Lost in Transition: How Trade Adjustment Assistance came up short (and where it succeeded)

As the power of generative AI tools has increased, concerns have intensified about the technology causing mass labor market disruptions. Although AI is likely not an important factor driving recent labor market deterioration,¹ it could increasingly contribute to labor turnover over time.  We need to be ready for this possibility. While improving the nuts-and-bolts functions of our unemployment insurance system is critical to help minimize the negative impacts of AI-driven job displacement,² monetary benefits alone may be insufficient. Direct job search support could be a more cost-effective way to help workers shift from declining sectors into the industries of the future.

The United States has been down this road before. Concerns about American workers who would be hurt by global trade deals led to the establishment of Trade Adjustment Assistance (TAA) in 1962. TAA was intended to assist workers whose jobs were eliminated due to increased imports or the relocation of manufacturing facilities overseas. Workers whose application for TAA was approved could receive job training and enhanced unemployment benefits. 

But TAA fell short of expectations. Persistent eligibility barriers, administrative complexity, and inadequate funding limited the program’s reach and success. After decades of mixed results, Congress allowed TAA to expire in 2022. TAA’s demise has made it all the more important to assess what went wrong with the program, and what lessons TAA offers for future worker adjustment efforts in an era of rapid technological change. This reexamination is already underway.³ As Senator Mark Kelly argued in his recent AI for America policy framework, “Past approaches to major disruptions, like Trade Adjustment Assistance (TAA), too often failed workers and were underfunded, underscoring the need to rethink and build a system that is more proactive, worker-focused, and responsive to the realities of technological change.”⁴

At the same time, policymakers should be careful not to overcorrect. Benefit inaccessibility should not be mistaken for benefit ineffectiveness. Recent research indicates that the workers able to overcome program barriers greatly benefited from TAA, and that the value of TAA support often exceeded the value provided by other workforce development programs. Legislative efforts that avoid TAA’s major weaknesses and account for its strengths will produce the best results.

Complexity and funding limitations hindered access to TAA

Congress periodically adjusted TAA over the past six decades, yet core issues with the program were never truly resolved. Workers struggled to be approved in a timely manner — or at all — because proving that one’s job was eliminated due to trade was difficult and resource-intensive. Trade-impacted workers often left unemployment and started new jobs before claims evaluations were completed. Even when unemployed workers were found eligible, they had to overcome more hurdles before actually benefiting. Funding limitations and restrictions often led to long wait times for training. These issues were present from TAA’s creation until its expiration.

Trade Adjustment Assistance was enacted in 1962 as part of a political compromise to eliminate the (frequent) use of escape clauses⁵ in trade agreements. TAA was supposed to cushion the blow for workers displaced by subsequent trade liberalization. Yet no applicants for TAA were certified to receive benefits in the first seven years of the program.⁶ To be approved, applicants had to demonstrate not only that they had been injured by U.S. trade liberalization, but that the trade liberalization had been the primary cause of their unemployment.⁷ Proving that trade was the single most important factor was a high bar to clear: Determinations required investigations into the business records of firms and their customers, which could take years.

After seven years without a single benefit approval, the Nixon administration helped modify program eligibility so that workers qualified if trade was proven to be a substantial cause (not necessarily the single most important reason).⁸ TAA enrollment increased, and by 1980 the program reached its peak level of recipients.⁹ Yet the fundamental tension at the heart of TAA remained unresolved. The program still required proof of a causal link between trade and workers’ job loss; the result was that a displaced worker’s first TAA payment was received 14 months, on average, after their job separation.¹⁰ Furthermore, less than 10% of TAA beneficiaries received job training services.¹¹

As the 1980s progressed, the program began to more heavily emphasize, and then require, job retraining.¹² But while the program was meant to help workers prepare for roles in new industries, the training infrastructure and funding failed to match its creators’ ambition. Congress approved limited funding for job training, which hampered access. Only 31% of eligible beneficiaries participated in job training services in the mid-1990s, a rate which fell below 20% at the turn of the 21st century.¹³ States reported waiting lists for job training programs, and sometimes had to suspend training efforts due to Department of Labor funding delays.

The training waiver system that emerged illustrates the breadth of the problem. States could grant waivers when training was unavailable or would not begin soon after workers qualified for support. Workers obtaining such waivers could then collect cash payments without enrolling in classes. By 1999, 38% of TAA participants who left the program had received some type of waiver.¹⁴

Were it not for the Workforce Investment Act (the predecessor to today’s workforce development system known as WIOA), many TAA applicants would not have received case management services — personalized career guidance — either. Under TAA, case management was deemed an administrative cost, and only 15% of program funds could be applied toward administration. This restriction limited how many applicants could receive this personalized guidance before benefits were approved.¹⁵ By connecting TAA applicants directly with WIA, workers could get valuable job search support even before being approved for training — a process hampered by TAA’s funding restrictions.

The largest underlying problems with TAA persisted despite these alterations. A small percentage of laid-off workers in industries heavily impacted by import competition received assistance via TAA.¹⁶ Initial determination times — while much improved from prior decades — still frequently fell short of the statutory requirements set out by Congress (between 40 and 60 days).¹⁷ It remained difficult and resource-intensive for officials to assess whether imports contributed to a worker’s unemployment spell. As documented in a 2000 Government Accountability Office report:

To evaluate whether this certification criterion has been met, Department investigators may survey up to six of the firm’s major declining customers to determine whether their reduced purchases from the firm were accompanied by increased imports of like or directly competitive products. These investigations go back to the 2 most recent full years, and the year−to-date statistics are compared to the same period of the previous years.¹⁸

Even after reforms in the 2000s, claimants still had to wait an average of eight months between when they submitted a benefit claim and when they were approved for training via TAA.¹⁹ Such delays were never fully resolved before TAA was allowed to lapse.²⁰ Workers in regions most affected by imports were much more likely to rely on disability insurance payments than to use TAA.²¹ Proving that trade was a distinct cause of job loss remained a challenging administrative task until the end.

Recent evidence shows real benefits for workers who accessed TAA

While facing up to TAA’s failures, policymakers should also take note of the efficacy of TAA benefits for those who accessed the program. The consensus for years has been: 1) TAA was bogged down by administrative issues, and 2) the program offerings had little impact. But recent evidence suggests the latter idea — that TAA program benefits were ineffective — is likely wrong.

To evaluate the effect of receiving TAA, one must compare workers who got access to TAA with otherwise similar workers who did not. Researchers have frequently used unemployed workers who received regular unemployment benefits as a comparison group. When controlling for demographics such as age, race, gender, education, and prior occupation, a prominent 2012 Department of Labor-sponsored study found that TAA recipients had worse outcomes than traditional Unemployment Insurance (UI) recipients.²² The authors evaluated recipients between 2004 and 2011, and came to the conclusion that workers enrolled in TAA ended up earning a few thousand dollars less than UI recipients in the comparison group four years after initial enrollment. 

This particular study has carried a lot of weight, but there are a few reasons to defer to more recent analyses. For one, a large portion of the DOL study occurred during the peak of the Great Recession. TAA recipients who entered the program in the mid-2000s were faced with a much worse labor market when exiting TAA than UI recipients who began claiming benefits at the same time but reentered the labor market earlier.²³ Likewise, comparing these respective groups four years after initial enrollment — as the DOL study does — could be too short of a time frame to capture a substantial portion of the total returns from TAA. Lastly, UI recipients may not be the best possible comparison group for gauging the effectiveness of TAA.

These potential drawbacks were addressed in a series of papers by Ben Hyman, formerly of the Federal Reserve Bank of New York. First, in a 2018 working paper, Hyman analyzed 20 years of worker-level data to gauge the effectiveness of TAA.²⁴ But rather than simply controlling for similar demographic characteristics, Hyman was able to estimate the causal effects of TAA claims being randomly assigned to program evaluators with different approval leniencies. Nearly identical workers could be approved or denied based on the claim evaluator they were assigned. Workers approved for TAA were found to have $50,000 more in cumulative earnings a decade later, and were more likely to be active labor market participants. Notably, the main benefits reflected TAA-approved workers receiving higher wages — not just higher labor force participation — indicating that the training provided workers with valuable human capital.

In addition to the direct returns to TAA participants, Hyman determined that TAA program offerings had high returns per public dollar spent. This assessment was performed by assessing the marginal value of public funds (MVPF), where the net returns from TAA were compared to the total change in the fiscal effects of TAA, the government’s outlays and receipts (including second-order effects, such as higher spending due to moral hazard, causing workers to stay on the program for longer, or increased tax revenues from increased earnings over time). An MVPF estimate significantly over one should be broadly interpreted as increasing social welfare, as each dollar spent results in a total return exceeding the initial expenditure; a value below one means that the social costs are greater than the social benefits.²⁵

In the case of TAA, the calculated MVPF was determined to be at least 1.14. This represents a lower-bound estimate under conservative assumptions — for example, that each recipient exhausts all TAA benefits available to them. When accounting for lower average costs per participant, and the fact that some recipients would substitute receipt of TAA for receipt of other benefits (for example, Social Security Disability Insurance), the denominator would fall (as the fiscal costs decline) and the MVPF would correspondingly increase.

This is not the only study that concludes TAA made a positive difference for workers. In a 2024 NBER working paper, Hyman and his coauthors also found that TAA’s wage insurance provisions had a positive effect on the employment of older, trade-disrupted workers.²⁶ They studied a 2002 change to TAA that allowed workers aged 50 and older on TAA to have up to 50% of the difference between their prior and current wage levels covered if they took a lower-paying job. This policy was, essentially, a strong reemployment incentive for these older workers. By using the eligibility discontinuity between workers right below and above 50 years old, the authors determined that eligible older workers were more likely to be employed after their job displacement as a result. Even though the employment effect did fade over time, this policy had a calculated MVPF not just in excess of one; it likely “paid for itself” (that is, the increase in government tax revenues plus the decrease in other government spending caused by TAA exceeds the government spending on TAA).

These effects are all the more impressive when considering the assessed impacts of other workforce development policies. MVPF estimates for workforce programs are commonly below one.²⁷ Measured returns often fail to justify the costs of such programs, calling their value into question.²⁸ So it is notable that TAA appeared to have social benefits greatly in excess of its social costs. The challenge ahead is capturing the strengths of the TAA program while avoiding its significant baggage.

Capturing the good without the bad

Lawmakers concerned about AI-driven job displacement ought to update their takeaways from TAA and incorporate new lessons learned from the program. TAA was meant to offset the pain of trade liberalization, yet the narrow program focus actively impeded workers from accessing it. Support arrived way too late, or not at all. These defects were tragic, because recent evidence suggests that the training and support were extremely valuable for enrolled workers.

In light of the successes and failures of TAA, congressional efforts to address increased labor market disruption in the wake of AI should follow two key principles:

• Base eligibility on employment status rather than the precise cause of unemployment.

The single greatest failure of TAA was the requirement that workers prove that their jobs were eliminated due to trade. Distinguishing between workers who lose jobs due to AI versus another factor will be just as difficult, if not more so, in an economy increasingly shaped by AI. Any system that attempts to condition eligibility on the cause of unemployment will move too slowly to be effective. Job adjustment assistance should be triggered by the occurrence of unemployment regardless of the underlying economic cause(s).

• Prioritize fixing and expanding the existing infrastructure that supports unemployed workers.

Congress should not try to reinvent the wheel with a bespoke “AI adjustment assistance” program. Setting up entirely new, AI-centric infrastructure risks a host of new implementation issues without addressing identified, and resolvable, problems with existing workforce programs — oftentimes, basic resource constraints. Instead, lawmakers should focus on improving and experimenting within the workforce development system already used by millions each year: scaling up workforce training programs like TAA’s, which delivered benefits that greatly exceeded its costs, and scaling down workforce programs that underperform.

Conclusion

If — or when — generative AI significantly reshapes the labor market, the temptation to build bespoke workforce programs specifically targeting AI-driven disruptions will be strong. Policymakers should resist it. By prioritizing robust reforms to our existing unemployment and workforce systems that benefit all workers — not just those who can prove that their jobs were eliminated due to AI — Congress can help the U.S. workforce retool without repeating the mistakes of TAA.

About the authors

Will Raderman is a Senior Policy Advisor at Searchlight, focusing on employment and social insurance policy issues. He previously served as a policy analyst at the Niskanen Center, where he testified before the House Ways and Means Subcommittee on Work & Welfare, and as a research fellow at Boston University, where he tracked and analyzed state policy actions during the COVID-19 pandemic, focusing on unemployment insurance and paid leave. Raderman’s work has been featured in The Washington Post, Barron’s, and the Wall Street Journal. Will hails from Dobbs Ferry, New York. He is a graduate of George Washington University and University College London.

Alex Mechanick is a Senior Policy Analyst in the State Capacity Initiative at the Niskanen Center. In that role, he focuses on developing reforms to administrative procedure and other topics that sit at the intersection of economics, law, and policy. Previously, Alex served in the front office of the Office of Information and Regulatory Affairs at the Office of Management and Budget, as part of the Senate Judiciary Committee staff of Senator Richard Blumenthal, and in the Global Modeling Studies section of the Board of Governors of the Federal Reserve. Alex received his A.B. in Economics and Philosophy from Brown University and his J.D. from Yale Law School.