, these policies have often been pursued to manage the political backlash that comes with displacement, as well as to address material harm.

As discussed further below, efforts to help workers and communities adjust to economic change are not implemented in a vacuum. These efforts—which are, at their core, redistributive—are often presented in contrast to “pre-distributional” policies that shape how the gains from economic growth accrue in the first place.⁴ But it is worth considering how pre-distributional policies and adjustment assistance efforts might work in tandem—both to limit the scale of dislocation to begin with and because assistance may be more successful when workers have stronger bargaining power.

Likewise, adjustment assistance efforts are shaped by the presence—or lack thereof—of strong social insurance programs. Economic displacement, for example, has higher stakes in a society where a worker’s health care coverage is contingent on their employment, as has been and still largely is the case in the United States. Similarly, more robust Unemployment Insurance for all workers creates a stronger floor for those who are displaced by more unique circumstances.

All of these policy remedies have long had merit even absent an elevated risk of worker dislocation, and there is every reason to imagine they—and more novel approaches that prevent displacement and reduce its consequences—may become more necessary. The flipside is that even as policymakers seek more universal reforms to lower the need for adjustment assistance measures, it is likely that some form of policy in this space will still be necessary, given the difficulty in preventing all displacement and the unique harm it causes.

A history of the Trade Adjustment Assistance program

As early as 1939, the economist Nicholas Kaldor wrote that it would be possible to tax those who benefitted from reduced tariffs and use the proceeds to compensate others who lost income. That could, as he described, ultimately lead to higher overall economic welfare, while leaving no individual worse off.⁵ But policies designed to aid workers, firms, and communities affected by surging imports gained prominence after World War II, as reducing barriers to trade became a central plank of U.S. Cold War foreign policy.

The first major proposal to assist those who lost their jobs due to trade was offered in 1954 by United Steelworkers President David McDonald, who was an appointed member of a commission on international economic policy created by President Dwight D. Eisenhower.⁶ The commission rejected McDonald’s proposal by a 16-1 vote, but the idea quickly built political momentum. Less than 2 years into his first term in the U.S. Senate, John F. Kennedy introduced a similar proposal as its own bill.⁷

After he was elected U.S. president in 1960, Kennedy made an early push to reduce tariff barriers. He had proposed the creation of an adjustment assistance program as a crucial component of his presidential campaign, helping to secure the American Federation of Labor’s endorsement. Arguing that “those injured by [increased foreign] competition should not be required to bear the full brunt of the impact,” President Kennedy proposed a program to help both workers and firms adjust, explaining that “the accent is on ‘adjustment’ more than ‘assistance.’”⁸

Under the Trade Expansion Act of 1962, workers who could demonstrate that increased imports caused by trade agreements were the major cause of their job loss were eligible for Trade Readjustment Allowances, which, combined with Unemployment Insurance, could pay up to 75 percent of a worker’s previous wage for a year, with another 26 weeks of allowances granted for those participating in free training programs.⁹ These workers could also apply for relocation payments. Trade-affected firms were eligible for loans, tax breaks, and technical assistance in adjusting to new market conditions.

Yet in the first 6-plus years following the program’s passage, not a single application for assistance was approved out of 26 filed. The applications were rejected as a result of overly strict legislative language that had been tightened during congressional negotiations, even stricter interpretation of that language by the tariff commission responsible for implementing it, and seeming neglect on the part of the White House under presidents Kennedy and Lyndon B. Johnson.¹⁰ It was not until November 1969 that the first petitions were certified.

Those early years of Trade Adjustment Assistance were a harbinger of things to come, with occasional efforts at expansion frustrated by both intentional attempts to undermine the program and extended periods of near-fatal inaction. After 1969, the tariff commission took a more liberal approach in its consideration of petitions, which increased the number of applications and approvals. But the program remained on shaky ground, as President Richard Nixon in 1973 proposed to shrink it and eventually phase it out as part of an effort to offer more generous unemployment benefits to all workers. The AFL-CIO also came out against the program temporarily as part of broader concerns over trade policy.¹¹

Yet as part of negotiations for the Trade Act of 1974, which created the modern “fast-track” process for congressional consideration of trade deals, Congress again attempted to make Trade Adjustment Assistance work. Alongside an increase in the size of benefits for workers, the revised program had a significantly loosened standard for eligibility, and adjudication had to occur within 60 days.¹² These changes resulted in a large increase in the number of workers who received assistance, with 1.3 million certified between 1975 and 1981.¹³

That increase was short-lived. Following a massive spike in applications in 1980, when nearly 700,000 were approved, President Ronald Reagan proposed, and Congress approved, a reduction in allowances in 1981. The Reagan administration adopted a harder line in adjudicating claims, as the approval rate shrunk from 65 percent in the late 1970s and 81 percent in 1980 to below 15 percent in 1981 and 1982, and proposed to end the program altogether.¹⁴

As a result, far fewer workers applied, which one expert attributed to a potential “widespread misapprehension that the program was abolished.”¹⁵ A long-term reauthorization in 1986 stabilized the program, and approvals increased in the latter years of the Reagan administration, though still far below where they had been before 1981.

In the 1990s and 2000s, the program’s fortunes were once again connected to broader debates over trade policy. As part of the North American Free Trade Agreement, TAA coverage was extended to cover workers who lost their jobs because their plants moved to Mexico or Canada, in addition to those affected directly by higher levels of imports.¹⁶ The passage of the 2002 trade bill, which followed China’s entrance into the World Trade Organization, produced another debate over the future of Trade Adjustment Assistance, with Democrats insisting on expanding the program to provide votes for an extension of trade negotiation authorities.¹⁷

That 5-year extension of the program to 2007 included the creation of a tax credit to cover health insurance for workers eligible for Trade Adjustment Assistance and a new program for farmers affected by trade. The extension also included an expansion of coverage to workers who produced goods that were components of finished products affected by imports.

By 2009, in response to the Great Recession of 2007–2009 and the impact of increased Chinese imports, pressure for a greater rethink of Trade Adjustment Assistance emerged again. As part of the American Recovery and Reinvestment Act of 2009, President Barack Obama signed into law the most significant expansions of the program since the 1970s. The expanded program provided allowances for up to 2.5 years for those engaged in training, increased the size of the health coverage tax credit and job search and relocation allowances, and made service workers eligible for the first time. The law also created a new wage-insurance program for workers older than 50 that covered up to $10,000 of the gap between what they might earn in a new job and what they had previously taken in.

A combination of this expansion and the Great Recession resulted in about 280,000 workers being certified in fiscal year 2010—a post-1980 high water mark for the program. But it did not reflect new stability for Trade Adjustment Assistance. The program was extended in 2015 as part of a broader continuation of trade authorities but soon became wrapped up in new fractures over trade policy—first over the potential passage of the Trans-Pacific Partnership, and then amid President Donald Trump’s reshaping of U.S. trade policy during his first term—making it impossible for any real effort to either build on or reform the program to take hold.

President Joe Biden sought to extend Trade Adjustment Assistance once again as it was set to expire, but due to congressional inaction, it lapsed in 2022, meaning its only remaining activities today relate to previously approved workers. The second Trump administration has proposed to finalize the wind-down of the program, ending it altogether.¹⁸

What the evidence on Trade Adjustment Assistance tells us

Trade Adjustment Assistance’s history reflected larger political dynamics. Yet its evolution and demise were shaped by a perception that it did not work as intended, as seen by expert views across several decades, including in:

• 1971: “The program has not achieved its major goals. It has failed to neutralize resistance to a liberal trade policy. And, the goal of economic adjustment intended by Congress has not been attained.”¹⁹
• 1987: “The program is now widely conceded to have been a failure.”²⁰
• 2000: “The TAA and NAFTA-TAA programs have spent more than $1.3 billion in the past 5 years to help workers make the transition to new jobs, but it is unclear how effective these programs are in achieving their goals.”²¹
• 2025: “As many observers have noted, [TAA] is woefully inadequate in scope and scale.”²²

The idea that Trade Adjustment Assistance was, at best, a flagging experiment and, at worst, an effort that was doomed to fail became common wisdom both on the left and the right. Even as organized labor generally pushed to maintain and expand the program, it was viewed as a secondary priority behind resisting greater tariff relief, as union support for the “grand bargain” reflected by the Trade Expansion Act of 1962—lower tariffs in return for greater adjustment assistance—began to fail. By the 1970s, “burial insurance” became the epithet of choice among unions to describe Trade Adjustment Assistance, reflecting a perception that it was paltry compensation in lieu of actually protecting the jobs and firms at risk from trade.²³

At the same time, any effort to make the program more expansive received pushback from conservatives and the business community, who argued it was a boondoggle. Fears that the program would ultimately either be mistargeted to support workers who were not actually harmed by trade, or would serve as the first tumble down a slippery slope toward a more expansive welfare state for all workers, created constant pressure to limit the program’s availability or generosity.

Formal efforts since the 1970s to evaluate Trade Adjustment Assistance, while not unanimous in their findings or their interpretation, have generally been used to justify these critiques. A 1981 study by Mathematica, a nonprofit policy analysis center, surveyed laid-off workers who began receiving adjustment assistance in 1976 and compared them to jobless workers who only received Unemployment Insurance. The study concluded that “TAA services did little to mitigate” the losses that workers experienced as a result of displacement.²⁴

A second Mathematica analysis of TAA recipients in the late 1980s found that the program was “well-targeted,” but that “TAA training did not have a substantial positive effect on earnings of TAA trainees, at least in the first three years after the initial UI claim.”²⁵ Subsequent research that used the same data but added controls for pre-layoff conditions of the workers and attempted to correct for selection bias among TAA participants versus UI exhaustees similarly found that Trade Adjustment Assistance did not seem to provide any earnings boost—although it did identify that recipients who received training, as opposed to those who got a waiver, were more likely to be employed.²⁶

Building on that finding, a 2011 study by Kara Reynolds of American University and John Palatucci of Rutgers University concluded that there was “no statistical evidence” that the program improved the employment outcome of TAA beneficiaries overall, although it did find that those workers who engaged in training “are more likely to obtain reemployment, and at higher wages, when compared to the TAA beneficiaries who do not participate.” They also found that the TAA training component increased the likelihood of employment by 10 percentage points to 12 percentage points and reduced earnings losses by 8 percentage points to 10 percentage points.²⁷

Perhaps the most widely cited evaluation of Trade Adjustment Assistance in recent years was also commissioned from Mathematica by the U.S. Department of Labor and published in 2012. It found that while after 4 years, TAA recipients “almost entirely closed the gap in employment and earnings, and by some measures, they had pulled slightly ahead,” compared to unemployed workers who did not participate in the program, it still did not make up for the time they were out of the workforce due to training.²⁸ As a result, TAA participation was found to have a negative impact on total income overall, even including the payments participants received as part of the program.

“Without considering the benefits of TAA stemming from the possibility that it promotes free trade, the estimated net benefit to society of the TAA program as it operated under the 2002 amendments was negative $53,802 per participant,” the study concluded, once taking into account lower earnings for participants, as well as the cost of government services. The study’s authors acknowledged limitations, however, in interpreting their findings—in particular, that the study period (4 years) may not have been long enough to measure the full benefits of TAA training, that the timing (during the Great Recession) may have confounded the results, and that TAA recipients may have differed from other workers in unobservable ways.

Recent assessments of Trade Adjustment Assistance by the economist Benjamin Hyman at the University of California, Los Angeles seek to correct for some of these challenges.²⁹ In a 2018 paper, Hyman analyzes 20 years of U.S. Census Bureau data on displaced workers, using a quasi-experimental approach that attempts to overcome issues of selection bias by looking at whether workers saw their TAA petition assigned to a more or less lenient U.S. Labor Department investigator. (The study takes advantage of the fact that some TAA investigators are much more likely than others to approve petitions in the same industry.)

Hyman “find[s] evidence of “large initial returns to TAA”—workers earn $50,000 more over the course of 10 years after engaging in TAA-related retraining, even as they forego roughly $10,000 in income while actually enrolled in the training program. Hyman also identifies that the gains from training decay over time, though these “diminishing returns are restricted to states with low training durations,” suggesting that participants receiving higher-quality training may continue to experience benefits.

Hyman also concludes that workers who participate in Trade Adjustment Assistance “are more likely to switch industries and move to labor markets with better opportunities.” Overall, he calculates—in contrast to the Mathematica study—positive social returns to TAA investments, even using a conservative approach and ignoring potential externalities.

In a separate study, Hyman, along with Brian Kovak of Carnegie Mellon University and Adam Leive of the University of California, Berkeley, evaluates TAA wage insurance by comparing workers just above and below the 50-year-old cut-off for eligibility, finding that the program increases employment probabilities by 8 percent to 17 percent during the 2 years following displacement (before fading to zero) and increases cumulative earnings by $18,000 over the 4 years following displacement.³⁰ Even before taking into account the potential reduced spending needed from other programs, such as disability insurance, they find that wage insurance “pays for itself,” returning more revenue to the government than it costs. 

The other important question is whether the program has had an impact on dampening the political backlash to trade. Using TAA applications as a proxy for displacement in studying the 2004 election, Yotam Margalit of Tel Aviv University finds that voters are “more sensitive”—measured by votes against the incumbent party—to job losses when it is due to foreign competition, and identifies that this effect is larger in areas where a greater share of workers applied for but were denied adjustment assistance.³¹

Margalit also takes care to reject a potential explanation that TAA certification itself was political, meaning that this result is not explained by higher denial rates in counties that are already more Democratic-leaning. Margalit concludes that “a government-funded compensatory scheme can do more than help workers readjust in the labor market. It can also serve as a political tool for politicians that want to advance trade liberalization but fear its electoral repercussions.”

Similarly, Sung Eun Kim of Korea University and Krzysztof Pelc of McGill University use 25 years of data to show that U.S. counties with a stronger history of successful TAA petitions are less likely to see petitions requesting tariff relief. By contrast, rejected TAA petitions do not have the same “dampening effect on protectionist claims.”³² And Melinda Ritchie of the Ohio State University and Hye Young You of Princeton University also find that higher approval rates on TAA petitions are associated with lower county-level vote share for President Trump in both the 2016 Republican primary and general elections, with the effects statistically significant, although modest in magnitude.³³

It is also worth noting that even as Trade Adjustment Assistance was often described as aiding communities hit hard by imports—recognizing that the harm caused to individual workers could spill over to others when concentrated within a single location—the program offered so little in place-based aid as to make evaluation of those efforts difficult. As part of the post-NAFTA provisions added to Trade Adjustment Assistance in 1993, the Community Adjustment and Investment Program was established to provide loan guarantees, loans and grants in counties that faced NAFTA-related job losses. The program ultimately was allocated less than $45 million.

Per a 2000 report by the General Accounting Office, the nonpartisan congressional watchdog agency now called the U.S. Government Accountability Office, it took “over 3 years to set up program guidelines and start disbursing program financing to distressed counties.”³⁴ Another GAO report from 2001 examining six communities that faced substantial job losses due to imports concluded that the communities “face[d] fundamental challenges in restructuring economies, while the available adjustment assistance is limited, targeted, and short-term.”³⁵

A more expansive “TAA for Communities” program was passed by the U.S. Senate in 2002 and then stripped from the overarching bill before the final legislation was enacted; a new version finally was signed into law under the 2009 expansions of Trade Adjustment Assistance. But that effort, too, was both underresourced and short-lived—it was repealed by 2012, under the argument that it duplicated the work of the U.S. Department of Commerce’s Economic Development Administration.³⁶

Indeed, that agency’s work does reflect the kind of community support that has often been envisioned for trade-affected communities, with its mandate defined as supporting communities harmed by “changing trade patterns.” But its modest funding—outside of recent supplemental investments during the Biden administration—have fallen far short of what might be necessary to provide a broad, robust response to the community impacts of concentrated job losses.³⁷

What, then, can we conclude from the research into Trade Adjustment Assistance? In part, it illustrates that assessing the program’s success or failure is often complicated by separating two different questions. One, did theworkers who the program was meant to reach actually receive assistance? And two, did the assistance help those who got it? The most sophisticated research provides some evidence that the best version of a TAA-like model could, in fact, have benefits for workers who are displaced, if they are positioned to take full advantage of its potential.

Yet millions of people who Trade Adjustment Assistance presumably was intended to support were ultimately left out. To offer one illustrative example, the economists Autor, Dorn, and Hanson estimate that as many as 2 million jobs were lost due to the China trade shock in the early to mid-2000s; by comparison, from 2004 to 2006, an average of only about 130,000 workers a year were covered by the program, and an average of fewer than 50,000 a year entered into training.³⁸ A program that works for those who participate, yet fails to reach most of those who it is meant to help, is still a failure—and is cold comfort, or worse, for those who have lost their jobs.

Despite often-good intentions, it is fair to judge the overall economic and political project of Trade Adjustment Assistance harshly, considering the evidence we have about both the experiences of trade-impacted workers and the political fallout—which may have only been exacerbated by a sense that promises had been broken. (And, of course, the fact that Trade Adjustment Assistance has been allowed to expire is a data point unto itself.) The question is whether there are lessons from that failure that might inform a better approach in the face of future displacement.

Are there lessons to be learned from past adjustment assistance efforts?

The conversations about potential future displacement as a result of AI have, to date, focused far more on trying to predict the scale of that displacement rather than on identifying specific solutions. At the moment, the range of potential outcomes due to AI seems particularly large, which can make it difficult to even assess what problem policymakers might be trying to solve. How many workers are displaced, whether and how the displacement is concentrated by occupation, industry, or geography, and where it hits on the education and income ladders are all key questions that should shape policy responses—and where there is substantial uncertainty around what might happen.

At worst, these conversations—particularly when framed by corporate leaders themselves—have assumed both that displacement is inevitable and that the primary responsibility for adapting must be on workers themselves to “retool” in response to a changing labor market. Those assumptions should be challenged: Policymakers should explore and pursue policies that help develop and deploy technologies that complement workers, rather than replace them, and they should seek to reduce incentives—in the tax code and elsewhere—for employers to substitute machines for workers.

But to the extent that at least some displacement is unavoidable, particularly in response to technological change, policymakers should learn lessons from the past on how to better assist workers. (The same is true for displacement due to other factors, including the clean energy transition. These can build on the tools that were included in efforts such as the Inflation Reduction Act’s programs directed specifically at “energy communities,” where workers are facing a transition away from fossil-fuel jobs.³⁹) And indeed, government officials, members of Congress, academics, and AI companies themselves have pointed to Trade Adjustment Assistance as a model for future programs, though often with little reflection on what worked or did not work in the past.⁴⁰

A threshold question is whether the fundamental approach that underpins efforts to provide assistance to displaced workers can work or whether, even if we learn from the failures of Trade Adjustment Assistance, these policies are designed to fail. The perception of these policies as “burial insurance” has led those who are not yet displaced and their advocates, such as unions, to understandably focus greater attention on preventing displacement in the first place.

Political leaders who are primarily interested in adjustment policies as a means of maintaining support for a broader economic agenda—even if they sincerely want to support those who lose out—may treat these workers as secondary to their overall goals. In addition, those already displaced may, by the very nature of their displacement, be ill-equipped to organize effectively as a meaningful political force. Finally, in the case of Trade Adjustment Assistance, one concerning observation is that even as the program failed to sufficiently protect workers from the fallout of increased imports, its existence was used as an excuse to sidestep trade policies that might have better avoided displacement in the first place.

The challenges that have hampered prior efforts at adjustment assistance offer a strong case that it cannot succeed as the only (or even primary) means to help workers who might be at risk from displacement and should not serve as an excuse to avoid wrestling with how to prevent dislocation. To offer a historical example, while Trade Adjustment Assistance as implemented was clearly insufficient to address the targeted displacement caused by the China shock after its entry into the World Trade Organization at the turn of the 21st century, even a more robust program would not necessarily justify the same trade policy choices with respect to China—including the failure to activate special safeguards that might have protected against the rapid surge in imports.⁴¹

Indeed, a first observation is that adjustment assistance may be most likely to succeed—both economically and politically—if it is not positioned as the sole solution to displacement risks. As noted earlier, closing coverage gaps in broader social programs would reduce the adverse consequences of displacement in ways that would require adjustment assistance programs to do less to cushion workers’ economic losses. Additionally, creating more bargaining power for workers within firms, including through collective bargaining, would increase pressure on companies to pursue alternatives to layoffs and provide greater support for those who do lose their jobs, as well as offer protections against other worker abuses that new technology such as AI may facilitate.

If we take warnings of AI-related job displacements at face value, then the importance of these policies should grow even stronger, both as a means of preventing job losses and lessening their cost to workers. Yet it is likely these policies—and other approaches to reduce displacement in the first place—may be necessary, but not sufficient.

The experience of Trade Adjustment Assistance raises one additional question for policymakers, which is whether there is any merit in targeting efforts at assisting displaced workers based on why they are displaced—rather than just improving supports for alldisplaced workers. This is not a new idea: President Nixon proposed to end Trade Adjustment Assistance in favor of more generous unemployment assistance for all. As discussed further below, there is little question that efforts to narrowly target support—and concerns about assisting the “wrong” people—were a key element in limiting the effectiveness of Trade Adjustment Assistance, preventing workers who clearly were harmed by trade from receiving aid in a timely fashion or at all.

Given that AI is a “general-purpose technology” likely to reshape a large share of jobs across industries—potentially including both blue-collar and white-collar occupations—one answer might be that the bulk of assistance efforts designed to address AI-related displacement should end up, in practice, being universal. To be sure, policymakers should be mindful of any particular differentiated harms that might be felt by those who lose jobs in industries where demand for their labor has been reshaped by artificial intelligence—and who, therefore, might face larger barriers finding a successful path forward. For example, trade-related displacement does appear to have had some outsized negative impacts on people and places, compared to job losses for other reasons, given its concentrated impacts.⁴²

At the same time, the potential for significant displacement should ideally provide the political impetus to offer support to the broadest possible set of displaced workers—establishing a much higher floor for the assistance provided to all workers. This is especially true if AI displacement hits workers further up the income ladder, given the perversity of providing much more generous assistance to someone who loses a higher-paying job due to AI than a low-wage worker laid off for unrelated reasons. As a result, additional targeted support for those who face unique harms due to a novel kind of shock—whether AI or the climate transition—should be seen primarily as a supplement to that universal program.

Principles and policies for helping people and places adjust

Learning from history, there are three broad principles and four specific design considerations that might inform efforts to help people and places transition when they lose out amid economic change. This paper will now detail each in turn.

Principle 1: Commitment

Government efforts to help workers and communities adjust have almost always been contingent and tenuous, in contrast to more successful social insurance programs. Trade Adjustment Assistance, efforts to help coal miners and timber workers in the 1990s, and recent investments in energy communities have all come into existence as attempts to “solve” a highly contested political dispute. They have rarely received the funding necessary to fully achieve their goals and, during most of their existence, operated at risk of termination.

As a consequence, transition efforts have rarely been as robust, flexible, or effective as they needed to be. The early years of Trade Adjustment Assistance, when not a single worker received assistance for 7 years, are reflective of this reality. Whether the excessively restrictive standard at that time was a result of statutory language or the administrators’ interpretations of it, the policymakers who had touted the program’s potential did little to make sure it worked.

Even as the program became better established, its stop-and-start nature limited its effectiveness. During the 1980s, for example, states expressed frustration that they were required to apply for new administrative funding each year, preventing them from building any lasting capacity. Simple fixes, such as making this funding guaranteed every year, would have allowed them to “hire permanent staff to take care of TAA clients, providing more individual counseling and assessment, as well as doing a better job of TAA outreach.”⁴³

In part, these challenges were the consequence of dedicated opposition to robust adjustment assistance efforts. But even proponents of these programs failed to approach them in a way that set them up for success. “Try harder” is a poor policy prescription, but the design and implementation of future efforts would benefit from a recognition that a different orientation is likely needed to get the economic and political benefits that transition assistance programs are meant to provide.

Future policymakers should err on the side of being over-aggressive—and over-inclusive—in their response to AI-related economic dislocation. Designers and implementers of Trade Adjustment Assistance too often accepted the premise that inadvertently reaching workers who experienced harm for a reason other than imports would reflect an inappropriate mission creep. As a result, they spent more time and energy trying to keep the wrong workers out at the expense of getting the right workers in.

Concerns about the program costing too much also weighed heavily on any debate, even as, in retrospect, its size remained tiny relative to the harm it was trying to address. Addressing future displacement will require a different mindset: Policymakers should seek to maximally reduce barriers to assistance both to ensure that those facing novel dislocation from AI are aided and because helping a broader set of dislocated workers is itself beneficial. 

One way to measure an effective adjustment program is whether it matches the scale of displacement. Policymakers should closely track data to make sure that eligibility—and, just as importantly, take-up—expands whenever dislocation increases. The Trade Act of 1974 created a trust fund for Trade Adjustment Assistance meant to grow with the level of imports, and though it was never implemented, it offers one model for a flexible pot that could be adapted based on how displacement unfolds—and potentially financed in ways connected to those benefiting from the transition, such as through taxes that fall on those who profit most from the deployment of AI.⁴⁴

More broadly, policymakers cannot treat adjustment assistance efforts for AI-driven economic displacements as a mere bargaining chip contingent on achieving other policies. In part, doing so means necessary fixes—even technical ones—are likely to be held up by larger debates. Moreover, this approach makes it harder to be sufficiently ambitious. Those who are concerned about displacement will be resistant to make other policy concessions, while those who want to push other policy changes will demand more in return for building a more successful adjustment assistance program. A strong effort at transition assistance must be pursued aggressively and ambitiously as a good unto itself—both as a matter of program design and as a prerequisite to build confidence among those at risk of displacement that policymakers truly have their best interests at heart.

Principle 2: Speed

Past efforts at helping workers adapt to economic change suggest it is not simply a question of whether workers are able to adjust, but also how quickly. A large body of research has shown that extended periods of unemployment have lasting impacts on workers’ earnings, health, and well-being, even after they are eventually reemployed.⁴⁵ The complicated TAA certification process meant that even workers who received assistance often waited months or years between when they were laid off and when they actually accessed TAA services.

Adjustment efforts should operate with a ticking clock, where the goal is to avoid any additional day a worker is out of work or not engaged on a pathway back to work. And indeed, the clock should not start at the moment when a worker is laid off or a firm closes, but rather when the risk of displacement becomes high. While tools such as WARN notices—which require employers to provide 60 days advance notification before mass layoffs—offer partial assistance, efforts should be made to provide greater incentives for firms to give longer-term warnings when workers are at risk of being displaced. This should be combined with initiatives to reach workers in occupations that are at risk of displacement from AI (or the climate transition), making them a priority for services that can prepare them for training.

More broadly, though, the greatest lesson from programs such as Trade Adjustment Assistance is that any complexity around adjudicating whether a worker is eligible adds a potentially fatal barrier to a program’s success. Not only can this complexity reduce access altogether, but in the case of the TAA program, it also meant the application-and-approval processes themselves took months or years even for workers who did ultimately qualify.

Over time, Trade Adjustment Assistance built in internal deadlines to guarantee adjudication. Similarly, any transition program for AI-driven displacements should specify a maximum number of days post-layoff by which any assistance should be approved. In addition to potentially attaching aid to pre-layoff notices, programs should operate with high degrees of presumptive eligibility, making it possible for significant numbers of workers to qualify with limited processes or paperwork.

That presumptive process could include designating workers in certain high-risk occupations or industries as automatically eligible for services or compensation once they lose their jobs (to the extent that assistance is not universal), offering access to transition assistance to workers before layoffs if they want to jump to a different career path, or even providing help to firms to transition workers to new roles in lieu of job cuts—all while raising the floor for the services available to all displaced workers regardless of the cause of their joblessness.

Principle 3: Agency

Defining policies as helping a person or a place adjust already reflects a sense that they are being required to adapt to something happening to them. And while the descriptions “compensate the losers” or “burial insurance” may not fairly characterize what adjustment assistance is meant to do—offering services and support that go beyond mere compensation—these tags have stuck for a reason. After all, these policies have been designed with a certain acceptance that displacement will happen and that recipients of assistance can either accept what they are given or get nothing at all.

Even as efforts such as Trade Adjustment Assistance have been pursued with politics in mind, they have rarely placed much emphasis on creating a sense of control for the workers or communities they reach. The process of receiving assistance—an opaque maze of applications, followed (if lucky) by services that can often be challenging to access and impersonal in their delivery—does not create conditions that empower workers.

Indeed, official evaluations of the implementation of Trade Adjustment Assistance have noted the challenges facing workers who are forced to make decisions about training programs right as they have seen their career paths disappear, with one report describing how “some were still suffering from the emotional turmoil of losing their jobs, a situation particularly common for those who worked for decades at traditional manufacturing firms.”⁴⁶

The question is whether and how adjustment assistance might be designed to allow workers greater agency. To be clear, agency alone is unlikely to be an effective tool. A long history of poor outcomes from weakly regulated for-profit educational and training institutions does not bode well for, to offer one example, viewing large training vouchers to workers as the solution.⁴⁷

But policymakers should reimagine not only which services are available, but also how they are provided. How is the availability of support communicated to workers when they lose their jobs? To what degree are they provided with knowledge and coaching about the options available to them, with the opportunity to make informed choices that reflect their preferences? When places are upended by economic change, how is the community engaged in building a path forward? Could communities be provided with some resources to deploy themselves?

Providing agency, of course, is harder in an environment where too many workers lack voice in the workplace to begin with, particularly due to the decline of private-sector unions. In unionized workplaces, worker power creates an opportunity not only to push back against efforts to displace workers when other alternatives exist, but also to ensure sufficient assistance for those who are let go, informed by workers themselves. Indeed, not only were unions a major filer of TAA petitions, but qualitative research also suggests they played an important role in supporting TAA take-up.⁴⁸

In the 1960s, during a previous wave of technological change, so-called automation funds were established through collective bargaining to help workers transition, with decisions made by a combination of labor, management, and outside experts.⁴⁹ These efforts did not fully address the damage caused by job losses, but they did help blunt their harms, while providing workers with a greater sense of agency in what came next.

Unlike Trade Adjustment Assistance, these automation funds were often nimble and iterative. In one well-studied case at the meatpacker Armour & Company, an automation fund’s leaders learned from early struggles how to better assist workers through a combination of relocation support, retraining, and other tools, including maximizing the use of government resources.⁵⁰ Similarly, workers today could be more formally engaged in the process of design and implementation of federal, state, and local adjustment programs, creating feedback loops to ensure the program is serving their needs.  

Design consideration 1: Use high-quality training models and help them evolve

As with Trade Adjustment Assistance, the efficacy of federal job-training efforts has typically been understood by both policymakers and expert evaluators as mixed at best, particularly for displaced workers.⁵¹ But in recent years, a considerable research base has emerged suggesting that so-called sectoral training programs have large and persistent benefits for participants. These programs target specific industries and occupations that have “strong current local labor demand and opportunities for longer-term career advancement,” as Harvard economist Lawrence Katz and his co-authors have defined them, and typically include a combination of training in both soft and occupational skills, job development and placement, and wrap-around services—all in close partnership with employers.⁵²

The efficacy of these programs depends, of course, on the presence of strong labor demand in some areas, pointing to the related challenge of policies that create high-quality jobs in the first place. And while these high-quality programs are usually shorter term than formal postsecondary education, they are intensive and therefore relatively high cost. But their returns suggest a model worth investing in to address displacement. Recent examples include investments in the health care workforce, as well as efforts under the CHIPS and Science Act of 2022, that apply these best practices—from engagement with employers to the provision of wrap-around services—to train semiconductor manufacturing workers.⁵³

At the same time, policymakers and training providers should identify how these models for AI-driven displacement will need to adapt amid the impact of AI and related new technologies, recognizing that they may need to be tailored for a changing labor market. There is already emerging evidence, for example, that AI is reshaping the job search and hiring processes in ways that may add bias or create challenges for marginalized workers. Any approach to assist displaced workers will need to develop tools that are effective in a changing labor market.

Design consideration 2: Ensure that compensation and adjustment go hand in hand

At many stages in the evolution of Trade Adjustment Assistance, experts evaluating the program realized that far more workers were receiving compensation alone, without engaging in training. The response was often to more directly condition the receipt of funds on participation in training programs. This only partly solved the problem because many workers simply chose not to take part, whether because they could not find quality training or because they did not consider it worthwhile. At the same time, any time spent in training and not working is costly, even for a worker who could only earn a lower wage than before they were displaced.

Alongside more robust and accessible Unemployment Insurance for all workers, providing additional support for those who have been displaced by changes in the economy makes sense both due to the magnitude of the adjustment they are likely to face and to pay them while they are in training and out of the workforce. The solution is to ensure that sufficient compensation and the right services to facilitate adjustment are provided in tandem. Either one alone is unlikely to succeed. (An exception is reasonable for older workers where, as the economic research described above suggests, there is a strong case for wage insurance as the optimal solution to keep displaced workers in the labor force, rather than investing in additional training.)

Design consideration 3: Lean into tools that help us understand where the labor market is going

Major investments need to be made not only in data collection and analysis through agencies such as the U.S. Bureau of Labor Statistics, but also in attempts to help translate and disseminate this information to workers, employers, and communities. Efforts to identify warning signs, including from AI companies and employers themselves, could help inform which occupations or industries are most at risk, as well as where higher demand might exist in the future.

These data are crucial for policymakers, as well as training partners, workers, and communities. Better labor market analysis will never allow for total foresight about the future, but it can help design a roadmap that makes it more likely that the services provided to displaced workers are best suited to prepare them for a new economic reality.

Design consideration 4: Explore place-based policies, even if displacement is more dispersed

Trade-based displacement has been geographically concentrated within particular communities—especially in areas such as the industrial Midwest—which exacerbated the harm it caused even to others whose jobs were not directly affected by imports. The degree to which future displacement will look similar is unknown, particularly with respect to AI. While there is considerable evidence that AI usage is significantly more prevalent in some areas of the United States than others, that may not match to the ultimate labor market impacts.⁵⁴

Policymakers can have higher confidence that the climate transition will have a disproportionate impact on certain places, given the location of existing energy jobs and the physical nature of that transition. But even in the context of AI—and potentially more diffuse displacement—place-based policies can still be a crucial tool for giving workers a more accessible, empowering avenue to transition into a better career path, and with a set of tools that we have some practice using.

Recent efforts such as the Recompete program or the Build Back Better Regional Challenge—as well as targeted efforts as part of the CHIPS and Science Act and the Inflation Reduction Act—reflect promising models for place-based development designed to create and fill in-demand jobs.⁵⁵ These models, however, will need to be pursued at a larger scale and with a flexible approach that adjusts to changes in labor market demand.

Conclusion

Past efforts to respond to economic displacement, including through Trade Adjustment Assistance, have not lived up to the hopes of those who devised them—reflecting failures of both design and implementation, with major consequences for workers and our broader political landscape. The potential for displacement from AI and other forces should spark policies to reduce the likelihood of painful job losses but will likely also require measures to help people and places who experience harm. In the coming months, Equitable Growth will offer more detailed recommendations on policies that could achieve this goal—even amid uncertainty around the magnitude of the disruption to come.

To be successful, these efforts must avoid repeating the mistakes of the past, recognizing the need to provide a response that is more robust, faster, and more empowering for those who feel left behind. Doing so will be made even harder because the shape of the future labor market is so unclear, but that is all the more impetus for building tools that can be effective under a range of scenarios. Policymakers should heed the lessons of trade and globalization: Even if only a small minority of workers lose out due to an economic shift, that loss can have significant consequences. Past failures to prevent those consequences should lead to both humility about the magnitude of the challenge we face and a greater dedication to taking it on.

About the author

Jacob Leibenluft is a visiting fellow at the Washington Center for Equitable Growth. Previously, he was the executive associate director of the White House Office of Management and Budget during the Biden administration, responsible for coordinating day-to-day budget policy work at the agency. He also served at the U.S. Department of the Treasury as a counselor to Secretary Janet Yellen and as chief recovery officer, where he led Treasury’s efforts to disburse more than $400 billion in American Rescue Plan and other COVID-19 relief funds. During the Obama administration, Leibenluft served in several roles at the National Economic Council, including as deputy director of the council and deputy assistant to the president. He earned his B.A. in economics and history from Yale University.