This piece is a guest contribution from Yuliya Yatsyshina, Program Manager of the Labor and Fiscal Policy Projects at the Mercatus Center. This spring, Labor Market Matters is kicking off a series exploring the intersection of H-1B visas and the broader labor market. Over the coming months, we’ll be sharing a few pieces examining how high-skilled immigration policy shapes wages, hiring, and economic opportunity.

Subscribe now

This month marks the first round of the H-1B lottery conducted under a new rule requiring U.S. employers hiring specialty occupation workers from abroad to pay fees of up to $100,000 per application. These unexpectedly high costs reportedly caught many H-1B employers off guard. Combined with a shift toward prioritizing higher-paid, “merit-based” positions, the policy was introduced by the Trump administration to curb alleged fraud and abuse in the temporary visa system.

In practice, however, the measure functions as a tax on foreign labor which risks generating long-lasting and unintended consequences for U.S. labor markets and the broader economy.

The stated policy objective

One of the administration’s main objectives with this policy is to protect American workers by discouraging reliance on lower-cost foreign labor while prioritizing the admission of exceptionally high-skilled workers.

But will such a drastic policy shift actually deliver a net benefit to the United States or to American workers in particular? Will it meaningfully reduce the alleged fraud it targets? And what broader economic consequences can we expect from raising new barriers to legal immigration?

Economic realities

The full effects of any policy change are difficult to predict. However, a fee of this scale, which is effectively a tariff on foreign labor, will likely cause firms to adjust their talent sourcing, leading to disproportionate effects across businesses and industries. While large firms may be able to absorb the added costs, small and mid-sized companies are more likely to face heightened barriers when competing for international workers.

At the same time, recent simulation evidence suggests the fee may have only a limited impact on employer demand in some cases. On average, H-1B workers earn about 15 percent less than comparable native workers, implying that firms may still be willing to pay a one-time fee to secure these hires. In practice, this means some employers could offset the added cost by adjusting compensation either for foreign hires or for domestic workers.

In the cases where wage gaps are already narrow, particularly in large tech firms such as Amazon, Google, and Microsoft, these adjustments are more likely to come at the expense of new hires regardless of origin, as firms rebalance overall labor costs. As Nvidia CEO Jensen Huang has noted, while a $100,000 visa fee could pose a meaningful barrier to opportunity, firms may still choose to pay it when the expected returns justify the cost.

Additionally, the United States does not produce enough STEM graduates domestically, creating a structural reliance within the tech sector on foreign-born talent, as reflected in National Center for Science and Engineering Statistics data (see figure below). So far, India remains a leading producer of high-quality technical graduates, while China continues to significantly outpace the United States in the production of STEM PhDs.

Given the relative scarcity of STEM talent, strong U.S. productive capacity, and intensifying competition from AI, demand for foreign-trained labor is unlikely to decline. When the cost of key inputs rises, firms adjust toward cheaper or more accessible alternatives.

A $100,000 H-1B fee raises the marginal cost of hiring foreign specialty workers, incentivizing substitution across several margins including increased offshoring to foreign affiliates, greater reliance on remote contractors, automation, or shifting operations to jurisdictions with more predictable immigration regimes. However, when considering offshoring, smaller firms cannot use that as easily as a Google or Amazon, so some may lose access to talent entirely.

Highly skilled workers remain a scarce and essential resource, and no technological shortcut, including AI, is likely to change that in the near to medium term.

While some argue that AI could offset domestic talent shortages, this view overlooks how technological change reshapes, rather than eliminates, labor demand. As AI advances, it generates new roles and shifts skill requirements, increasing the need for specialized expertise. According to Pew Research, nine in ten H-1B requests are for jobs requiring some high-level STEM knowledge, which suggests that employers are not exploiting the system but simply solving the economic problem of persistent shortages of skilled labor.

Ample research confirms that work visa restrictions generate many unintended consequences. In an earlier policy brief, Liya Palagashvili explains that limiting H-1B visas does not reliably achieve the stated goal of protecting American workers.

For example, using H-1B visa microdata and Bureau of Economic Analysis data on U.S. multinational firms, Britta Glennon (2020) shows that restrictions on high-skilled immigration lead multinational firms to expand foreign affiliate employment abroad such as in Canada, India, and China. She finds that for every visa rejection, the average firm hires approximately 0.42 employees overseas, rising to 0.93 for the most globally integrated firms. In other words, the firms are likely to add approximately two to three overseas employees for every visa rejection.

Evidence on domestic labor markets points in a similar direction. Palagashvili’s review of the literature finds that H-1B workers do not reduce aggregate employment or wages for American workers. While there may be localized or occupational shifts, such as within computer science fields, there is no evidence of overall job loss attributable to the H-1B program. Indeed, the preponderance of evidence indicates that restricting H-1B visas may undermine, rather than advance, the goal of strengthening outcomes for American workers.

Firms also adjust through regulatory arbitrage. Employers adapt by navigating around immigration restrictions through alternative channels to bring in foreign talent. When compliance costs increase sharply, firms have incentives to search for alternative legal visa categories that allow them to avoid the fee. In practice, this may mean restructuring job classifications, shifting workers to other nonimmigrant visa categories, partnering with third-party staffing firms that are able to bring talent exploring alternative visa systems, such as from OPT (F-1) to H-1B.

All of these adjustments require time and resources, meaning the policy’s ultimate impact will depend less on its intended deterrent effect and more on how firms substitute across these margins. The upcoming FY2027 registration data will provide early evidence on whether the reform meaningfully reshapes high-skill labor markets or primarily induces offshoring and reallocation.

Notably, the policy is expected to generate between $10 and $20 billion in annual revenue. This is a transfer from employers to the federal government. Basic economic theory predicts that such a tax will create deadweight losses by preventing mutually beneficial exchanges. These losses may take the form of fewer hires, reduced innovation, and diminished consumer benefits. While federal revenues may rise, the burden is to fall on firms, workers, and consumers.

On the issue of fraud

Visa fraud is a complex issue, and in the H-1B context, it is essential to distinguish between different types of violations and their prevalence. Unfortunately, little recent peer-reviewed research exists. The most frequently cited evidence remains a 2008 USCIS study, which identified problems in roughly 20 percent of applications which lumped together both outright fraud and minor technical violations.

Only 85,000 H-1B visas are issued annually, compared to more than 400,000 F-1 student visas and nearly 6 million B1/B2 tourist visas. Incidents of fraud and noncompliance issues are far more likely with these visa categories. Focusing enforcement on H-1B alone therefore targets a relatively small share of overall visa activity.

Moreover, the new fee is unlikely to meaningfully reduce fraud given the availability of alternative pathways for high-skilled workers. In addition to OPT (F-1) opportunities, employers can turn to other visa categories such as O-1, EB-1, EB-2, and EB-3. As a result, raising the cost of H-1B applications may shift behavior across programs rather than address underlying compliance issues.

More problematic is the lack of sufficient resources and incentives for USCIS and Consular services to identify, investigate, and reduce fraud before an application is approved. For these reasons, a fee-based deterrent is a poor fit for the problem of fraud since it takes numerous forms such as misclassified job roles, speculative filings by staffing intermediaries, and wage manipulation, none of which are meaningfully discouraged by raising the entry price.

A more effective approach would emphasize targeted enforcement such as enhanced auditing, stronger penalties for bad actors, and more robust site visit requirements. These measures would better address abuse and avoid placing unnecessary burdens on the many employers and workers who participate in the program in good faith.

Furthermore, recent research by Adam Ozimek and Sarah Eckhardt finds that H-1B holder who do apply in good faith generate substantial positive balances: “the average H-1B household contributes $30,050 net annually — 2.6 times the $11,530 contribution of a typical U.S. household.”

What’s actually at stake

A substantial body of evidence shows that high-skilled immigration strengthens the resilience of the U.S. economy, supports labor market dynamism, fuels innovation in strategic industries, and contributes positively to the nation’s fiscal health. In short, the benefits far outweigh the risks.

However, an increasingly outdated immigration system with rising fees and increasing barriers to high-skilled immigration creates a significant policy trade-off. The U.S. government may stand to gain billions in additional revenue, but the costs to productivity, innovation, and competitiveness may far outweigh that hoped-for windfall. Congress has already introduced a bill for fee waivers for H-1B healthcare workers in recognition of the policy’s initial negative consequences.

Rather than continuing to patch a system that distorts incentives and constrains growth, policymakers should reconsider the broader framework. At stake is not just access to talent, but America's position as a global leader in technology, a standing that is tested continuously given intensely competitive global landscape.

Ultimately, high-skilled immigration policy is not merely a domestic labor question, it is a strategic lever with consequences that extend well beyond U.S. borders. When talent can move more freely across borders, the American and global economies prosper.

When I first started talking to policymakers in February 2020 about portable benefits, the idea was still largely theoretical—something that lived mostly in research papers and policy conversations, not in actual law.

Over the years, I’ve testified before Congress and in state legislatures across the country on this issue, and my conversations with policymakers consistently reflect a genuine interest in addressing this gap.

Today, that idea has moved firmly into the mainstream.

In the past three years, 18 states have introduced, enacted, or piloted portable benefits reforms—turning what began as a single-state effort into a rapidly expanding, multi-state policy movement. At the same time, two major congressional bills are advancing similar safe harbor frameworks at the federal level.

As more states adopt these frameworks, portable benefits are quickly moving from a policy experiment to a central part of the national labor policy debate.

The Map: A Policy Movement in Real Time

What’s striking is not just where portable benefits are appearing—but how fast.

States like Alabama, Idaho, Tennessee, Utah, West Virginia, and Wyoming have already enacted voluntary portable benefits frameworks. Others—including Pennsylvania, Maryland, and Georgia—have launched pilot programs. And a growing number of states—from Connecticut to Kansas to Hawaii—are actively considering legislation. Below is a list of state-level portable benefits legislation introduced in 2026:

• Connecticut — H.B. 5041 (health care benefits)

• Florida — H.B. 1431

• Georgia — H.B. 987

• Hawaii — S.B. 2088 (health care benefits)

• Kansas — H.B.2602 (includes tax advantage)

• Kentucky — H.B. 732

• Louisiana — H.B. 301

• Mississippi — H.B. 1072

• New Hampshire — H.B. 1245

• Rhode Island — H.B. 7365

This kind of policy diffusion is rare. It signals something important: lawmakers across very different political and economic environments are converging on the same idea.

This momentum has been supported by a growing coalition of policy organizations, researchers, worker advocates, and industry stakeholders working to expand access to portable benefits.

What Portable Benefits Actually Do

At their core, portable benefits solve a simple but deeply embedded problem in U.S. labor policy.

Our benefits system was built around a single-employer model. But today, roughly 30 million Americans work independently, generating nearly $1.5 trillion in economic activity annually.

The independent workforce has grown by 97 percent since the late 1990s.

And yet, many of these workers lack access to benefits—not because companies are unwilling to offer them, but because the law makes it risky to do so.

Portable benefits directly address this problem. They are worker-owned accounts that follow individuals across jobs and allow multiple companies or clients to contribute. The key policy change is simple: remove the legal risk that offering benefits will trigger worker reclassification.

Under current law, the presence of benefits can be used as evidence that a worker is an employee. Portable benefits laws clarify that this should not be the case.

In other words, voluntary portable benefits safe harbor reforms don’t mandate benefits—they make it possible to offer them. By contrast, other portable benefits bills are typically mandatory or limited to app-based industries.1

Three Reasons This Is Taking Off Now

There are several reasons this momentum is accelerating.

First, the workforce has already changed. Independent work is no longer niche—it now spans industries ranging from professional services and healthcare to transportation and the arts. Surveys increasingly show that Gen Z is more interested in self-employment and entrepreneurial work than previous generations at the same age—a shift that is already shaping how policymakers think about the future of work. This approach has also been supported by state policymakers, including through a 2023 framework developed by the National Conference of State Legislatures.

Second, workers themselves are asking for it. About 80 percent of independent workers prefer to remain independent, and roughly 81 percent say they want access to portable benefits.

Third—and most importantly—this approach breaks a long-standing policy stalemate.

For years, the debate has been framed as a binary choice: either force workers into traditional employment or leave them without benefits.

Portable benefits offer a third path: expanding access to benefits while preserving flexibility.

Are There Downsides? Addressing the Biggest Concerns

As portable benefits spread, a predictable set of questions comes up. These are worth addressing directly.

1). Does this increase misclassification?

No.

Portable benefits policies do not change the legal definition of employment. They do not alter existing classification tests—whether it’s the ABC test or common law standards.

They simply clarify one narrow point: the presence of benefits cannot be used as evidence of employee status.

For example, states like Connecticut, which already use one of the strictest worker classification tests in the country (the ABC test), can adopt portable benefits without changing how that test is applied.

This is not about redefining work—it’s about removing a barrier that prevents benefits from reaching workers who are already independent.

2). Why not just make everyone employees?

Because most independent workers don’t want that.

Independent work provides autonomy and flexibility that traditional employment often cannot—especially for caregivers, parents, students, and those managing multiple income streams.

Portable benefits meet workers where they are. They expand access to benefits without requiring workers to give up the arrangements they prefer.

In states that have taken far more restrictive approaches, the results have often backfired. In California, for example, AB5 had to be partially rolled back with over 100 occupational exemptions. Our research shows that for occupations that were not exempt, independent work declined without any corresponding increase in W-2 employment.

3). Will companies actually contribute?

Yes—and we’re already seeing it happen.

Once the legal barrier is removed, companies begin to participate. Portable benefits programs in multiple states show that businesses are willing to contribute when they have clarity and certainty.

As an example, in Pennsylvania, workers in the DoorDash pilot program received more than $1.3 million in benefits. Nearly 75 percent of workers who previously lacked access to benefits gained access through a portable benefits program. Among them, 81 percent reported feeling more financially secure, and 95 percent said they would feel even more secure if the program were made permanent.

The demand is there—from both workers and companies. The missing piece has simply been legal clarity.

What Does the Data Actually Show?

The best real-world evidence so far comes from Utah—the first state to adopt a voluntary portable benefits framework.

My research team analyzed outcomes in Utah, where a portable benefits policy took effect in 2023. The early evidence shows that traditional W-2 employment and self-employment continued to grow at similar rates as before. The policy simply allowed independent workers to access benefits.

In essence, W-2 employment growth remained stable after the bill’s enactment. Self-employment continued its steady growth pattern unchanged. There was no disruption to overall labor market composition.

This reinforces a broader point: independent work and traditional employment are not in conflict—they coexist and often complement one another. Most workers still rely on W-2 employment, while many also engage in independent work as a supplement or alternative depending on their needs.

Portable benefits fit within that reality.

They are complementing, not replacing, traditional employment—expanding access to benefits without reshaping or undermining the existing labor market.

What This Means

Taken together, these developments point to something bigger than a single policy reform.

We’re watching the early stages of a structural shift. States are experimenting and codifying portable benefits frameworks in real time, pilot programs are demonstrating proof of concept, and federal policymakers are beginning to take notice. If the past three years were about state-level innovation, the next phase is likely national—where a federal safe harbor could allow these models to scale across industries and states.

But even without federal action, the direction is clear.

Portable benefits are no longer a niche idea—they are becoming a standard part of modern labor policy.

For decades, the U.S. benefits system forced a tradeoff: flexibility or security. Portable benefits are beginning to break that tradeoff.

And what’s most striking isn’t just the idea itself—it’s the speed at which it’s spreading. What started as a fringe concept is now moving state by state, law by law, into the mainstream of American labor policy.

Subscribe now

Helen Andrews, a conservative commentator and author, argues that rising female participation is reshaping how institutions function. The mechanism, in her account, runs through norms. As gender composition shifts, so do the informal rules governing conflict, debate, and internal competition.

Institutions become more conflict-averse, more focused on consensus, less tolerant of the kind of blunt internal contestation that she associates with high performance. Over time, she argues, this gradually degrades how institutions execute their core functions. That is a causal claim, and it has observable implications.

Two empirical tests of that claim happen to be available. The first looks at moments when female participation shifts rapidly and institutional performance is easy to observe. A war economy – where objectives are explicit, output is measurable, and failure is catastrophic – provides the sharpest possible version of that test.

The second looks at the long run. If feminization erodes institutional effectiveness, the decades-long movement of women into high-skill professions should leave a productivity footprint: the institutions Andrews worries about most should show signs of declining output as female share rose. We have evidence on both.

Neither test was designed to answer Andrews’ question. But both speak to it. And the results are not what her framework predicts.

A war economy is an allocation problem

A modern industrial war is the most unforgiving kind of macroeconomics. Countries must convert finance into output while operating under tight constraints: steel, shipping, skilled labor, factory capacity, logistics. Strategy matters, but much of wartime success comes down to something simpler. Institutions have to allocate resources toward the sectors that produce war materiel.

A neat paraphrase of that logic: money wins wars; wars redistribute talent; talent makes money.

That’s a useful frame for what follows. When Andrews argues that feminization changes how institutions behave, she is implicitly telling us which variable she thinks belongs on the right-hand side of the institutional performance equation¹ – that is, which factor she believes is driving institutional outcomes. A war economy is where that claim meets its hardest test.

What Andrews is claiming

Andrews begins from an observation that is hard to deny: women now occupy far more of the professional and institutional terrain than in the past, and she asks why we assume institutions will function under these “novel circumstances.”

She is careful, in one important way. She does not rest her argument on women being less talented. She argues instead that “female modes of interaction” are not well suited to the goals of many major institutions – academia, corporate life, politics – where she associates institutional success with open debate, competitive striving, and truth-seeking.

Set aside whether those are the only goals of these institutions. The analytic move is that she treats gender composition – or what she thinks gender composition does to norms – as a structural determinant of performance.

You might think this is a semantic point. It isn’t. It changes what evidence we should care about.

If composition is the primitive, then rising female share should predict institutional degradation in settings where performance is observable and stakes are high. WWII is exactly such a setting.

The numbers that moved

In 1940, the U.S. civilian labor force included about 13 million women, roughly a quarter of all workers. By July 1944, that number had risen by more than 6 million – nearly half the 1940 level – with over nine-tenths of the new entrants in non-agricultural work. By July 1945, women were 35% of civilian workers. Those are not marginal adjustments. They are the kind of labor supply shock that normally only appears in textbooks in hypothetical settings – except here it happened in a handful of years, driven by mobilization.

Manufacturing is where the constraint was tightest, so it is the most relevant sector to examine. Between 1940 and March 1944, women’s employment in U.S. manufacturing rose by more than three million, and women’s share of total female employment in manufacturing climbed from 21% to 34%. This was not clerical work or supporting roles at the edges of production. It was entry into the sectors producing war materiel – the sectors where the output function was most clearly defined.

Britain faced tighter constraints. By 1944, over seven million women were engaged in war work. The Essential Work Order of March 1941 tied workers to essential jobs and restricted dismissals without Ministry of Labour approval. By early 1941, women aged 18 to 60 were required to register for war work; conscription followed.

Economists Acemoglu, Autor, and Lyle show that this was not coincidence. Exploiting variation in WWII mobilization rates across U.S. states, they find that higher mobilization rates were associated with larger increases in women’s labor supply – consistent with a supply response to male labor withdrawal rather than slow-moving cultural change.

At this stage, a sympathetic reader might say: of course, women worked. Wars force everyone to work. The question is whether institutions stayed good at what they were supposed to do. That is the right question – and it forces us to be precise about the mechanism.

What the causal model should look like

Andrews’ thesis puts gender mix in the driver’s seat: change the composition of the workforce, and the norms change with it. But WWII inverts that story. Institutions first changed their objectives and restructured incentives around binding constraints. Women entered because the demand for their labor shifted -- not because norms did. Composition was the outcome, not the input.

In Britain, the Essential Work Order exists precisely because turnover, misallocation, and poaching are performance problems in wartime. In the U.S., the wartime expansion of women into manufacturing and war industries is not explained by a sudden national embrace of feminist workplace norms; it is explained by a binding scarcity of labor in the sectors that produced war materiel.

You might think: perhaps wartime institutions performed despite feminization, not because of it. But that concedes the key analytic point. If “female modes of interaction” are structurally incompatible with institutional goals like output maximization, open contestation, and high-tempo execution, the setting in which those incompatibilities should bite hardest is a war economy – when “safetyism” is a luxury. Yet the institutional response was to widen women’s participation, not restrict it.

That doesn’t prove institutions are always gender-neutral machines. It does tell us that gender composition is rarely the primitive variable. Incentives and constraints are.

The part we should not overstate

This is not an argument that wartime feminization produced permanent transformation, or that WWII proved gender to be irrelevant to all institutional dynamics.

The postwar reversion was real and swift. From July 1945 to January 1946, the Women’s Bureau recorded a drop of nearly four million women from the labor force. Economist and Nobel Laureate Claudia Goldin’s retrospective work finds that almost half of wartime female entrants had exited by 1950, and concludes that WWII’s direct influence on women’s long-run employment was “considerably more modest” than popular narratives suggest. Economist Evan Rose’s work confirms limited lasting effects on female participation by 1950 in the most mobilization-exposed areas.

Andrews might read this as confirming her view: the wartime feminization was temporary because it was externally imposed, and when the coercion lifted, equilibrium re-sorted back toward the “natural” gender distribution.

But this reads the evidence backwards. The reversion is explained by the same causal model we have been using: when the constraint relaxed and the objective function changed – when the government stopped prioritizing war production and started prioritizing veteran reabsorption – the labor market equilibrium shifted. Women’s participation moved with those incentive changes, not against them. That is not evidence of an underlying female preference for domestic life reasserting itself over institutional resistance. It is evidence that labor supply responds to incentive structures, which is the whole point.

The war test asks a narrower question than the postwar numbers do: under maximal stakes and tight constraints, could institutions still function with rapidly rising female participation? The answer is yes, unambiguously. They did not collapse into conflict-avoidance.

But that wartime result should be read carefully. It does not tell us that institutional norms are irrelevant, or that long-run cultural change never matters. It tells us something narrower and more useful: gender composition alone was not enough to prevent institutions from meeting demanding, measurable objectives. Under wartime pressure, performance turned on incentives, constraints, and efficient allocation of labor.

A study that helps clarify the mechanism

The wartime evidence tells us what happened under pressure. A 2019 study by economists Hsieh, Hurst, Jones, and Klenow tells us why the expansion of women into high-skill work continued afterward and what drove that long-run shift.

1. Falling barriers, not changing norms, did most of the work.

HHJK explicitly decompose occupational convergence into two potential drivers: declining frictions, such as labor market discrimination and barriers to human capital, and changing preferences. Their finding (Table V, column 2 vs. column 1): changing occupational preferences “explain little of U.S. growth during this time period.” The discrimination and barriers to human capital alone explain 41.5% of market GDP per person growth; add preferences in and you get 40.8% – preferences contribute essentially nothing and may slightly reduce growth.

Andrews’ explanation is fundamentally about norms. HHJK point to a different mechanism. Much of the occupational convergence she reads as feminization looks, in their framework, like the removal of barriers to comparative-advantage allocation. In the baseline decomposition, preferences contribute little to growth, while falling frictions do most of the work. That does not rule out every role for norms. But it does suggest that the large movement of women into elite occupations is better understood, in economic terms, as a correction of misallocation.

2. The highest barriers were in exactly the institutions Andrews worries about.

Figure 4 in HHJK is one of the strongest exhibits. The vertical axis shows the size of the barrier women faced when entering each occupation. Higher values mean larger barriers. In 1960, the barriers were very large in professions like law and medicine. Over the following decades, they fell sharply. Construction remained high over time, which the authors note could reflect comparative disadvantage in physical-strength tasks.

More broadly, HHJK find that large barrier reductions were concentrated in high-skill occupations, and in robustness checks they say much of the gain comes from women moving into lawyers, doctors, scientists, professors, and managers. These overlap closely with the institutions Andrews worries about. HHJK do not measure “deliberation” or “truth-seeking” directly. But they do show that as barriers in these professions fell and participation broadened, the resulting reallocation was associated with higher aggregate market output, not lower.

A different diagnosis

This does not prove that every institution improves with every increase in female participation. It frames the burden of proof. A thesis that treats rising female participation as inherently corrosive has to argue that the performance gains from tighter “masculine” norms exceed the output lost to misallocation. Andrews does not state that trade-off in those terms. Her essay is not a growth accounting note. It is a diagnosis of institutional dysfunction, with gender composition as the upstream driver.

The WWII stress test pushes us toward a different diagnosis. The allocation evidence points the same way. When institutions fail, start with incentives, constraint management, and selection. Do not assume the gender mix is the primitive variable.

The modern institutional failures Andrews points to are real. Conflict avoidance, bureaucratic risk-aversion, the disappearance of frank internal debate – these are genuine problems worth diagnosing. But the diagnosis that follows from the WWII case, and from the long-run evidence on talent allocation, is not “too many women.” It is “misaligned incentives and blurred objectives.”

That is a more tractable problem. It is also, I think, the right one.

¹ In economic models, the variable being explained is usually written on the left-hand side of an equation, while the factors believed to determine it appear on the right-hand side.

Subscribe now

In 2023, the contract between UPS and the International Brotherhood of Teamsters, one of the world’s largest private sector unions, was widely described as historic. The agreement delivered large pay increases, expanded benefits, and introduced new work rules governing how labor is scheduled, deployed, and compensated. It was heralded as a turning point for workers in the shipping and logistics sector. For many observers, it appeared to demonstrate that aggressive bargaining could reverse years of stagnant wages and restore labor’s leverage.

Two years later, UPS is in the midst of a sweeping restructuring.

Since that contract, the company has eliminated 48,000 operational jobs, announced plans to cut another 30,000 positions, and closed or consolidated more than 100 facilities. Executives describe the effort as a necessary “right-sizing” of the business, driven by lower package volumes, higher operating costs, and a strategic shift away from less profitable delivery segments. As UPS’s chief financial officer has put it, “With less volume, we need fewer positions in order to support that volume.” At the same time, UPS is accelerating investments in automation and other capital-intensive operations, explicitly aiming to build a less labor-intensive delivery system.

It is true that the 2023 contract delivered meaningful gains. Many UPS workers now earn higher pay and enjoy improved benefits. But that is rarely the end of the story.

The question, then, is not whether the gains are real, but how the trade-offs unfold. Why do headline-grabbing contracts so often coincide with downsizing, automation, and job losses in sectors governed by exclusive, monopoly bargaining arrangements? When short-run wage gains are secured through monopoly bargaining power, where do the adjustments occur—and who ultimately bears the costs?

The evidence suggests that this pattern is not accidental, but a structural feature of monopoly bargaining. Our recent study (with Revana Sharfuddin), Do More Powerful Unions Generate Better Pro-Worker Outcomes?, helps explain why the sequence now unfolding at UPS is not an anomaly. Drawing on 147 studies, the paper shows how monopoly union power tends to shift costs into the future, where they often appear as reduced employment, lower investment, and faster automation—often to the detriment of workers over time. This suggests that improving long-run worker outcomes requires rethinking not worker voice itself, but the monopoly structure through which it is exercised.

Ultimately, this is not a story about unions versus employers, nor a judgment about intent or good faith. It is a story about institutional design and the incentives it creates. Specifically, it is about how systems of exclusive, monopoly representation shape bargaining behavior and firm responses in ways that can produce short-run gains for some workers while weakening long-run employment and opportunity for others.

What’s happening at UPS

According to UPS’s chief financial officer, Brian Dykes, the company is now in the most difficult phase of a large-scale restructuring. With lower package volumes, UPS is consolidating buildings, reducing hours, and eliminating positions. As Dykes explained, “When we started out with the labor contract, we knew that we were going to be investing in order to create a lower labor-intensive network.”

The numbers are stark. When the contract was ratified in 2023, UPS employed approximately 340,000 Teamsters-represented workers. Since then, the company has announced roughly 48,000 job cuts, with another 30,000 planned. UPS describes these reductions as roles concentrated in its logistics and delivery network, though it has not specified how many of those positions are covered by the Teamsters contract.

If the reductions were concentrated primarily among union-represented workers—and not offset by additions—the bargaining unit could theoretically be as low as roughly 290,000 today. There is currently no public confirmation that this is the case. Even so, if only a portion of the announced cuts affected Teamsters positions, the scale of workforce adjustment is substantial relative to the size of the 2023 bargaining unit.

At the same time that UPS appears to be shrinking parts of its workforce, it is also investing in automation and other capital-intensive operations.

From an economic perspective, this adjustment is unsurprising. The 2023 agreement significantly increased labor costs and altered work rules, while even the threat of a strike disrupted business activity. Firms respond to sustained cost increases and lost volume in predictable ways.

Of course, there were real gains for workers who remain employed. But when the marginal cost of labor significantly rises without a corresponding increase in productivity, firms adjust—by reducing headcount, accelerating automation, or reshaping their business models.

Isolating a single causal factor behind job losses is always difficult. But several pieces of evidence suggest that the 2023 contract—and the strike threat surrounding it—likely contributed to UPS’s subsequent adjustments.

First, timing. At the peak of the summer 2023 negotiations, roughly 1.5 million packages per day were diverted from UPS amid strike concerns. By October, only about 600,000 had returned, and FedEx reported retaining approximately 400,000 daily shipments. Executives also attributed roughly $1 billion in lost revenue to the disruption.

Second, the composition of losses. UPS deliberately shed its largest customer, Amazon, cutting volumes by roughly 50 percent because those shipments were no longer sufficiently profitable under the new labor-cost structure. This was not a general market contraction; it was a strategic response to changed economics.

While macroeconomic headwinds affected all logistics firms, UPS lost volume faster and recovered more slowly than competitors facing similar conditions. Its post-contract cost structure changed the profitability of certain routes and customers in ways its competitors did not face.

The margins firms adjust on

A central lesson from labor economics is that firms rarely respond to higher labor costs by simply accepting lower profits indefinitely. Instead, they adjust along several margins:

• Employment: reducing headcount, limiting new hires, or increasing attrition

• Capital substitution: investing in automation and technology

• Scale and scope: shrinking operations, closing facilities, or exiting lines of business

• Location: shifting activity geographically

This is exactly what we observe at UPS. The company has shed less profitable business, closed facilities, reduced staffing, and accelerated automation.

A similar pattern appears in the historical record. Research on the decline of Rust Belt manufacturing from 1950 to 2000 finds that powerful unions and frequent labor conflict played a significant role in the region’s employment losses—more so than globalization in the early decades. Wage premiums persisted even as investment slowed and firms gradually shifted operations elsewhere. When labor costs significantly rise without corresponding productivity gains, firms adjust over time. The Rust Belt shows this dynamic unfolding over decades; UPS illustrates it in real time.

What 147 studies tell us

Our paper reviews 147 studies examining the relationship between union power, employment, investment, productivity, and firm performance across the U.S. and Europe. The results are strikingly consistent.

Unions operating under exclusive, monopoly representation tend to raise wages and benefits in the short run. But those gains are frequently accompanied by:

• Slower employment growth

• Lower investment and capital formation

• Reduced profitability and productivity growth

• A higher likelihood of downsizing, automation, or firm exit

In other words, monopoly bargaining power often shifts costs into the future, where they appear as fewer jobs, fewer opportunities for marginal workers, and greater exposure to technological substitution.

This raises a recurring question: whether union “victories” can sometimes be illusory—real gains today that obscure longer-run costs borne by workers tomorrow. The evidence suggests this trade-off is not incidental; it is a structural feature of monopoly bargaining.

Why monopoly representation matters

The issue is not union representation. Worker voice is essential. The issue is how representation is structured—and how concentrated bargaining leverage within a monopoly framework can shape long-run firm and employment outcomes.

Under U.S. labor law, a single union typically becomes the exclusive representative for all workers in a bargaining unit. Workers cannot choose alternative representation or negotiate individual arrangements outside the collective framework, as most non-union workers do. Once certified, the union faces limited competitive pressure.

This monopoly structure shapes incentives in important ways:

• Bargaining emphasizes immediate compensation over employment margins

• The costs of aggressive bargaining are often delayed and concentrated in future employment adjustments

• Marginal workers—new entrants, part-timers, and lower-seniority employees—are the first to lose

The UPS case illustrates this clearly. When the 2023 contract was ratified, the Teamsters represented roughly 340,000 UPS workers. While the company has not disclosed how many of the announced job cuts affected union-represented roles, if a substantial share were concentrated in the bargaining unit—and not offset by new hiring—the number could be closer to 290,000 today.

A pro-worker path forward

The evidence points to a more constructive alternative. In labor markets where worker representation is competitive or pluralistic, outcomes tend to be more balanced—with less extreme swings between wage gains and employment contraction. Workers retain voice and bargaining power, while firms maintain flexibility to invest and grow.

Some European systems separate worker voice from wage bargaining, allowing multiple forms of representation within firms. These systems reduce winner-take-all dynamics and better align compensation with productivity and long-run firm performance.

Competition disciplines behavior—for firms and for worker representatives alike.

If policymakers want to improve long-run worker outcomes, the focus should shift from strengthening monopoly bargaining power to expanding worker choice and representation options.

One reform could do this directly: allow employees who opt out of union membership to negotiate directly with their employers, while union members remain covered by collective agreements. Workers who opt out would also opt out of union representation—eliminating free-rider concerns and forced monopoly representation. This resolves a long-standing tension in labor policy by eliminating both forced representation and free-rider concerns.

Another reform would be to apply antitrust principles more consistently in labor markets. Simple rule changes—such as limiting excessive union market share or clarifying that a single union should not represent multiple firms whose merger would violate antitrust law—would help prevent the entrenchment of labor monopolies while preserving worker voice.

Such reforms would strengthen worker voice by making representation more responsive, more accountable, and more compatible with both wage growth and employment stability.

UPS is the symptom

UPS is not an outlier. It is a case study in how monopoly bargaining can generate short-run wins that give way to long-run adjustment costs. The layoffs, automation, and restructuring now underway are not a repudiation of workers’ worth. They are the predictable response to institutional incentives.

If we want labor markets that deliver both higher pay and durable jobs, we need to rethink how worker representation is structured.

Subscribe now

Something that strikes me in much of the AI‑and‑work debate is how narrowly it is framed. The dominant question tends to be whether AI will replace workers in existing jobs, and if so, in which ones. Even the more nuanced task-based approaches focus primarily on which tasks AI can perform. What gets far less attention is whether AI changes how work is organized in the first place.

From an economic perspective, that omission matters. Technological change doesn’t just substitute capital for labor—it also reshapes the boundary between firms and markets. And on that margin, AI may be exerting a transaction‑cost shock that quietly expands self‑employment and contract‑based work across a much wider set of occupations than we usually associate with the “gig economy.”

This is not a claim that AI is destroying jobs. Rather, it is an analytic observation about how AI may be lowering the costs of operating outside traditional employment relationships—and why that possibility deserves much more empirical attention than it has received so far.

Beyond Substitution: An Organizational Lens on AI

Ronald Coase’s The Nature of the Firm offers the key insight: firms emerge when the costs of using markets such as finding contractors, negotiating terms, ensuring performance exceeds the costs of organizing production internally.

When technology reduces those costs, the logic reverses. Work that once needed to be organized inside firms can instead be purchased on the market.

This mechanism has been central to understanding the rise of freelancing, contracting, and platform‑mediated work over the past decade. In earlier work with my co-author Seth Oranburg, “Transaction Cost Economics, Labor Law, and the Gig Economy” (published in the Journal of Legal Studies), we built on Michael Munger’s book Tomorrow 3.0 to identify four specific transaction costs—triangulation, transfer, trust, and measurement—that explained how digital platforms enabled peer-to-peer transactions that previously didn’t happen. Uber worked because it reduced the costs of consumers finding drivers, transferring payments, and trusting strangers with rides.

AI operates through similar transaction cost logic, but with different mechanisms dominating. To understand its effect on work organization, we need to examine the costs that determine firm boundaries.

1. Production and Measurement Costs (The Alchian-Demsetz Framework, Extended)

The core insight from economists Armen Alchian and Harold Demsetz is this: when output requires interdependent team production and you can’t tell who contributed what, you need a firm with a manager. The firm monitors inputs, hours worked, effort expended, as proxies for output. But when individual contributions are separable and measurable, you can more easily contract for the output directly.

AI affects this margin in two ways: it makes individual contributions more separable when work is collaborative, and it reduces the need for collaborative production in the first place.

First, AI improves measurement and separability in collaborative work. Consider software development. Modular programming already made coding more contract-friendly by creating independent pieces that could be evaluated separately. You could verify whether a specific module worked without needing to understand the entire system. AI development tools push this further, not just in how code is structured but in how it’s specified, tested, and verified. Automated testing can validate each module independently. For example, AI tools can automatically record who proposes a change, what problem it’s intended to solve, and whether it passes standardized tests—creating a clear audit trail that ties individual contributors to specific outputs. This makes individual contributions more transparent and verifiable even in team contexts.

Second, and perhaps more significantly, AI substitutes for complementary roles that previously made team production necessary. This is about reducing the minimum efficient scale of production. Historically, firms existed partly to efficiently bundle complementary capabilities: you need analysis plus editing plus design plus formatting to produce a deliverable report. Individuals working independently couldn’t access all these capabilities easily.

AI changes this by providing many complementary services directly to individual workers. An independent consultant can now produce client-ready reports with AI assistance for data visualization, editing, and formatting, which previously required support staff at a consulting firm. A policy researcher who once needed a think tank’s editorial review, design staff, and institutional publication channels can now use AI tools for editing and professional formatting, while publishing directly through Substack or personal websites reaches audiences without institutional backing.

Large-scale survey research demonstrates this clearly. Traditionally, it required firms with teams of interviewers, coders, and analysts working in coordination. Commercial AI interview platforms like Outset and Listen Labs already enable researchers to conduct hundreds of simultaneous AI-moderated interviews with automated follow-up probing and analysis.

Anthropic’s AI Interviewer—currently an internal research tool—demonstrates that the underlying technology continues to advance in this direction, with systems capable of conducting thousands of in-depth interviews using standardized prompts, automatically transcribing and coding responses, and producing analyzable data at scale.

As these capabilities mature and diffuse into commercially available tools, a solo researcher will be able to execute projects that previously required a research firm’s infrastructure. The independent researcher delivers large-scale results without relying on the firm’s bundle of complementary labor inputs.

For workers, both channels create new possibilities. Better measurement means your work speaks for itself—clients can evaluate what you’ve actually produced rather than relying on someone else’s assessment of your effort. But the substitution effect might be even more important: you can now produce complete, polished deliverables independently when you previously needed a team’s complementary skills. A consultant delivers client-ready analysis with professional formatting and visualization. An analyst produces publication-quality research without needing editorial and design support.

Digital platforms reduced search and payment friction, but AI represents something different: it changes which tasks can be evaluated at all. When AI can directly assess knowledge work quality that previously required managerial judgment, the boundary between what needs firm-based monitoring and what can be contracted shifts fundamentally.

This connects directly to research I did with Paola Suarez, “Employee vs. Independent Worker: A Framework for Understanding Work Differences” (published by the Mercatus Center), examining how work characteristics differ between independent and traditional employment. We found that independent work relies significantly less on:

• Team production and coordination

• Supervision and responsibility for others’ outcomes

• Working in interdependent groups

Figure 1 from our prior research shows these differences clearly—independent work scores lower on precisely the dimensions that favor firm‑based employment.

AI expands the frontier of viable independent work through both better measurement and reduced need for teams.

2. Asset Specificity and Relationship-Specific Investment (The Williamson Framework)

The second mechanism comes from economist Oliver Williamson’s work on asset specificity. When workers invest heavily in learning firm-specific systems, processes, and relationships, they become locked into employment relationships. Those investments are worthless if they leave, creating switching costs that favor long-term employment over independent work.

AI can reduce these relationship-specific investments by standardizing tools and making knowledge more portable. When workers can use the same AI development environments, analytical tools, and knowledge systems across different clients, their skills travel with them.

This standardization effect depends critically on whether AI tools themselves become standardized across firms or remain proprietary. If firms develop sophisticated internal AI systems that require significant learning investments, this could recreate the lock-in dynamics—a possibility I’ll return to in the countervailing forces.

A worker proficient in ChatGPT, Claude, GitHub Copilot, or standard AI-assisted workflows can move between projects more easily. The training and expertise they build aren’t tied to one employer’s proprietary systems.

This portability works in both directions. It’s easier for firms to bring in contractors who can hit the ground running with standardized tools. And it’s easier for workers to maintain independent practices because their capabilities aren’t locked into any single firm’s ecosystem. When skills are more general-purpose, workers face lower risks in operating independently.

3. Search and Specification Costs (The Basic Coasean Framework)

The third mechanism is more incremental because digital platforms already did much of this work. But AI provides some additional improvements in helping both sides find each other and define what needs to be done.

AI-powered platforms can better match workers to appropriate projects and help both sides define deliverables more precisely. For workers, AI tools can analyze market rates to help price services competitively and can translate vague client requirements into clear project specifications before commitment. This reduces information asymmetry—workers know what’s expected and what similar work commands in the market. Less ambiguity means fewer disputes and more confidence in taking on independent work.

But honestly, LinkedIn, Upwork, and similar platforms already reduced these costs substantially. AI’s contribution here is real but incremental. This mechanism matters less than measurement costs and asset specificity.

A Second Thought: Countervailing Forces

Before going further, we should consider what might work against this thesis.

Returns to scale in AI deployment. Large firms may have advantages in building and deploying sophisticated AI systems. Firms with proprietary datasets or specialized AI might make employees more productive than those same workers operating independently. When workers invest in learning firm-specific AI tools, this creates the relationship-specific investments that favor employment—the opposite of standardization.

Input monitoring could favor employment. Here’s an interesting wrinkle from the measurement cost literature: if AI makes it easier to monitor inputs (like real-time tracking of what workers are doing, keystroke monitoring, or attention tracking), this might actually push firms toward employment rather than contracting. When you can cheaply monitor whether employees are working effectively, the traditional disadvantage of not being able to measure output directly becomes less costly. This could reinforce employment relationships in some contexts even as output measurement improves in others.

Coordination complexity in AI-augmented work. Some AI applications might increase rather than decrease the value of tight coordination and co-location. If AI creates new forms of complex interdependencies, this could favor team-based employment over contracting.

These countervailing forces are real. The net effect of AI on firm boundaries will depend on which mechanism dominates in different occupations and industries. But they don’t negate the core mechanisms—they constrain how broadly and deeply the shift toward independent work can go.

The Empirical Questions We Should Be Asking

The theoretical logic is clear: AI reduces several transaction costs that historically favored firm-based employment. What we don’t know yet is whether this is actually reshaping how work gets organized and if so, how broadly and in which directions.

The mechanisms outlined here push toward expanded independent work, but they won’t operate uniformly. Firms with sophisticated AI infrastructure may create stronger productivity gains for employees than those same workers could achieve independently. The shift isn’t universal or inevitable—it’s conditional on occupation, industry, and the specific AI applications involved.

There are concrete empirical questions we could examine. Are AI-intensive occupations seeing faster growth in self-employment? Are workers in fields with high AI adoption more likely to transition from employment to contracting? Do occupations that already exhibit low coordination requirements and high output separability—the “independent work profile”—adopt AI faster?

These questions are empirically tractable, though data limitations exist. Yet I haven’t seen them seriously examined in the AI-and-work literature. Getting answers matters because the implications extend well beyond understanding labor markets—they shape how we design the institutions that will govern work in an AI-augmented economy.

Why This Matters (Even Without All the Answers Yet)

Much of our social insurance system assumes people work as W-2 employees for single employers. If AI makes independent work more productive and feasible—not by pushing people out of jobs but by making contract-based arrangements genuinely more efficient—then we face growing institutional mismatches.

When someone works on projects for multiple clients simultaneously, employment-based health insurance stops making sense. Labor law built around measurable work hours becomes difficult to apply. Labor policy scholars Seth Harris and Alan Krueger have argued this challenge is fundamental for independent workers: minimum wage protections may simply not fit when work hours cannot be reliably measured or attributed to specific intermediaries.

AI seems to intensify this challenge: when a developer uses AI to compress three days of work into six hours while juggling multiple clients, what would applying hourly wage floors even mean? The question becomes not just administratively difficult but conceptually puzzling—you’d need to determine whose ‘hours’ are being worked when the traditional relationship between time, effort, and output has fundamentally shifted.

The portable benefits conversation has moved from think tank papers to actual state pilots and federal proposals. That momentum reflects a growing recognition: safety nets built around W-2 employment don’t work well when work happens across multiple clients and projects. If this AI-driven reorganization accelerates, the mismatch between how we organize social insurance and how work actually gets done will only widen.

A Different Kind of Reorganization

AI may be changing the boundary between firms and markets—and we’re not paying nearly enough attention to this margin. The dominant frame in AI-and-work discussions is substitution: will AI replace workers or augment them?

But Coase taught us that the firm-market boundary is endogenous to transaction costs. When those costs shift, work organization shifts too. Whether this produces a major labor market reorganization remains uncertain—evidence is developing and effects vary by occupation. But the theoretical mechanisms warrant empirical attention now, not after the fact.

The conversation needs to expand beyond “will AI take my job?” to “how will AI reorganize how we work?” That question will shape not just labor markets but social insurance, labor regulation, and economic measurement for the economy we’re building.

Final Note: If you’d like to read the full JLS paper but can’t access it through the paywall, just drop a comment or send me a message—happy to share the published version.

Subscribe now

If you talk to job seekers right now, a common complaint is: “It’s not that everyone is getting laid off—it’s that nothing is moving.” Hiring feels slow. Offers feel scarcer. And yet we don’t see the classic recession headline: mass layoffs, unemployment spiking, separations exploding.

It only looks like a contradiction at first glance. It is, in fact, a coherent labor‑market regime—one that labor economists often describe as low churn (few hires, few quits, few layoffs) and that, in steady‑state terms, can look “stable” in the aggregate while becoming less opportunity‑rich underneath.

From mid‑2025 into early fall, the headline data increasingly fit this story: unemployment drifted up, but the engine of job loss did not shift into high gear; instead, much of the softness showed up as subdued hiring and slower reallocation—a labor market that looks close to a steady‑state, low‑churn equilibrium.

The key idea I want readers to take away is simple:

A labor market can cool mainly by becoming less “opportunity‑driven” (fewer hires, fewer quits, fewer openings) rather than more “separation‑driven” (more layoffs). That kind of cooling can raise unemployment gradually—without producing the layoff wave that defines a recession.

I start with the intuition and then provide a disciplined read of the late-2025 data.

Steady state in labor economics: the “stocks vs flows” way to think.

Most headlines fixate on stocks: the unemployment rate, the number employed, the payroll level. Labor economists care at least as much about flows: How many people are losing jobs (separations, especially layoffs)? How many people are finding jobs (hires/job‑finding)? How many people are moving voluntarily (quits)?

A good mental model is a bathtub:

• The water level is the unemployment rate (a stock).

• The faucet is inflows into unemployment (separations).

• The drain is outflows from unemployment (job‑finding).

In a steady state, the bathtub level is roughly constant because the faucet and drain balance.

A one‑line “sidebar equation” (intuition, not derivation)

A standard steady‑state approximation is:

u ≈ s / (s + f)

Where:

• u is the unemployment rate

• s is the separation (job‑loss) rate into unemployment

• f is the job‑finding rate out of unemployment

Read it like this: unemployment is high when separations are high or job‑finding is low.

So here is the crucial point for late‑2025:

You do not need a surge in layoffs (s) to see unemployment drift up. If hiring slows and job‑finding probability falls (f declines), unemployment can rise gradually even with stable separations.

That’s the “no‑hire, no‑fire” logic in steady‑state form.

(Technical footnote: the model object f is an unemployed‑to‑employed transition hazard. The JOLTS hires rate is not the same object, but it is an empirically useful indicator of how “opportunity‑rich” the market is.)

What a Hiring Freeze Looks Like in 5 Charts

A brief housekeeping note: the 2025 federal government shutdown caused a brief CPS disruption, concentrated in one or two months. I’ll focus on multi‑month trends and cross‑check CPS (unemployment, participation, EPOP) against JOLTS (openings, hires, layoffs). The signal in late‑2025 is a market cooling via slower hiring and weaker churn, not a surge in layoffs.

1) Hiring is trending down, while layoffs remain low

This is the cleanest “low‑churn” picture. Over 2022–2025, the hires rate drifts down while layoffs/discharges remain low and relatively stable. The market is cooling primarily on the job‑creation / match‑creation margin, not the job‑destruction margin.

2) Job openings have normalized from extraordinary highs

Openings soared in 2021–2022 post pandemic, then moved down markedly. By 2024–2025 they look more like a market that is no longer overheated—but also not collapsing. That matters because openings are the labor‑demand side of the matching story: fewer openings generally means fewer opportunities.

3) Unemployment drifted up, not spiked

Unemployment rises gradually across 2023–2025 rather than jumping the way it typically does when layoffs surge. That pattern is consistent with a fall in job‑finding, not a sudden wave of job destruction.

4) Participation is not doing the explanatory heavy lifting

Participation is relatively stable in this window (with modest month‑to‑month movement). That makes it harder to explain the unemployment drift as “just” labor force entry. In a steady‑state lens, that pushes you back toward the idea that job‑finding is getting harder at the margin.

5) The Employment‑Population Ratio signals a hiring freeze

The employment-population ratio (EPOP) is the share of the population that’s working. It’s a simple way to see whether employment is still expanding relative to population. The figure above shows a clear post-COVID arc: a sharp collapse in 2020, a strong rebuild through 2022, and then a long plateau around ~60. The key late-2025 signal is the final leg: EPOP drifts down through 2024–2025, indicating that employment growth has slowed relative to population.

Put that next to unemployment inching up and you get a coherent read. This looks like a hiring-freeze phase—slower job creation and fewer new matches—rather than a labor market breaking via mass layoffs.

Why low hiring and low firing can coexist

From a search‑and‑matching perspective (Mortensen‑Pissarides), the labor market is a market for new matches as much as a market for existing jobs. Job creation and job destruction are endogenous and can respond differently to shocks.

Firm‑side caution: fewer postings, slower backfilling

When firms are uncertain about demand, costs, financing conditions, or policy, a common response is not to lay people off immediately. It’s to: post fewer vacancies, take longer to fill them, not backfill departures quickly, reallocate internally rather than expand headcount. That reduces hires without requiring a spike in layoffs.

Worker‑side caution: fewer quits, fewer job‑ladder moves

A big part of churn is voluntary. When quits fall, job ladders slow down: fewer voluntary exits means fewer openings created by churn, which means fewer hires.

Matching and reallocation: low churn can mean less dynamism

Low churn isn’t automatically good. It can mean fewer mismatches get resolved, fewer productivity‑improving reallocations occur, and outside options weaken—pressuring wage growth and mobility.

Why this isn’t a classic recession pattern—yet

The canonical recession labor market is separation‑driven: layoffs rise, job losses broaden, unemployment rises quickly.

What your JOLTS figure shows is closer to a hiring slowdown than a layoff shock: layoffs/discharges remain low by recent historical standards, while the hires rate has been drifting down.

That’s why the honest read is: softening, not breaking.

But “not a recession” is not the same as “healthy.” And that’s where composition and opportunity matter.

Relative to 2015–2019, the Beveridge curve shifted outward after 2020—openings were unusually high even at low unemployment—consistent with elevated matching frictions. Since 2022, the market has moved down/right as hiring cooled, and the curve appears to be edging inward, though it still looks more vacancy-heavy than the 2015–2019 baseline.

The Beveridge Curve says this is a hiring slowdown more than a firing shock)

One of the cleanest ways to see whether a labor market is cooling through fewer new matches (vacancies/hiring) versus more job destruction (layoffs/unemployment) is the Beveridge curve—a scatter/line plot of the job openings rate (vertical axis) against the unemployment rate (horizontal axis).

It’s usually downward-sloping: when the economy is hot, openings are high and unemployment is low (upper-left). When the economy cools, openings fall and unemployment rises (down-and-right). The key is how you move.

The Beveridge curve makes the shift easy to see. For example in March 2022, openings were 7.4% with unemployment at 3.7%—a market defined by abundant opportunities and fast churn. By November 2022, openings had already cooled to 6.4% even though unemployment was still 3.6%. Fast‑forward to November 2025: openings are down to 4.3% while unemployment is 4.6%. The headline change isn’t a sudden unemployment spike—it’s the steady drop in openings. A market that feels stuck even without mass layoffs.

That pattern is exactly what you’d expect from firm-side caution under uncertainty. Instead of immediately shedding workers, firms often: post fewer vacancies, slow down hiring, take longer to fill roles, and don’t backfill departures quickly.

In Beveridge-curve terms, that shows up first as a drop in vacancies—often before you see a dramatic jump in unemployment.

A second takeaway is about tightness: a rough proxy is vacancies per unemployed (V/U). For example, in March 2022, the chart implies roughly 2 openings per unemployed worker (7.4/3.7 ≈ 2.0). By September 2025, it’s closer to one-for-one (4.6/4.4 ≈ 1.05). That’s a big shift in bargaining power and outside options consistent with slower job-ladder moves, fewer quits, and softer wage pressure even without a layoff wave.

Finally, the post-2020 path in the figure sits above earlier “normal” decades at comparable unemployment rates. That’s often read as some combination of lingering mismatch/frictions and lower matching efficiency—another way of saying: even when there are openings, the market isn’t translating them into hires as smoothly as in the past. A low-churn environment can do that.

The elephant in the room: tariffs and a labor market stuck in wait-and-see mode

In a textbook cyclical downturn, the labor market usually “breaks” first on the match-destruction margin—layoffs and separations surge as firms unwind existing matches—so the fact that layoffs remain anchored is strong evidence that late‑2025 is not primarily a broad demand-driven recession shock.

Instead, the pattern fits an external cost/uncertainty regime that raises the option value of waiting, and the elephant in the room is tariffs: economists are overwhelmingly agreed that tariffs are borne largely at home through higher prices/costs (not “paid by foreigners”), as reflected in near‑unanimous expert-panel responses.

The most rigorous recent macro evidence for the U.S. finds that tariff shocks depress trade, investment, and output persistently—and reduce employment—while trade-policy uncertainty is contractionary as well. Consistent with your “no‑hire, no‑fire” mechanism, state‑level evidence shows that greater exposure to trade policy uncertainty reduces hours and employment mainly via the extensive margin, because firms postpone hiring.

And the last trade war’s micro evidence lines up: import protection did not deliver local job gains in newly protected sectors, while retaliatory tariffs produced clear employment losses—exactly the kind of environment that freezes vacancies and mobility without requiring a layoff wave.

What about AI?

I would be cautious about making “AI job destruction” the leading explanation for late‑2025 labor‑market softness. The empirical literature on generative AI so far reads more like within‑job task change and productivity effects than economy‑wide displacement.

More importantly for interpretation, if AI were already “stealing jobs” at scale, it should be visible in the labor market’s most basic destruction margin: layoffs. That’s where true displacement shows up first. The JOLTS data simply don’t show a large, broad-based rise in layoffs or discharges—suggesting that whatever AI is doing right now, it is not manifesting as an economy-wide wave of job destruction.

That doesn’t mean AI won’t matter. It means the most defensible claim from the 2025 flow data is: the labor market is slowing through hiring and reallocation, not through mass separations.

Bottom line: steady ≠ healthy

The labor market in late‑2025 looks increasingly consistent with a steady‑state, low‑churn regime: hiring is subdued, and firing is still low. That combination can keep unemployment from exploding—even as it drifts up—because the economy is adjusting through reduced flows rather than a surge in separations.

But a freeze is the opposite of dynamism.

A labor market can look calm because it’s moving slowly—like a river that isn’t turbulent because it isn’t flowing much at all. In that world, the key margin isn’t whether layoffs spike tomorrow; it’s whether the market keeps creating new matches, new opportunities, and upward moves.

The risk isn’t only recession. The risk is quiet stagnation: fewer ladders, weaker bargaining power, and a labor market that stays “steady” by offering less.

This is Part II of our two-part series (Part 1 here), where we examine employment patterns across labor market sectors, as well as ongoing administrative efforts at reducing the size and scope of the federal workforce.

When we look past the headline numbers, the real story of the labor market comes into focus. Job numbers have weakened lately, and even the strong ones often hide what’s really going on beneath the surface. Some industries are humming while others are shedding workers, and that reveals how both market forces and policy choices are quietly reshaping where opportunity lives—and where it’s drying up. These sectoral shifts matter most for young workers just entering the job market.

This second part looks at two big stories behind the numbers: the sectors driving recent job gains and losses, and the federal workforce reduction effort that’s been accelerating since January.

Subscribe now

Sectoral Shifts: Winners and Losers

Healthcare continues to dominate job growth—adding about 30,000 jobs in August and roughly half a million over the past year. This sectoral strength isn’t just a data point; it’s a window into understanding recent patterns in youth unemployment. John Burn-Murdoch at the Financial Times makes a compelling case that rising unemployment among recent graduates reflects largely a sectoral reallocation.

The gender divide is especially striking. Women are entering healthcare at high rates, and healthcare employment keeps expanding. Men, meanwhile, have clustered more in tech and computer-related fields—precisely where job losses have been creeping in. The data don’t tell us exactly which tech occupations are shrinking, but the overall pattern suggests that occupational choice—not just skills or credentials—is shaping labor market outcomes for young workers..

Manufacturing tells a different story. Employment there has been slipping for years, but the decline has quickened since April. In August alone, 12,000 factory jobs disappeared. Since “Liberation Day” (April), the sector has lost 42,000 workers—and 78,000 compared to a year ago.

One likely culprit: tariffs that have driven up input costs. More than half of U.S. imports—between 54% and 56%—are industrial supplies, materials, and capital goods used in production. A tariff on steel might sound like a win for steelmakers, but for every worker who produces steel, dozens more use it to build cars, appliances, batteries, and machinery. When input costs rise, those industries hire less, not more.

Together, these sectoral patterns show how market forces and policy decisions interact to reallocate labor across industries—boosting healthcare while constraining manufacturing and tech. Yet the federal government exerts its own influence on the labor market too, through its own hiring and staffing choices. And lately, those choices have tilted toward contraction.

Federal Workforce Reductions

Two weeks ago, the Office of Management and Budget sent agencies a memo to prepare reduction-in-force plans in case of a government shutdown. The move followed a funding standoff in Congress over additional healthcare spending.

Using the threat of layoffs as leverage isn’t exactly a model for thoughtful reform—but it’s part of a broader administrative push that began back in January. The federal workforce stood at a little over 2.4 million workers at the start of 2025. In the 25 years leading up to that point, it grew by roughly 550,000.

The idea of trimming government’s size isn’t new—or partisan. President Obama once called for a “leaner, smarter” government in 2012. And the last major workforce reduction came under President Bill Clinton in the 1990s.

Clinton began by ordering each agency to cut 4% of its civilian staff under his “Reduction of 100,000 Federal Positions” initiative. A year later came the Federal Workforce Restructuring Act, which offered voluntary buyouts—up to $25,000—to employees willing to retire or resign early. By the end of Clinton’s first term, the federal headcount had fallen by 331,000; by the end of his second, by 434,000.

Fast forward to today: the current administration’s approach echoes that era. In January, President Trump signed Executive Order 14210, aimed at downsizing government in two key ways:

• A hiring freeze: Agencies can hire no more than one employee for every four that depart (extended every three months, next expiration on October 15)

• Mandatory reduction plans: Agencies must submit formal strategies to shrink their workforce.

There’s also a voluntary exit route—the Deferred Resignation Program (DRP)—modeled loosely on Clinton’s buyouts. Employees who resign or retire early continue to receive pay and benefits through September or December.

So far, the numbers are striking. Since January, 85,300 federal jobs have fallen off the payroll, and about 154,000 workers have taken buyouts under the DRP. Most of those reductions come not from layoffs, but from the hiring freeze preventing agencies from replacing those who leave.

At this pace, federal downsizing is running nearly twice as fast as Clinton’s early cuts. But it’s still early days. During Clinton’s first term, the deepest reductions came later on. By this same point, he’d cut about 54,000 positions. The current total—85,300—is ahead of schedule, but it remains to be seen whether the momentum lasts.

If this government is to match Clinton’s first-term cuts, federal employment needs to drop below 2.08 million workers by 2028. Reducing the federal workforce back to the Clinton-era lows of 1999 would require shrinking it down to 1.88 million.

Cutting Jobs, Shifting Jobs, and the Shape of What’s Next

Healthcare’s steady expansion and manufacturing’s contraction capture the reallocation story at the heart of today’s labor market. Where jobs grow—and where they disappear—helps explain why some young workers thrive while others stall. Meanwhile, the federal workforce reductions mark one of the most significant government pullbacks in decades, echoing the reform-minded cuts of the 1990s.

In the months ahead, it’s worth watching not just how many jobs the economy adds or loses, but which ones. Where workers land and what they produce will tell us if this reshuffle leaves the economy stronger.

We had planned to discuss the September jobs report, but as many of you know, the ongoing government shutdown prevented BLS from releasing its report as planned. Instead, this gave us an opportunity to dive deeper into something more fundamental: how economists measure employment in the United States, why the current system is under scrutiny, and what we can do to restore trust in these critical statistics.

In Brief:

• The BLS’s delayed September report offers a window into how U.S. employment data are produced—and why confidence in them has eroded.

• Recent benchmark revisions show the U.S. added about 900,000 fewer jobs than first reported, exposing systematic weaknesses in data collection.

• Declining survey participation and outdated statistical models—not political bias—drive most of the errors in job estimates.

• The BLS remains far more accurate than private-sector alternatives like ADP, despite falling response rates.

• Simple procedural improvements—quarterly benchmarking, refined imputations, and greater transparency—could substantially improve accuracy.

• Restoring public trust in labor data depends on practical reforms, not structural overhauls.

Signs of Weakness

From a high level, we’ve been seeing increasing signs of a weakening labor market. Unemployment crept up to 4.3% in August, after fluctuating between 4 and 4.2% for more than a year. According to ADP Research, a private firm that provides economic data in parallel to the government statistics, September payrolls declined by 32,000 (though ADP numbers tend to be more volatile). The consensus forecast from economists was around 50,000 new jobs for September—so a negative number or one closer to zero would certainly be disappointing.

Beyond the headline unemployment rate, there are deeper signs of slack. The number of people not in the labor force but who indicate they want a job has surged from 5.48 million at the start of the year to 6.35 million in August—almost 900,000 people in just seven months. This signals hidden labor market weakness among those who haven’t actively looked for work in the past four weeks.

Another troubling indicator: the number of job openings fell below the number of unemployed in July, and the gap has widened since.

Then came the bombshell. On September 9, BLS released preliminary revisions showing the economy added 911,000 fewer jobs than initially estimated for the year through March. If these revisions hold, average monthly job creation would be about half of what we initially thought—71,000 rather than 147,000.

The most likely culprits? Weaker than inferred job creation at new businesses and sampling errors caused by declining survey response rates. One economist noted that roughly two-thirds of the downward benchmark revision appears to stem from faulty business creation counts.

A Crisis of Confidence

This is an unusual moment for labor economists. The BLS jobs report used to be the kind of thing you’d never discuss at a party or family brunch. Now your uncle corners you: “Don’t you work on this stuff?” President Truman famously complained, “Give me a one-handed economist! All my economists say, “On the one hand... on the other hand…’” These days, we find ourselves cast as the “but on the slightly positive side” economists in the dismal science.

So here’s our silver lining. Unprecedented public interest in how BLS collects, compiles, and reports employment data creates an opening. Yes, there’s confusion and skepticism. But confusion is teachable. This is our chance to explain how the machinery actually works and to identify straightforward fixes that could restore confidence.

Start with what happened this summer. On August 1, the BLS released the July jobs report showing 73,000 new nonfarm payrolls. Within hours, President Trump fired the BLS commissioner. The trigger wasn’t July’s modest number; it was the substantial downward revisions to prior months buried in the same release. May collapsed from 144,000 down to 19,000; June fell from 147,000 to just 14,000.

Then came the twist. The September 5 release, after the Commissioner’s firing, revised June down again, from +14,000 to –13,000. Not weak job growth, actual losses. The institutional machinery kept running anyway. Career staff applied the standard revision process, indifferent to political pressure or leadership changes.

The broader timeline is equally revealing. President Trump voiced concerns that the BLS numbers were artificially boosting VP Harris before the election. But consider what the benchmark revision actually showed. On August 21, 2024—before the election—the BLS announced a preliminary downward revision of 818,000 fewer jobs for the Biden-era period from April 2023 through March 2024. Then, on February 7, 2025, after President Trump took office, the final revision reduced that number to 589,000 fewer jobs.

In other words, the preliminary data released during the Biden administration made that period look worse, and the final revision released under Trump made it look less bad. If anything, the timeline cuts against claims of partisan manipulation. The real problem isn’t bias; it’s something deeper about how we measure employment in the first place and why our measurement system keeps missing by such large margins.

How the Sausage Gets Made

To understand why the revisions are so large, you need to understand where this data comes from in the first place. The BLS Employment Situation report aggregates two separate surveys:

• The CES (establishment survey) samples worksites—businesses answer “how many people are on your payroll?”

• The CPS (household survey) samples households—individuals answer questions about their work status.

Survey responses get weighted to population-level estimates, then augmented with something called birth-death model imputations. This is where things get tricky. The birth-death model is a statistical forecast that estimates jobs created by brand-new businesses not yet captured in the survey sample, plus jobs lost from businesses that closed before the BLS knew they existed. There’s always a lag between when a company opens or shuts down and when it appears in the BLS sampling frame.

The model relies on historical patterns of business formations and closures by industry. That works fine in stable times. But it systematically misfires at economic turning points. When the economy cools, the model overestimates job growth. It assumes startups that aren’t actually happening. When the economy rebounds, it underestimates growth. It misses surges in new firm formation. This is a major source of those massive benchmark revisions.

The initial monthly number gets revised twice as late survey responses trickle in. Then, once a year, comes the benchmark calibration to QCEW (Quarterly Census of Employment and Wages) administrative records. QCEW is essentially the near-universe of unemployment insurance tax records, covering roughly 95% of in-scope employment. This is when the BLS compares its survey-based estimates to actual administrative data and discovers it was off by hundreds of thousands of jobs. That’s the benchmark revision that generates headlines and fuels suspicion.

Why BLS Jobs Numbers Keep Changing: A Visual Guide

The Accuracy Question

The BLS surveys 121,000 businesses each month and asks them how many people they employed in the week including the 12th day. Employment attrition between these weeks means total accuracy is impossible.

Declining survey participation is one real concern. Before the COVID-19 pandemic, about 60% of businesses responded to the CES survey. In recent years, this has dropped to closer to 40%. But even with this decline, the estimates remain accurate. The June revision that initially upset the President was a correction of just a 0.16% error in employment numbers, not a huge miss in the grand scheme of things.

Some critics have suggested that private organizations might be better placed to estimate these numbers. We already have monthly jobs data from a private payroll company, ADP. But while BLS typically has an error rate between 0.1% and 1%, ADP has a median error rate of about 9%—that’s typically about 30 times as inaccurate as BLS data when measured against actual QCEW counts. Private payroll data doesn’t seem like a viable alternative if what we’re seeking is better accuracy and reliability.

Solutions: Low-Hanging Fruit

Here’s the good news. None of these solutions requires reinventing how we measure employment. They’re improvements to existing infrastructure. Some are purely procedural, while others would need additional resources, but none would require starting over from scratch.

Here’s what we can do:

• More frequent benchmarking – move from annual to quarterly QCEW reconciliation

• Better imputation methods – use small-domain estimation instead of similar industry coarse size classes

• Birth-death model transparency – publish a companion series without the forecast

Start with more frequent benchmarking. Right now, the jobs number benchmarks annually with QCEW, but QCEW data is available quarterly. Move to quarterly benchmarking and you immediately shrink the magnitude of revisions. Corrections would become smaller and more frequent rather than massive annual shocks. Smaller, predictable adjustments are harder to spin as partisan manipulation. Yes, QCEW has a couple of months of reporting lag, but those can be shortened with additional staff and faster data processing.

Fix the imputation method. When the BLS publishes the first monthly report, it has to fill in gaps for businesses that haven’t responded yet. Currently, CES imputes these nonrespondents by borrowing growth rates from “similar” establishments, defined by industry and size class. When a small restaurant doesn’t respond to the survey, the BLS imputes its employment data using responses from similar small restaurants that did.

The problem is that “similar” isn’t similar enough. Size classes are coarse: “10–49 employees” lumps together very different businesses. Worse, within each class, the donor pool skews toward larger, more responsive firms with sophisticated HR systems. This means the BLS is effectively applying growth rates from a 40-employee chain restaurant with a dedicated payroll department to estimate employment changes at a 12-employee family diner. These aren’t comparable establishments, and that matters for the quality of the imputation.

The fix is small-domain estimation. Give each industry × size × region cell its own model-based estimate that borrows information across genuinely similar cells, rather than letting big donors dominate the signal from small businesses.

Then there’s the birth-death model, where the biggest distortions live. This is a pure statistical forecast, not real-time data. It estimates jobs created by brand-new businesses not yet in the survey sample, plus jobs lost from businesses that closed before the BLS knew they existed. The model runs on historical patterns of business formation and closure.

Here’s the core problem: big firms are visible immediately, but small firms and startups stay invisible for months. When the economy hits a regime change—pandemic, rate shocks, credit crunch—the model keeps projecting “normal” business formation based on 2018 or 2019 patterns even as reality shifts underneath it. At turning points, this creates systematic bias in predictable directions: (1) during slowdowns, the model overstates job growth by assuming startups that aren’t actually happening; 2) during recoveries, the model understates job growth by missing surges in micro-firm formation.

The transparency fix is straightforward: publish a “no birth-death” companion series every month. Just show what actual survey respondents reported, without the forecast layered on top. Let users see the raw signal before statistical adjustments.

None of these fixes requires Congress to pass legislation or for agencies to undergo major reorganization. They are methodological improvements that the BLS career staff could implement with modest budget increases and leadership buy-in.

The Path Forward

We’ve covered a lot of ground—from how employment data are produced to the political pressures surrounding them, from federal workforce trends to sectoral shifts reshaping the economy. Here’s what often gets lost: the BLS is actually very good at its job. When compared with administrative records, BLS data remain far more accurate than private alternatives. The reforms outlined above require effort to implement, but they build on existing infrastructure rather than starting from scratch. Quarterly benchmarking, small-domain estimation, and greater transparency in the birth-death model are refinements to an already strong system. This period of public scrutiny—uncomfortable as it may be—offers an opportunity to make a good measurement system even better.

I’ve been thinking a lot about something that doesn’t get enough attention in labor economics debates: what happens when workers can walk away more easily?

In my recent Hill column, I explored the connection between portable benefits, job lock, and worker power—noting how portable benefits are gaining momentum—with Senator Cassidy’s bill in the Senate, similar legislation in the House, and Senator Warner’s earlier efforts. But there’s a deeper story here about power and leverage that I want to unpack.

The Monopsony Problem

Let me start with a concept that sounds academic but is actually quite simple: monopsony power.

You’ve probably heard of a monopoly—when one seller dominates a market. Well, monopsony is the flip side: when one (or just a few) buyers dominate a market. In labor markets, that “buyer” is your employer. And when employers have monopsony power, they can pay you less than what your work is actually worth—because where else are you going to go?

Here’s the thing: you don’t need to live in a company town with one employer to experience monopsony power. It happens if the cost of leaving your job is too high. Maybe you need the health insurance. Maybe your skills are specialized to your current workplace. Maybe good alternative opportunities are too difficult to find. Whatever the reason, if switching jobs feels risky or difficult, an employer has more leverage over a worker—and that might show up as lower wages and/or fewer on-the-job amenities or worse conditions.

The Economic Policy Institute has documented this extensively. Their research shows that when workers face a 10% pay cut, only 20-30% actually quit. Meaning, if employers cut your pay by 10%, seven or eight out of ten workers stay anyway.

The Hidden Handcuffs

One of the biggest culprits? Employer-sponsored health insurance.

Survey data shows that 80% of workers cite employer-provided coverage as an ‘impactful’ reason when deciding to stay in their current job. In other surveys, 56% of workers said their health plan is why they remain with their employer. A 2021 West Health-Gallup survey found that among workers in households earning less than $48,000 a year, 28% answered “yes” to the question: “Are you currently in a job that you want to leave but don’t because you are afraid of losing your health insurance benefits?” Previous research has documented that for chronically ill workers reliant on employer-provided coverage, job mobility drops 40 percent.

Our entire tax code and regulatory structure favors employer-sponsored plans, making individual alternatives prohibitively expensive for most people. The Affordable Care Act created individual marketplaces, but it didn’t eliminate the structural advantages that make employer plans more affordable.

This creates what economists sometimes call “job lock”—you’re not staying because the job is great, you’re staying because leaving means losing benefits your family depends on. Other economists have studied this as “entrepreneurship lock”—a situation in which workers would pursue entrepreneurship but for their reliance on employer-provided benefits, particularly health insurance.

There are of course many other reasons you don’t want to leave your job—for example, you’re afraid that you won’t find other sources of income easily or the job market is not great right now.

Enter the Multi-Earner Economy

Now here’s where things get interesting.

More Americans are earning income from multiple sources—driving for Uber part-time, running an online business, freelancing, teaching on the side, doing high-level consulting. This isn’t necessarily about people being desperate for extra money (though for some it is). It’s about the fundamental structure of how people earn and think about household income. In the 20th century, it was one job or one career for your whole life. In today’s economy, jobs and careers are more fluid, and up to 60 million Americans are earning income from many different sources.

Of course, the experience varies widely—for some workers, juggling multiple jobs is exhausting necessity, not empowering choice. But for millions of others earning supplemental income by choice, the dynamics are different.

I’m starting to believe that this shift into the multi-earner economy is reducing monopsony power in ways we haven’t fully appreciated yet (not for all workers, but for many).

When you have multiple income streams, you’re simply less dependent on any single employer. If your full-time job cuts your pay or becomes untenable, you already have other revenue coming in—and you already have relationships with other clients or platforms. The cost of walking away drops dramatically.

That’s not theoretical leverage. That sounds more like real bargaining power.

Two Pieces of the Puzzle

This is where portable benefits fit into the picture—and why I think the current legislative momentum matters more than people realize.

Think about it as two complementary forces that could shift power toward workers:

1. Multi-earning directly reduces dependence

When you’re earning from multiple sources, no single employer controls your economic fate. You have options. You have fallback income. The exit costs are lower. This isn’t about everyone becoming full-time freelancers—most independent workers maintain a primary W-2 job and supplement their income. But that supplemental income changes the equation.

2. Portable benefits remove the handcuffs

If benefits weren’t tied to a single employer—if you could take your health insurance, retirement contributions, and other benefits with you—then the biggest barrier to job mobility disappears. You gain the freedom to move, negotiate, or pursue opportunities without risking your family’s healthcare or financial security.

Together, these two forces could fundamentally change the employer-employee dynamic.

Imagining a Different Labor Market

I want to be clear: the independent work system isn’t perfect. There are some independent contractors who are genuinely vulnerable and deserve thoughtful policy protections. This isn’t an argument that portable benefits solve all problems or that independent work is without challenges—I’m trying to understand the relationship of work for the majority of independent workers who choose (and are not forced) into independent work and who use it as supplemental income.

Right now, we have a system where benefits are controlled and granted by your employer. What if it wasn’t? What if workers had genuine freedom to leave, to negotiate, to explore better opportunities without the fear of losing those benefits?

That’s not just about making independent work more secure. It’s about giving all workers more leverage—because when people can walk away more easily, employers have to compete harder to keep them.

An Open Question

So here’s what I’ve been mulling over: if we see a movement toward a system of portable and flexible benefits (along with a continued growth in multi-earner economy), could we be looking at the first major shift in worker bargaining power in decades?

When workers aren’t locked in by benefits, and when they have other income opportunities readily available, does the balance finally tip?

I don’t have all the answers. But I think we’re at an inflection point where the right policy changes could unlock something powerful: a labor market where workers aren’t stuck, where mobility is real, and where having options isn’t a privilege—it’s the norm.

Whether this holds empirically remains an open question. Do workers in sectors with more multi-earning opportunities respond more strongly to wage changes? Do portable benefits laws drive reduced monopsony power? The answers would tell us if we’re seeing a structural shift or a limited phenomenon. But the theoretical case is strong: when workers have genuine exit options, bargaining power shifts. The question for policymakers is whether our institutions are suppressing options that workers and markets would otherwise create.

Subscribe now

From Fringe to Front Page

Five years ago, when I first began sharing my research with lawmakers about a voluntary portable benefits approach for independent and gig workers, it felt too unconventional — a niche idea that didn’t quite fit into the standard policy debates of the time.

Now, that once-niche idea has gone mainstream. Sen. Bill Cassidy recently introduced a legislative package to support portable benefits — building on a Senate HELP white paper that cites our research. Earlier this year, Rep. Kevin Kiley introduced a House version, which passed in a committee vote yesterday. Meanwhile, states are paving the way: Wisconsin, Tennessee, and Alabama have all passed portable benefits reforms in 2025, following Utah’s lead in 2023. Governors in Pennsylvania, Georgia, and Maryland have also approved pilot programs to test the model.

And earlier this week, the Wall Street Journal editorial board published an endorsement of this approach. The revolution is here — and it’s voluntary, flexible, and worker first.

I’ve been fortunate to contribute to this conversation from the beginning — by publishing research and policy guides that examine outdated assumptions about work and benefits. I’ve shared these findings with Sen. Cassidy’s and Rep. Kiley’s team, as well as with every congressional or state lawmaker who showed interest — and have testified more than a dozen times before Congress and in state legislative hearings.

It’s surreal (and deeply encouraging) to see how far we’ve come. Let’s talk about how this once-niche idea just went mainstream.

States Led the Way

When I first brought my research to congressional staff, interest was limited. So I shifted focus to the states — and that’s where the breakthrough happened. Utah became the first to adopt the voluntary portable benefits model in 2023, and since then, nine other states have introduced similar bills. Three of them have passed legislation, and governors in Pennsylvania, Georgia, and Maryland have approved pilot programs — facilitated by the benefits company Stride. These early state-led efforts proved that the model works — and helped pave the way for federal action.

Other portable benefits models exist too — notably in California (Prop 22), Washington, and Massachusetts — where companies like Uber, Lyft, DoorDash, and Instacart are legally required to provide a menu of benefits to gig workers.

The voluntary model, informed by our research, takes a different path. It doesn’t mandate benefits — it removes the legal barriers that prevent companies or clients from offering them in the first place. It also applies broadly to all independent workers, not just those in a single industry.

Five Reason Why the Ideas Are Taking Off

1. Not Your Grandfather’s Workplace
   The U.S. labor market is no longer built around lifelong, single-employer jobs — and hasn’t been for a long time. Today, people move in and out of jobs, take on side gigs while caregiving or studying, freelance professionally, launch entrepreneurial ventures, or earn income from multiple sources. Roughly 27 million Americans now work independently — whether in a primary job or to supplement other income — and that number is only expected to grow. As I emphasized in my May 2025 testimony before Congress, portable benefits offer a modern solution for a modern workforce — one that reflects how people actually work today.
   

2. It’s What Independent Workers Want
   Most independent workers aren’t looking to be reclassified — they simply want access to benefits without giving up their autonomy. According to the BLS, over 80% say they prefer their current work arrangement, and fewer than 9% say they’d rather be a traditional employee. Portable benefits accounts offer a way to access health insurance support, retirement savings, and paid time off — all without changing how or why they work. In fact, 81% of self-employed workers say they want portable benefits solutions.
   

3. The Research on Reclassification: Fewer Jobs, Not Better Ones

   Attempts to force legitimate independent workers into traditional employment through laws like California’s AB 5 don’t convert contractors into W-2 employees — they just reduce work opportunities. My research team’s study found that after AB5 took effect, self-employment in affected occupations dropped by over 10%, without an accompanying increase in W-2 workers for those affected occupations. Similar effects have been found in states that implemented ABC test. Our preliminary results show ABC tests laws reduce both self-employment and standard W-2 jobs across the board.

4. Bipartisan Support
   Portable benefits reform has attracted rare bipartisan support. Center-left groups such as Brookings and Progressive Policy Institute have long backed the idea, while center-right organizations such as the Independent Women’s Forum and Americans for Prosperity emphasize its value for preserving worker flexibility. Bipartisan federal legislation encouraging pilot portable benefits programs from Sens. Warner and Young and Rep. DelBene reflects this growing momentum. At the state level, Democratic lawmakers in Maryland and Pennsylvania have allowed for pilot programs in their states, and across the country portable benefits legislation is often co-sponsored by both parties — a sign of how practical and widely resonant this idea has become.
   

5. Marginal Steps Toward a Better World

   These reforms are practical and incremental — designed to give companies and clients the legal clarity to offer benefits voluntarily. In places like Utah and Pennsylvania, workers are already using funds from portable benefits accounts to pay for health insurance premiums, save for retirement, or take paid time off. The contributions may be modest at first, but for many, this marks the first time they’ve had access to any kind of benefits safety net. These early steps are building a foundation for what could become a new pillar of economic security for the flexible workforce.

What’s Next?

For the next few months, all eyes are on Washington D.C. These are the two key federal bills that could shape the future of portable benefits nationwide:

• S.2210 - Unlocking Benefits for Independent Workers Act (Cassidy’s Senate bill): Introduced in July 2025, this bill creates a legal safe harbor for companies and clients to contribute to benefit accounts for independent workers without legal risk. Last week, Senate HELP held a hearing on this bill.

• H.R. 1320 - Modern Worker Security Act (Kiley’s House bill): Introduced in February 2025 and passed in a committee vote yesterday. Like Sen. Cassidy’s bill, it creates a legal safe harbor and is aimed at expanding voluntary portable benefits across sectors.

If passed, these bills would open the door for national adoption — giving workers more security and companies the legal clarity to offer those benefits.

Research and Resources

Selected Testimonies:

• May 2025 Congressional testimony

• April 2024 Congressional testimony

• Nevada testimony (May 2025)

• Massachusetts testimony (March 2024)

• Utah testimony (February 2023)

Selected Op-eds:

• “Could Congress Finally Delivery Benefits for All Workers?” (National Review, July 2025)

• “How Portable Benefits Can Drive US Innovation and Job Growth” (The Hill, February 2025)

• “Here’s One Way America’s Many Pro-worker Candidates Can Deliver” (The Hill, October 2024)

• “Utah is the First State to Truly Welcome the Gig Economy” (Salt Lake Tribune, May 2023)

Policy Guides:

• Federal Policy Guide on Portable Benefits

• State Policy Guide on Portable Benefits

• Response to Senator Cassidy’s Request for Information on Portable Benefits

Research Studies:

• Flexible Benefits for a Flexible Workforce: Unleashing Portable Benefits for Independent Workers and the Gig Economy (2023)

• Assessing the Impact of Worker Reclassification: Employment Outcomes Post–California AB5 (2024)

• Understanding Non-Traditional Work and the Policy Landscape (2023)

• Barriers to Portable Benefits Solutions for Independent Workers and the Gig Economy (2020)

Welcome to all our new readers! This post is part of our new series on AI & Labor—a space to explore how intelligent machines are reshaping the very nature of work.

Over the coming weeks, we’ll continue to share insights on how AI is transforming labor markets—led by our Labor Policy predoctoral researcher, Revana Sharfuddin.

Author’s note: This is Part II of a two-part essay on preparing for the AI-augmented economy. In Part I, I focused on the coming labor transitions and how to think beyond automation toward augmentation. In this second installment, I dig into what readiness really means—not just as workers, but as people. Not just for the economy, but for society. Readiness is not simply a matter of skills; it’s a matter of meaning. As Tyler Cowen recently argued, we’re entering a future where flourishing will require a radical shift in how we view purpose, collaboration, and value.

Preparing for that future means confronting a series of cultural, institutional, and policy challenges. It will require mindset shifts at the individual level, renewed collaboration between companies and civil society, and smarter policy that removes roadblocks rather than creating new ones.

Subscribe now

We Need Each Other: Working Together, Resting More, Rebuilding Society

Since the rise of the internet and the IT revolution, we have been living in the era of the technical elite, where technical skills have reigned supreme. The message has been clear: master coding, data analysis, and computational thinking, and you will thrive. Soft skills—like communication, collaboration, and emotional intelligence—have often been sidelined as secondary. Even in team-based fields like software development, direct interpersonal interaction has been reduced to virtual workspaces where developers collaborate through shared screens and chat channels, creating digital systems with minimal face-to-face interaction.

But perhaps it is time to step back and reconsider what we have left behind.

As coding and automation-related jobs surged during the computerization boom, professional skills that emphasize relationship-building, collaboration, and emotional intelligence were systematically devalued. The COVID-19 pandemic accelerated this shift, particularly for younger workers whose first jobs were fully remote. Workplace networking, mentorship, and informal social learning—once essential for professional growth—were suddenly absent. Dating and friendships, which once flourished through organic, in-person interactions, became mediated almost entirely through screens. A generation of workers entered adulthood without developing the soft human capital that underpins trust, adaptability, and social cohesion.

Survey research shows that valuing interpersonal skills has the largest generational gap, with older STEM workers valuing them far more than younger ones.

For many, remote work became a convenient escape from the often exhausting unpredictability of human interaction. And understandably so—social engagement takes effort. It requires learning each other’s rhythms, making space for different personalities, navigating rapidly shifting social norms, and granting others the benefit of the doubt. In an age where a single misplaced word can trigger outrage, where conversations feel like navigating a minefield of political correctness, and where miscommunication is met with instant dismissal rather than understanding, it is no surprise that many have opted out.

The result? A loneliness crisis, increasing political polarization, and a workforce that has learned it is simply more efficient to work from home, alone, free from the complications of human interaction.

But in optimizing for efficiency, have we sacrificed something fundamental? Have we, in the process, lost a bit of what has historically been our greatest comparative advantage—being human?

To be clear, remote work offers undeniable benefits—it allows for greater flexibility, more time with family, and better support for caregiving, all of which are crucial in an era of declining fertility and weakening social cohesion. But as we weigh these advantages, we must also consider what we risk losing. Technology has consistently freed up time from routine labor, yet whether that extra time enhances our lives depends on how we use it.

Consider this striking statistic: In 2015, an average U.S. worker could have maintained the income level of an average worker in 1915 by working just 17 weeks per year. AI has the potential to accelerate this trend—reducing time spent on repetitive tasks and increasing leisure to allow for a greater focus on family life, child-rearing, and deeper interpersonal relationships. However, the full benefits of this shift can only be realized if we lean into our comparative advantage as humans.

History shows that the jobs that endured were those that leveraged human strengths—judgment, creativity, and social intelligence. If AI is poised to take over routine technical tasks, the real opportunity lies not in competing with machines but in mastering the distinctly human skills that allow us to wield AI to its fullest potential.

Taxing the Future? Unpacking the Risks of AI Penalties

For decades, conventional economists have made a critical mistake: underestimating the burden placed on workers displaced by technological shifts and globalization. Foundational economic models assumed that technological progress was neutral, benefiting all workers. In practice, technological change can disproportionately help some while displacing others in the transition period. While advancements boost productivity and raise overall living standards, they do not always distribute their gains equally.

Too often, the response has been cold rationalism: Your job is obsolete? Adapt or risk being left behind. But economic agents are not just anonymous data points in a labor market model. They are people, with families, financial obligations, aspirations, and identities deeply tied to their work. The assumption that displaced workers can “easily” retrain and transition “seamlessly” into emerging fields ignores the real costs of career shifts. Middle-aged and low-wage workers face steep learning curves, financial constraints, and cognitive exhaustion from balancing work, childcare, and financial stress.

The lower a worker is in the income distribution, the more cognitive capacity is consumed by immediate survival concerns—worrying about bills, rent, and daily necessities—leaving little capacity for retraining or investing in new skills. When these realities are ignored, trust in institutions erodes.

Political polarization is, in part, a cautionary tale about the consequences of dismissing economic displacement. When workers are told that the economy is better off on average, while their own prospects remain bleak, it breeds disillusionment and a sense that the system no longer represents them.

In recent decades, many market-oriented economists have largely remained silent about proactive solutions for supporting workers amid technological disruptions. This absence has allowed other groups to step forward with proposals—mostly centered on taxation or redistribution. Consider, for instance, a recent policy memo by Acemoglu and his colleagues suggesting taxes on AI technology.

While their recommendation emerges from rigorous, influential research, we must examine carefully the assumptions underlying their proposals. In traditional growth frameworks, such as the Solow or Ramsey–Cass–Koopmans models, technological advancement is modeled explicitly as labor-augmenting productivity—captured by a parameter. In these scenarios, technological progress complements workers uniformly, enhancing productivity without replacing jobs.

Of course, this is a simplification—real-world effects are uneven, and some degree of substitution between labor and technology does occur, depending on the task. However, Acemoglu et al.'s task-based model differs fundamentally. Their framework explicitly addresses automation—where technology replaces labor—without an exogenous aggregate productivity shifter. In other words, the economy evolves as tasks shift between capital and labor, not by scaling up labor productivity.

Consequently, their optimal tax calculations are primarily aimed at addressing cases of marginal ("so-so") automation, where AI substitutes for human work without meaningful productivity improvement.

If we broadly apply an optimal tax rate designed for labor-replacing AI to all AI technologies—including those that enhance human capabilities—we risk penalizing the very productivity gains essential for wage growth and job creation.

Time and again, history has taught us that the best way to support displaced workers is through retraining programs. But who will lead this effort? The answer is almost always the government. Yet do we really trust policymakers to understand, predict, and effectively respond to one of the most complex technological transformations in modern history? Consider the now-famous congressional hearing where TikTok CEO Shou Chew testified before the House Committee on Energy and Commerce—a moment that starkly illustrated how disconnected many policymakers are from even the most basic workings of technology. If these are the same institutions we expect to guide workers through the AI revolution, should we not be thinking more critically about alternative solutions?

Two Market-Based Reforms to Support a Changing Workforce

While comprehensive solutions will require broader effort, here are two practical reforms that could help us begin moving in the right direction:

First, fix how the tax code treats investments in human capital. Currently, businesses can deduct expenses for worker training only if the training improves skills for their current job—not if it qualifies them for new types of work. This creates a perverse incentive: companies have a tax advantage when training workers to stay in the same roles but a disincentive when investing in training that allows employees to transition into higher-value, future-proof jobs. If we treated worker training the same way we treat equipment purchases—allowing businesses to fully deduct all training expenses—firms would have a stronger incentive to invest in reskilling their workforce. Unlike industry-specific tax credits, which introduce complexity and distort incentives, full expensing of all forms of investment—both human and physical capital—ensures neutrality and efficiency.

Second, embrace a universal form of Individual Training Accounts (ITAs)—a worker-centered approach designed to empower job seekers by letting them choose training that aligns with their interests and market demands. However, an even more impactful approach to meet the same goals would be the adoption of Universal Savings Accounts (USAs). Compared to ITAs, USAs are simpler, more flexible, and easier for individuals to manage on their own. Many low-income families struggle with navigating numerous tax-advantaged savings accounts, each with distinct eligibility requirements and penalties. This cognitive burden discourages savings and undermines financial security, particularly for those most vulnerable. USAs simplify the process by offering a single, accessible savings account where contributions grow tax-free and withdrawals can be made at any time, without taxes or penalties. Experience from Canada and the United Kingdom shows this simplified approach increases savings participation among low-and moderate-income families, giving workers the agency, dignity, and self-determination they need to invest in their own futures.

Building Coalitions: Why We Need Everyone at the Table

A shift of this magnitude demands collaborative effort—businesses, workers, policymakers, and civil society must all be at the table because the stakes are too high for complacency. Recall my earlier discussion in Part I on the business potential of automation versus augmentation (as illustrated by Erik Brynjolfsson). History shows that automation alone has limits, while augmentation—when coupled with structured retraining—has the power to unlock immense economic dynamism. Companies must take the lead in human capital development because augmentation builds stronger workforces, reduces displacement anxiety, and creates the foundation for shared economic growth.

Private industry needs to be proactive—not just funding retraining but shaping it. Companies should partner with local community and technical colleges to design micro-credential programs that match regional labor market needs. Just as important, they should help workers navigate these options through clear communication and guidance. Training tied to local demand, clearly communicated, is how anxiety about obsolescence can become ambition for what's next.

The momentum is already building. Since my last Substack post, I was invited to share my research with members of Capitol Hill staff. I am happy to report that there is growing bipartisan interest in supporting workers and communities amid technological change. Additionally, the newly launched AI-Enabled Policymaking Project (AIPP) has convened AI labs, startups, civil society organizations, academic experts, policymakers, government representatives from defense and intelligence, and independent technologists to share specialized knowledge and identify practical solutions. Small and medium enterprises (SMEs) are also proactively engaging researchers to integrate AI into their workflows. Today, I'll join the Fairfax County Small Business Forum to further these important conversations.

Public-private partnerships are essential, but equally important is a thoughtful reimagining of the labor unions’ role. Unions should no longer be seen merely as relics of industrial-era conflict or obstacles to innovation. Instead, they can serve as strategic partners—helping businesses and workers navigate change, supporting effective retraining, representing worker voice, and fostering shared understanding of future market directions. Union involvement in the SB 1047 debate signals a willingness to engage constructively in shaping AI’s role in the economy. However, for unions to realize their full potential, a reorientation is necessary—reforming institutional laws and structures, shifting away from traditional adversarial approaches toward collaboration, and embracing the possibilities of augmentation rather than opposition.

In the United States, existing labor laws grant certified unions exclusive bargaining rights, which can inadvertently encourage more adversarial stances. In our new paper, we explore alternative institutional structures that could enable unions to evolve into collaborative, forward-thinking partners, fostering constructive dialogue and cooperation between workers and firms. This shift can transform AI from a perceived threat into a catalyst for economic growth and shared prosperity.

Conclusion

The lesson from history is clear: when societies harness technological change rather than fear it, they emerge stronger. But achieving this strength requires purposeful action. It demands individuals willing to revalue and cultivate uniquely human strengths—creativity, empathy, and collaboration. It necessitates proactive businesses that invest thoughtfully in worker training and forge genuine partnerships with educational institutions. It calls for policymakers to facilitate, rather than obstruct, innovation by aligning incentives and reducing cognitive burdens on workers seeking reskilling. And it requires re-imagined labor unions and civil society to work together to enhance worker agency and dignity.

The future of work need not be a contest between man and machine. It can be a stage for rediscovering what only humans can do—if we choose to build that future together.

A Personal Mission Reaches Capitol Hill

For the past several years, I’ve educated policymakers to highlight a major flaw in our labor system: benefits are tightly tied to W-2 employment, leaving millions of self-employed and independent workers without access to these benefits. Through my research and briefings, I’ve outlined a set of reforms to modernize this outdated model—one of the most important being the need to decouple benefits from employment classification. This would allow companies to offer benefits voluntarily, without the legal risk of misclassification. 

At the time, these proposals were viewed as too niche to gain widespread attention. But now they’ve moved to center stage. 

In a remarkable shift, Senator Bill Cassidy, Chair of the Senate HELP Committee just released a white paper that directly addresses the need for portable benefits. His report reflects many of the core recommendations I’ve advanced in my work. Earlier this year, Rep. Kevin Kiley introduced a bill in the House that reinforces these same principles at the federal level.

At the state level, momentum is growing too. Utah, Tennessee, and Alabama have all enacted legislation to clear the path for portable benefits. In Pennsylvania and Georgia, governors have approved pilot programs that allow companies to offer benefits to independent workers without triggering reclassification rules.

Put simply, a once-obscure policy idea has entered the national spotlight—and it’s just getting started.

The First Briefing: Planting the Seed

I held my first portable benefits lecture on the Hill on February 27, 2020. At the time, I had just finished the first draft of my study on “Barriers to Portable Benefits Solutions for Gig Economy Workers,” and I was eager to educate policymakers about how our current laws restrict the voluntary flow of benefits to independent and self-employed workers — and also how our tax laws promote W-2 employees while discouraging self-employment and entrepreneurship. 

My pitch was simple: our laws create legal barriers that block independent workers from accessing voluntary benefits. If a company or individual provides benefits to a freelancer, that act alone could trigger reclassification as a W-2 employee. The result? The hiring party risks being liable for back taxes, benefits, and penalties—so most simply won’t take the chance. It’s a system that unintentionally penalizes the flexible workforce.  Indeed, 80 percent of independent workers want to remain in those arrangements (as opposed to being W-2 employees), and instead the vast majority of them have indicated they desire portable benefits. 

The first step toward fixing this is just as simple: legalize access to benefits. Federal law should make clear that the presence of benefits cannot be used to determine a worker’s classification status. 

That’s exactly what Rep. Kevin Kiley’s Modern Worker Security Act now proposes. But when I first started sharing my research and these ideas with policymakers, they landed with silence. No follow-up questions. No interest. Of course, the pandemic soon took over everyone’s attention—but the conversation was quiet even before that.

Still, my team kept up the educational efforts—meeting with congressional offices, writing op-eds, and making the case for reform wherever we could.

By 2022, we decided to broaden our efforts to the state level—and that’s when the breakthrough came.

Shifting to the States: The Utah Breakthrough

While there was little movement at the federal level, a different kind of momentum was building in the states.  In November 2020, California voters passed Proposition 22, a mandatory portable benefits framework limited to app-based transportation and delivery workers. It established that these workers would remain independent contractors, but required companies to provide a menu of benefits. In 2022, Washington state followed with a similar law, also mandatory and restricted to gig transportation platforms.

This was a different approach from the voluntary framework I had outlined in my original study. The California and Washington models were:

1. Mandatory,

2. Defined the specific ‘menu of benefits to be provided,’ and

3. Limited to app-based transportation and delivery companies.
   

Still, even as the mandatory bills advanced, we continued educating policymakers and the general public about the benefits of a voluntary, flexible approach—one that could apply across industries and support all types of independent workers.

In a key policy brief for lawmakers, Flexible Benefits for a Flexible Workforce, I emphasized that while federal action could unlock even broader reform, states could take the lead by clarifying that the presence of benefits cannot be used in determining employment classification. The brief also outlined three guiding principles to help shape early portable benefits efforts:

• Voluntary participation for all parties

• Multiple funding sources—including clients, companies, and the workers themselves

• Portability, so benefits move with the worker, not the job
  

Then came the Utah breakthrough.

In mid-2022, Caden Rosenbaum of the Libertas Institute reached out after reading my study on barriers to portable benefits. He wanted to bring these ideas to Utah. Caden shared the policy framework with Utah lawmakers, and shortly after, Senator John Johnson introduced the first voluntary portable benefits bill in the country. The bill stipulated that the presence of benefits could not be used to determine a worker’s classification status. I was invited to testify before the committee on the bill, which passed unanimously out of the Senate committee.

Just a few days later, it cleared the full Utah Legislature with little opposition.

On March 23, 2023, Governor Spencer Cox signed the Utah Portable Benefits bill into law. I wrote an op-ed in the Salt Lake Tribune explaining what the law means for the future of flexible work. 

And just like that, the portable benefits reform idea moved from theory to reality—with Utah leading the way.

Proof of Concept: Stride as Portable Benefits Provider

A crucial component of the portable benefits model is the presence of a third-party provider—one that can manage benefits accounts independently of any single employer, ensuring the benefits are truly portable and tied to the worker, not the job.

When Utah passed its portable benefits law, Stride, a California-based company, became the first to operationalize the model as envisioned: acting as the third-party facilitator that keeps benefits with the worker, regardless of whom they’re contracting with. 

Founded in 2013, Stride partners with major platform companies like Uber and DoorDash and helps self-employed individuals, gig workers, and part-time employees navigate access to benefits like health coverage and tax deductions.  

So within a year of the Utah law taking effect, Shipt (a Target-owned delivery platform) launched a portable benefits pilot for drivers in the state, with Stride facilitating the program. Six months later, Lyft announced its own pilot, contributing 7% of driver earnings into individual accounts to help workers pay for health coverage, retirement savings, or time off.

The model quickly began spreading to other states.

In Pennsylvania, with the backing of Governor Josh Shapiro, DoorDash rolled out a pilot portable benefits program, again facilitated by Stride. Positive results from that program helped pave the way for expansion. Next came Georgia, where Governor Brian Kemp approved a similar pilot, also powered by Stride.

As this momentum built at the state level, federal policymakers began paying closer attention—seeking to understand how portable benefits could be scaled nationally and what kind of legal clarity would be needed to support such innovation.

The Portable Benefits Future Arrives: Congress Take Notice

By 2023, portable benefits were no longer a niche policy idea—they were gaining traction on Capitol Hill–several key think-tanks and individuals joined our efforts and proved to be pivotal to growing the movement. 

That year, I did almost two dozen congressional lectures and briefings on portable benefits reforms. This time, the response was different: staffers and members were engaged, asking follow-up questions and requesting additional research and policy resources. The federal momentum had finally begun.

In April 2023, I was invited to provide expert testimony on independent work before both the Senate Aging Committee and the House Education & Workforce Committee. These hearings marked a turning point in how seriously Congress was beginning to consider portable benefits.

Exactly one year later, in April 2024, the House Education & Labor Committee invited me back—this time specifically to testify on portable benefits. They wanted to explore how federal policy could support the kinds of models being tested at the state level.

Here were the three big takeaways from my testimony:

1. Independent workers' lives would be significantly improved if they had access to benefits.
   

2. States are already experimenting with various portable benefits approaches, giving workers more options than the outdated choice between rigid employment or total flexibility without support.
   

3. Federal policy can play a key enabling role by creating a safe harbor—clarifying that providing benefits should not be a part of the worker’s classification determination. 

As a personal note: I brought my four-month-old baby with me to that April 2024 testimony—the same room where I testified one year earlier while pregnant with her –– life in full circle! 

By June 2024, Senator Bill Cassidy issued a formal Request for Information on portable benefits—a clear signal that Congress was actively exploring legislative options. My response laid out key recommendations and design considerations.

A few months later, I published—along with my co-author Jonathan Wolfson—a Federal Policy Guide and State Policy Guide on portable benefits, bringing together years of research and lessons from state experiments. Since then, I’ve continued informing lawmakers across the country with my research on independent workforce and portable benefits. 

Where We Stand Now: A Bipartisan Future for Portable Benefits

A large part of the reason why there is so much interest in portable benefits reform is because it’s a truly bipartisan idea. It allows properly classified independent workers to access benefits without increasing misclassification and without reducing traditional W-2 employment, as shown in our analysis of Utah’s portable benefits law.

Center-left groups have long supported this approach. The Progressive Policy Institute published a study arguing for removing the presence of benefits from the worker classification determination. The Brookings Institution published this report on portable benefits last week, and a year earlier, they published Economist Jonathan Gruber’s portable benefits proposal. A 2015 Brookings report on modernizing labor laws—alongside earlier work by the Aspen Institute—helped inspire my own research in this space.

At the same time, right-of-center organizations are also supporting these reforms. The Independent Women’s Forum published a policy focus on portable benefits, as did Americans for Prosperity, emphasizing the importance of flexibility and choice in how workers access economic security.

On the legislative side, Democrats Senator Mark Warner (VA) and Representative Suzan DelBene (WA), along with Republican Senator Todd Young (IN), introduced a bill to grant $200 million to states and localities to pilot portable benefits models. While the bill doesn’t directly address the core legal barriers, it supports the broader movement toward portable benefits. 

The takeaway is clear: portable benefits enjoy support from across the political spectrum—a rare thing in today’s policy landscape.

What’s Next?

My motto: The Future of Work is Flexible Benefits for a Flexible Workforce. At its core, the portable benefits reform movement recognizes a simple truth: current policies force workers to choose between flexibility and benefits. Flexible and portable benefits aim to bridge that gap—so that true independent workers and self-employed individuals can access benefits without being forced into a traditional employment structure.

This is a common-sense, non-partisan solution, and I expect we’ll continue to see growing interest from scholars, organizations, and lawmakers across the country.

At the state level, both Alabama and Tennessee passed portable benefits bills last month, joining Utah in leading the way on voluntary reform. Several other states are now showing interest in similar legislation. And at the federal level, given Senator Cassidy’s recent white paper and growing momentum, it’s very possible that federal legislation may be on the horizon.

For anyone new to these ideas, additional resources are below.

Additional Resources

• Independent Work and the Gig Economy: The Basics

  • My long-form study for American Enterprise Institute unpacking independent work data and policy reforms

  • Addressing the Real Challenges for Independent Contractors and the Gig Economy

  • Labor Myth Busting Series: The Gig Economy

  • A Fresh Look at the Independent Workforce with New BLS Data

• Women as Independent Workers

  • Celebrate Moms by Increasing the Availability of Flexible Work

  • The Future of Flexible Work is Female

• Portable Benefits Reforms

  • Policy Brief: Flexible Benefits for a Flexible Workforce 

  • Designing Benefits for Platform Workers (Jonathan Gruber)

  • Federal Policy Guide on Portable Benefits

  • State Policy Guide on Portable Benefits

  • Congressional testimony (April 2024)

  • Massachusetts testimony (March 2024)

  • Utah testimony (February 2023)

• Independent Work Reclassification Policies

  • Assessing the Impact of Worker Reclassification: Employment Outcomes Post–California AB5

  • New Study: From Gig to Gone? ABC Tests and the Case of the Missing Workers

  • New Study on California AB5

  • A Deep Dive into our Study on California’s AB5

Welcome to our new series on AI & Labor—a space to explore how intelligent machines are reshaping the very nature of work.

Over the coming weeks, we’ll share insights on how AI is transforming labor markets—led by our Labor Policy predoctoral researcher, Revana Sharfuddin.

“Leaning into our Humanity in the Age of AI” is Part 1 of this series.

Subscribe now

In the ongoing debate within labor economics, a central question is whether AI will ultimately benefit or harm workers. The answer hinges on a crucial distinction: will AI be deployed primarily to automate jobs—by completely replacing human workers—or to augment them, enhancing their productivity and capabilities? The balance between automation and augmentation will determine whether AI acts as a threat to job stability or a catalyst for new opportunities in labor.

We should not be helpless bystanders in this technological shift—we can influence AI’s direction before it defines ours. We can do this by examining early experimental research to gain insights that shape the conversation proactively.

The three key takeaways from research:

1. AI can significantly boost worker productivity—especially for lower-skilled or less experienced employees—by serving as a real-time assistant, not a replacement.

2. Even as automation advances, human traits like judgment, empathy, and creativity remain irreplaceable—and may become even more valuable in an AI-enhanced workplace.

3. If deployed thoughtfully, AI can reverse job polarization by empowering non-experts to perform higher-skill tasks, helping to rebuild the middle-skill workforce hollowed out by past technological shifts.

Is AI Augmenting Human Jobs or Automating Them: The Best Evidence Yet

A recent study by Brynjolfsson, Li, and Raymond finds that a GPT-3-powered AI tool augmented workers' performance, increasing productivity by 15%. Their paper, Generative AI at Work, was just published in the Quarterly Journal of Economics, one of the top journals in economics. This paper offers some of the best causal evidence to date on how generative AI affects workplace productivity and job quality—an area where most prior research has been more predictive than causal. While many studies have estimated AI’s potential impact by categorizing jobs and tasks, establishing true causal effects requires either experimentation, randomization, or well-designed observational studies that leverage quasi-experimental variation. That’s exactly what Brynjolfsson, Li, and Raymond provide.

Their study examines the rollout of a GPT-3-powered AI assistance tool among 5,172 customer service agents at a large (but unnamed) Fortune 500 business software firm. The researchers leverage a staggered implementation—often referred to in empirical work as “staggered rollout”—to compare workers who gained access to AI at different points in time. This allows for a natural difference-in-differences approach, creating a quasi-experimental setting.

Customer service is a strong test case for AI: high turnover, steep learning curves, and pressure from frustrated customers make it ripe for AI support—particularly with real-time response suggestions. Importantly, this wasn’t automation—agents weren’t required to use the AI’s suggestions, and they retained discretion over how to respond. This distinction matters: when AI performs a task that a human otherwise would have done, it forces us to ask whether that task truly requires human input at all. The broader question, then, is not just how AI augments work, but what work remains for humans when AI can handle specific tasks

Here is the breakdown of the 15% productivity gains:

• Agents spend 8.5% less time per chat.

• AI allows for better multitasking, increasing chats per hour.

• Event studies show an immediate boost after adoption.

• After the second month, productivity gains persist for at least five months (likely also end of study observation period).

And here is the most interesting finding: The least skilled agents (bottom quintile pre-AI) saw a 36% productivity jump, while top-performing agents saw almost no improvement. Similarly, new hires (with less than a month of experience) saw an approximately 34% improvement, while workers with more than a year of tenure showed no significant productivity gains.

What do these results tell us? AI doesn’t boost productivity evenly—it provides the biggest lift to those who need it most – less-experienced and lower-skilled workers. This is likely because AI serves as a real-time assistant, filling knowledge gaps, providing decision support, and reducing cognitive overload. Meanwhile, highly skilled and experienced workers, who already operate at peak efficiency, see little to no change.

This is exciting because it suggests AI could help reduce inequality—a topic I will explore more in a later section.

Typically, discussions around AI and labor focus on the risks to low-skilled workers, but this research flips that narrative. Rather than replacing them, AI appears to enhance their productivity, potentially increasing their value in the workforce. That’s a striking shift—one that could create new opportunities for those who have historically been the most vulnerable to technological change.

Of course, this also raises questions about how AI interacts with highly skilled, experienced workers. If their productivity remains largely unchanged while lower-skilled workers become more efficient, we may see some adjustments in wage structures over time. But the broader takeaway is clear: AI has the potential to level the playing field, making economic mobility more accessible.

Strength in our Humanness: Tacit Knowledge, Intuition, Empathy

In 1900, 41% of American workers were in agriculture; by 2000, that dropped to just 2% due to automation and mechanization. Innovations like tractors, assembly lines, and digital tools consistently replaced manual labor with machine precision. Yet, despite this trend, overall employment hasn't fallen. Why? Because automation may reduce labor per task, but it often increases the value of human work elsewhere.

The key lies in the strengths humans and machines each bring. Automation excels at routine, codifiable tasks, while humans thrive with problem-solving, adaptability, and creativity. Tasks requiring judgment or common sense remain hard to automate. A narrow focus on what automation replaces overlooks a fundamental economic dynamic: automation raises the value of tasks only humans can do. That’s why median wages have grown more than tenfold since 1820.

Economic progress has always depended on the interplay between labor and capital, intellect and muscle, precision and judgment. These elements don’t just coexist—they reinforce each other, shaping the evolving landscape of human productivity. When one improves, the others often become more valuable. As automation handles routine tasks, human traits like creativity, empathy, and adaptability matter more than ever.

Think of a disorganized kitchen—orders get mixed up, ingredients aren’t prepped, and dishes come out late— even the best chef can’t succeed there. But when the team runs smoothly, the chef’s expertise shines. Similarly, when AI improves routine tasks, it boosts the value of the human work that remains. When one part of a process becomes more efficient, the remaining tasks often become more valuable as they operate within a more optimized system. This is how AI, when integrated thoughtfully, can drive greater demand for labor.

A great example comes from banking. One study found that ATMs didn’t eliminate tellers—in fact, teller employment rose from 1980 to 2010. By lowering branch operating costs, ATMs indirectly increased demand for tellers: the number of tellers per branch fell, but urban bank branches grew (partly to deregulation on banking branches). Crucially, while routine tasks were automated, tellers shifted to “relationship banking”—advising customers and offering products using empathy and judgment.

Even when productivity shrinks a sector, overall employment may not fall—because income saved is spent elsewhere. When cars replaced horses, jobs didn’t disappear; new industries emerged, like motels and roadside diners. Rising incomes also boost demand in “technologically lagging” sectors, such as hospitality and personal services, which don’t directly benefit from automation. As technology lowers costs, spending shifts to new goods and services. Indeed, from 1940 to 1980, jobs shifted from physically demanding, low-paid work to skilled blue- and white-collar work (graph below).

Some specific takeaways over this time period:

• Agricultural employment fell by nearly 4 percentage points per decade

• Professional and technical jobs grew by 2.5 points

• Managerial roles grew by 0.5.

• Middle-tier jobs remained stable—clerical and sales roles expanded, while manual labor declined.

The data reinforce a clear pattern: as technology advances, employment doesn’t disappear—it evolves.

After the 1970s, however, job polarization emerged as automation increasingly displaced middle-skill occupations involving explicit, routine, and codifiable tasks. Computers excelled at these "routine tasks," reducing employment in clerical, administrative, and repetitive manual roles. Yet the substitution was limited: many human tasks remain inherently tacit—understood intuitively, impossible to fully codify, and thus difficult to automate.

David Autor’s take on Polanyi’s paradox highlights a key challenge for automation: we often perform tasks we can’t explain—like cracking an egg or recognizing a bird. These intuitive, tacit skills are hard to codify, making them difficult to automate. Computers handle logic well but struggle with common sense, judgment, and flexibility. That’s where AI shows promise—not by replacing these human abilities, but by amplifying them. As we saw in the earlier study, AI offered real-time suggestions, but human customer services agents ultimately decided how—and whether—to use them.

Augmenting human capabilities with technology creates vast new possibilities, as illustrated by Erik Brynjolfsson below. Machines perceive what humans cannot, act in ways humans physically cannot, and grasp concepts humans find incomprehensible. More profoundly, technologies that help humans invent better technologies not only enhance our collective capabilities, but also accelerate the rate at which these capabilities grow.

For instance, while AI may outperform radiologists at reading mammograms, it can’t handle the other 26 tasks required for the role—like comforting patients or coordinating care. In fact, across 950 occupations, Erik Brynjolfsson and coauthors found that although AI could perform certain tasks in nearly all jobs, there was no occupation that machines could fully handle alone. David Autor’s research also shows that as tech augments work, it creates new roles: across eleven of twelve broad occupational categories increases in augmentation-oriented innovations strongly predict the emergence of new occupational titles (figure below). About 60% of today’s jobs didn’t exist in 1940—a reminder of technology’s profound capacity to reinvent and expand human labor

Can AI Be a Force for Good in the Battle Against Job Polarization?

Although literature of income inequality in the US lack methodological consensus, it is clear from the patterns discussed in the sections above that technological shifts since the 1970s have contributed significantly through job polarization. Earlier waves of technological change automated a broad range of middle-skill occupations—jobs in administrative support, clerical work, and blue-collar production—pushing many of the approximately 60% of American adults without a bachelor's degree into low-paying, low-skilled service jobs.

A central reason for this widening gap is that information alone is merely an input for a much more valuable economic activity: decision-making. Since 1970s decision-making power has increasingly become concentrated among a relatively small group of highly educated professionals, typically those with college or graduate degrees. The widespread availability of cheap computational power amplified this concentration, dramatically raising the economic value of elite expertise at the expense of middle-skill workers.

Yet the unique promise of AI lies in its ability to reverse this pattern and broaden decision-making beyond a narrow elite. By combining information, explicit rules, and learned experience, it can amplify human judgment and enable workers with basic training to take on tasks once reserved for doctors, lawyers, or engineers. Rather than replacing expertise, AI can democratize it—revitalizing the middle-skill, middle-class core hollowed out by automation and globalization.

While some fear AI will replace human expertise, historical and economic patterns suggest otherwise. Like calculators or machines, AI tends to amplify, not substitute, human skill. Indeed, this mirrors the shift from artisanal work to mass production in the 18th and 19th centuries. Though craftspeople were displaced, new roles emerged—electricians, machinists, telephone operators—each requiring new forms of expertise built on literacy, numeracy, and training. As tools evolved, so did the value of mastering them.

However, the rise of the Information Age and the widespread use of digital computers after WWII reduced demand for “mass expertise”—the middle-skill jobs that powered the industrial age. Computers automated routine tasks, hitting middle-skilled roles hardest. Meanwhile, high-paid professionals like doctors and engineers remained insulated, as their work relied on judgment and creativity—qualities machines couldn’t replicate, as Polanyi’s Paradox suggests. This shift boosted elite productivity but displaced the middle class. Ironically, low-paid service jobs—like cleaning and caregiving—survived, deepening job polarization and income inequality.

AI holds real promise to reverse this trend. Unlike past automation, AI can augment human judgment instead of just replacing routine tasks. For example, an MIT experiment showed that giving writers tools like ChatGPT significantly improved both productivity and quality—especially among lower performers. Similar results appeared in the earlier customer-service study: AI boosted lower-skilled workers' output, narrowing gaps with experienced peers.

Crucially, AI didn’t replace expertise—it complemented it, requiring workers to apply judgment alongside AI’s suggestions. Thus, rather than deepening inequality, AI could democratize expertise, allowing a broader range of workers to participate meaningfully in tasks involving judgment, creativity, and decision-making—ultimately revitalizing the middle-skill workforce.

This snippet from the Macro Musings podcast with Dean Ball highlights how AI could challenge credentialism in academia. Ivy League professors (we know from literature are disproportionately drawn from university-educated families and higher socio-economic backgrounds) often have an army of research assistants, allowing them to outproduce skilled professors at mid-tier institutions (who probably do not share the same background characteristics). But what if all professors had access to their own AI-powered RAs? And what if human RAs had AI RA of their own? Deep research is exactly doing that. AI has the potential to level the playing field, expanding research capacity beyond elite institutions and democratizing academic productivity.

Conclusion

The future of work will not be defined solely by the power of AI, but by how we choose to use it. If deployed to augment rather than automate, AI can unlock human potential, expand access to opportunity, and rebuild the middle of the labor market. The evidence is clear: when paired with human judgment, AI improves outcomes—especially for those who have historically lacked access to expertise or upward mobility.

But realizing this promise will require intentional choices—from businesses, policymakers, and society as a whole. ​Rather than fearing disruption, we must embrace augmentation, foster resilience, and remain optimistic about the inclusive opportunities it presents.​ If we do, AI won't just change how we work—it will elevate what it means to work at all.

Subscribe now

The Federal Trade Commission (FTC) recently announced a new directive to form a Joint Labor Markets Task Force aimed at addressing “unfair labor-market practices that harm American workers.” This initiative merges the FTC’s consumer protection and competition mandates to tackle a range of concerning practices in labor markets, from wage-fixing and no-poach agreements to noncompete clauses and labor market monopsonies.

While the FTC’s focus on anti-competitive behavior in labor markets is commendable, the directive should also include an important source of market concentration: powerful labor unions with monopolistic characteristics. Just as we should be concerned about corporate monopolies that harm consumers, we should also recognize that excessive union market power can lead to outcomes that harm workers. 

When Labor Becomes Concentrated: The Case of America's Ports and Parcel Delivery

There is an inherent tension between competition and labor unions that has led to decades of legal controversies and judicial inconsistencies about whether certain union practices breach antitrust laws. The two examples below–dockworkers and UPS workers–illustrate how competition law is tested when just one or two powerful labor unions control a substantial share of an industry’s workforce.

The first example is the case of dock workers. Currently, the International Longshoremen's Association (ILA) represents dockworkers at all major ports along the East and Gulf Coasts. Similarly, the International Longshore and Warehouse Union (ILWU) represents workers at all major West Coast ports. Together, these two unions control labor at virtually all major U.S. ports—giving them extraordinary market power over the flow of goods throughout the economy.

Of the 39 major ports (Table 5) across the entirety of the United States, only 1 of them is not unionized by either ILA or ILWU. These two unions alone control labor representation at 97 percent of America's major ports. Even if ports with limited or partial ILA representation are excluded, these two unions still represent 82 percent of all major U.S. ports—a remarkably high concentration. 

The ILA’s partial or limited representation refers to ports with one or more terminals that have ILA contracts (often for container or Ro-Ro cargo), while specialized bulk or petroleum terminals may be under other unions or non‑ union labor. These limited representation ports are typically in the Gulf or Great Lakes regions–for example, the ILA has limited presence at Corpus Christi because that port handles more petroleum and bulk cargo than containers. 

Upon closer examination, we also see that 100 percent of all major U.S. ports on the East Coast are represented by one single union, the ILA. Likewise, on the West Coast, a single union—the ILWU—controls every major port. Such dominance clearly falls far short of any reasonable definition of a competitive market. 

Other unions typically have much smaller representation in the dockworker category, often focusing on specialized roles or facilities. In fact, if we include all dockworkers across the U.S, ILA and ILWU remain the dominant unions, with approximately 85-90% of all unionized dockworkers belonging to one of these two organizations. 

Furthermore, the ILA monopoly seems to also meet the “bad acts” requirement under antitrust laws. In October 2024, the ILA initiated a major strike not just to improve wages and benefits of workers, but also to ban any use of automation at the ports—including, even, automated gates. The ILA’s strike shut down every major port from Maine to Texas—a move its president described as an intentional way to “cripple the economy.” This isn't an exaggeration; when essential infrastructure is controlled by a single labor organization, the economic consequences of work stoppages can be dire.

We wouldn’t accept this behavior from corporate monopolies, yet we seem to overlook this behavior when it comes to labor union monopolies. 

The International Brotherhood of Teamsters presents another striking example of labor market concentration. The union represents between 97 and 99 percent of all UPS workers that handle or transport packages (generally includes drivers, sorters, loaders, and administrative personnel at the company's operational facilities).1  According to recent data, the Teamsters’ membership at UPS now tops 340,000 workers, effectively accounting for over 98% of UPS’s total bargaining-unit employees. 

This concentration of labor power in a single organization gives it substantial leverage not just over UPS, but over the entire ground delivery ecosystem.

These unions’ monopolistic control over critical sectors raises important questions about the balance between worker representation and market competition. While the FTC's directive identifies “labor market monopsonies” as harmful when created by employers, it should also  acknowledge that similar concerns arise when labor organizations concentrate market power on the supply side.

The Evidence on Union Monopoly Power and Worker Outcomes

My new upcoming research (co-authored with our predoctoral researcher Revana Sharfuddin) on labor unions reveals that union monopoly power doesn't necessarily translate to better outcomes for workers. While powerful unions historically secured higher wages for their members, these apparent "big wins" at the bargaining table often came with significant trade-offs for workers. 

The evidence shows that union monopoly power doesn’t just boost wages indefinitely; in fact, when unions push too hard, this backfires and results in slower employment growth and fewer job opportunities for unionized workers. Companies facing excessive union demands often cut investments—particularly in R&D and long-term capital—reducing productivity and profitability. This weakens businesses over time, increasing the likelihood of downsizing or closure, which ultimately leads to job losses and a smaller unionized workforce. These dynamics played out in the Rust Belt, where powerful unions and frequent labor conflicts were responsible for approximately 55% of the region's manufacturing employment decline. 

The pattern continues today. After the Teamsters secured a significant contract for UPS workers in 2023, the company announced in early 2025 a “network reconfiguration” that “could result in the closure of up to 10 percent of our buildings... and a decrease in the size of our workforce.”  Industry analysts estimate this may ultimately reduce UPS’s full-time headcount by nearly 10,000 positions, illustrating the potential cost of steep labor concessions. Similarly, Boeing faced a seven-week strike by the International Association of Machinists and Aerospace Workers that cost the company $9.7 billion, followed by an announcement to cut 10% of its workforce.

These examples highlight a fundamental economic reality: When unions use their monopoly power to extract concessions that companies find unsustainable, workers often end up worse off through reduced employment opportunities.

But it doesn’t have to be this way. Several countries in Europe demonstrate that unions can be effective when they are more collaborative, flexible, moderate, and balanced–thereby mitigating the negative effects on workers while ensuring the long-term viability and survival of both the union and the company. As part of our upcoming study, we found that the monopolistic characteristics of labor unions are formed because of U.S. laws and bargaining structures. Indeed, in some European countries, labor unions are much less monopolistic and yet more pro-worker. 

Applying Antitrust Principles to Labor Markets—Including Unions

The FTC’s new directive represents an opportunity to take a more comprehensive approach to labor market competition. Historically, labor unions received exemptions from antitrust laws through the Clayton Act of 1914 and the Norris-LaGuardia Act of 1932, which prioritized unions’ role in counterbalancing employer power.

However, the modern economy calls for a fresh assessment of how we balance worker representation with the benefits of competition. Just as the FTC scrutinizes corporate mergers that could harm consumer welfare, it should consider the anticompetitive effects when a single union controls a significant share of an industry's workforce.

Indeed, the FTC’s Bureau of Economics and Office of Policy Planning are both positioned to play a key role in researching labor markets to identify barriers to competition—including those created by government laws and regulations. By studying these dynamics, the FTC can publish research and spotlight how certain government-imposed rules or union protections may inadvertently stifle competition and harm workers.

To address unfair union monopoly practices and provide a more balanced approach, a new legislative framework is likely needed. What could it look like? Several possibilities merit consideration:

1. Limits on union market share: Perhaps we should reconsider allowing a single union to represent more than a certain percentage (say, 30-40%) of workers in a given sector. For instance, a cap on labor market share—similar to the Herfindahl-Hirschman Index (HHI) used for corporate mergers—could be applied to ensure no single union far exceeds this threshold.

   1. One simple variation of this principle could be as follows: Stipulate that one single union cannot represent two or more companies’ employees if antitrust laws would also prevent those same companies from merging.

2. FTC review of multi-employer bargaining units: When a single union negotiates with multiple employers simultaneously, this may function similarly to price-fixing arrangements that antitrust laws normally prohibit. 

3. Restrictions on ‘most favored nation’ clauses: These provisions mandate uniform terms across employers, potentially stifling competition and innovation in labor markets.

   1. Indeed, in many cases, courts and regulators have scrutinized ‘most favored nation’ clauses in labor agreements for anti-competitive effects, especially when they discourage employers from negotiating better deals with new labor entrants or competing unions. 

4. Greater flexibility for employers to enter or exit multi-employer bargaining arrangements: This would enhance competition while preserving workers' collective bargaining rights.

A Path Forward for Competitive Labor Markets

Workers deserve both fair treatment and robust work opportunities. The evidence in our upcoming study suggests that moderate union power—characterized by balanced demands and increased flexibility—better preserves the benefits of collective bargaining while avoiding the downsides of monopolistic union behavior.

The FTC’s new Labor Markets Task Force presents an opportunity to take a truly comprehensive approach to labor market competition. By acknowledging that anti-competitive behavior can come from both employers and labor organizations, the FTC can develop policies that protect workers not just from anti-competitive corporate behavior, but also from the harmful effects of labor market concentration.

If we truly want pro-worker policies, we need to recognize that concentrated power—whether wielded by corporations or unions—rarely benefits workers in the long run. True pro-worker policy embraces both fair representation and healthy competition in labor markets.

Subscribe now

Today, we’re celebrating International Women’s Day by recognizing how the labor market has evolved to better support women of diverse backgrounds and needs. The Labor Policy team at Mercatus is a team of three women who benefit from a dynamic labor market in different ways. Our team Director Liya is a working mom of two young children, our pre-doc Researcher Revana is an immigrant, and I’m the Program Manager, supporting my husband through grad school, while also managing my own academic pursuits. 

While our needs and wants from the labor market are different, all three of us rely on an innovative, flexible, and dynamic labor market to help us balance the unique trade-offs we face at work and in our personal lives. Today, we want to highlight three major innovations that are distinctly benefiting women (ourselves included): artificial intelligence, flexible work, and portable benefits. 

Artificial Intelligence 

In a 2010 TedTalk, Statistician Hans Rosling gives the “magic washing machine” as an example of how economic growth and technological innovation have dramatically improved people’s lives. He says proudly, “to my grandmother, the washing machine was a miracle.”

While we have all grown accustomed to the washing machine, Rosling emphasizes an important point: technological innovations have been making work inside and outside the home easier since the Industrial Revolution. The advent of email, spreadsheet software, and video conferencing were all major innovations that we now consider a mundane part of our day at work. 

We are experiencing the next technological innovation with the explosion of artificial intelligence. AI has made our team more productive through big things like improving our coding abilities and small things like taking meeting notes. Overall, the data points to AI improving worker productivity – one study showed that access to AI tools increased worker productivity by 14% and increased productivity for novice and low-skilled workers by 34%. 

Increased productivity isn’t the only valuable economic benefit from AI. AI also helps increase women’s employment and attachment to the labor force. A study published in NBER found that exposure to AI, especially in occupations more susceptible to AI enabled automation, increased the share of female employment by 2.2-2.9% overall. This effect is higher in countries with higher female educational attainment and where women’s labor force participation rates are already high. AI has the potential to greatly enhance women's economic opportunities and help close gender gaps in professional and technical fields. 

While there are concerns about how automation can replace jobs, it appears that, so far, AI is proving itself to be more of a complement to the labor force than a substitution. 

The washing machine was so incredible because it freed up women’s time. No longer were women forced to scrub their clothes on a washing board, they could put clothes in the wash and walk away. The same is true for artificial intelligence. Workers can delegate tasks to AI and focus their attention on more complex, cognitive skills which in turn increases their productivity. For women, this results in better labor market outcomes. Indeed, all three of us with the Labor Policy project use and greatly benefit from AI, experiencing firsthand how these tools enhance our productivity and professional development. 

Flexible Work

Flexible work arrangements make it easier for moms to juggle parenting and their careers. Years of economics research show that women tend to self-select into more flexible jobs, largely because work needs to be structured around taking care of children. The rise of part-time and intermittent work contributed to the growth of women in the labor force throughout the 1980s and 90s. 

Women value flexibility so much that many women would take a pay cut in order to get more flexibility at work. Work flexibility may also help moms move back into the workforce. One survey found that nearly two-thirds (62%) stay-at-home mothers said that they would need work flexibility in order to go back to work. 

Flexible work also makes it easier for young couples to decide to have children. During covid, when many young couples were working from home, the fertility rate for young women rose with many women citing their unique ability to adjust to the changes a baby would bring in the age of forced remote work. Flexible work may also increase the number of children families have. Indeed, 33% of young moms feel that their desire to work combined with inadequate childcare support contributes to not wanting to have more children. Accessibility to flexible work arrangements – which young workers are less likely to have – can help bridge the gap between childcare and desire to work. 

Not every woman demands flexibility in their work, but for some women it is a benefit they can’t manage their work and personal life without. This is why a dynamic labor market generating a diversity of jobs is so important – dynamism allows women, and all workers, to pursue jobs that work for them and their families. 

Flexible and Portable Benefits 

While work-from-home is one important type of flexible work, the growth in nontraditional work like freelancing, app-based, or self-employment, is providing another important avenue for women. Today, anywhere from 11 to 60 million people work in the independent workforce and many of those workers are mothers of young children. Many women have chosen to leave their traditional 9-to-5 jobs to pursue independent work, mainly because they prefer flexibility and work autonomy. There is great diversity in what these women choose to do – some women may choose to shop for groceries with Instacart or set up online shops on Etsy or provide technical services on Fiverr. But all are examples of women making a living in the independent workforce. 

However, our labor policies are not set up to support these types of work. Laws that tie benefits like health insurance or retirement plans to W-2 employment go back for decades. When these laws were created, the majority of workers were traditional employees. But now, our antiquated labor laws create a conundrum for women and other independent workers: take an inflexible job with benefits or a flexible, independent job without benefits. 

To address this, state and federal policymakers have championed the idea of flexible and portable benefits – or benefits tied to the worker, instead of the employer. In 2023, Utah was the first state to pass a law that removed the presence of benefits as a condition to determine if someone was an independent worker or a W-2 employee. This allowed organizations like Shipt and Lyft to provide benefits to workers in that state. Copycat bills have popped up across the country and a similar bill has been introduced at the federal level by Rep. Kevin Kiley. 

Portable benefits remove the false dichotomy between work flexibility and access to benefits. This can be an especially powerful tool to meet the needs of working moms. 

The Future of Work is Female

The innovation and dynamism of the labor market are worth celebrating, but it’s especially worth celebrating this International Women’s Day when we think about how much these changes have benefited women. Rarely do academics get to feel so closely connected to their research, and we all feel that it's a privilege to study and walk in the shoes of women in the workforce.

Most notably, stricter worker classification laws, such as the ABC test, have emerged as the key response to tackle potential cases of misclassification (e.g., when a worker is functionally an employee but legally classified as an independent contractor). An ABC test limits the circumstances under which a worker can legally work as an independent contractor. Supporters of ABC test laws argue that those laws can address misclassification and induce employers to reclassify independent contractors as employees to comply with the law, thereby increasing the share of workers who are employees. 

The challenge is that ABC tests make it more difficult for workers to engage in freelancing, gig work, or other types of independent contracting roles, even if those workers are already properly classified. This is partially because state policymakers have adopted broadly applied ABC tests that impact a wide array of industries, even in occupations and cases where there is no misclassification. 

For example, when California codified their ABC test into law with Assembly Bill 5 (AB 5) in 2019, it led to a widespread backlash from legitimate freelancers in California who were no longer able to work (see, for example, the reports by Los Angeles Times, The New York Times, and Business Insider). As a result, California had to exempt over 100 industries and occupations from Assembly Bill 5. So, while policymakers aimed to address true misclassification, the ABC test functionally led to an overkill of properly classified independent contractors who lost their jobs. 

Despite the interest and growth in ABC tests, there are no empirical studies that investigate the causal impact of those laws. Last year, our team at Mercatus published a study on California’s AB5–it was the first of its kind analyzing the employment outcomes in California post-AB5. Our AB5 study found that employment and self-employment fell in California for affected occupations post-AB5, but our study did not yield causal estimates due to the noisy nature of the data available (see our previous Substack posts “New Study on California AB5” and “A Deep Dive into our Study on California’s AB5”). 

We now have those causal results. Coauthored with Markus Bjoerkheim, our new study provides the first causal analysis of state ABC test laws on labor market outcomes. 

Our study asked the following question: What is the impact of an ABC test on traditional (W-2) employment, self-employment, and overall employment in states that have adopted these tests?

Our key results:

• The introduction of an ABC test caused significant declines in traditional (W-2) employment, self-employment, and overall employment.

• The ABC test reduced traditional (W-2) employment by 4.73%

• Self-employment fell by 6.43%

• Overall employment fell by 4.79% 

• Occupations with high shares of independent contractors experienced the largest reductions in employment.

These results suggest that contrary to the intended goal, ABC tests are not altering the composition of workers and leading to more workers becoming traditional W-2 employees, but they are reducing employment for both W-2 employees and self-employed workers. 

Our full study will be published in the coming months. In this post, we preview our findings. 

How did we do the study?

To evaluate the labor market impacts of the ABC test, we utilized what’s called the Callaway and Sant’Anna Difference-in-Differences (DiD) approach and an event-study framework tailored to settings with staggered policy adoption. These methods allowed us to leverage variation in the timing of the ABC test's implementation across states to identify causal effects, without introducing biases from diverse treatment effects or treatment timing. 

We constructed a state-level annual panel dataset using the Bureau of Labor Statistics and Census Bureau’s Current Population Survey (CPS) monthly data from January 1990 to September 2024. This dataset allowed us to analyze employment outcomes, including overall employment, traditional (W-2) employment, and self-employment.

To analyze the worker-level labor market effects of the ABC test, we designated states that adopted the ABC test as the treatment states, and states that continued using right-to-control or common law standards as the control states. Through our legal history review, we identified nine states that implemented the ABC test for any domain, and we recorded the timing of the test’s adoption and the severity of its application. In some instances, the ABC test had been on the books for decades but only became applicable during our study period due to developments in case law. We consulted with a law firm to confirm the ABC test states and the timing of when the ABC test went into effect. These states are New Jersey, Massachusetts, New Mexico, Hawaii, Illinois, New Hampshire, Maine, Nevada, and California. 

What do our results mean? 

Our results show that when a state adopts an ABC test, it causes a reduction in traditional (W-2) employment, self-employment, and overall employment in both the short-run and the long-run (15 years). 

Our key results:

• The implementation of an ABC test caused a 4.73% decrease in W-2 employment, highlighting a substantial shift away from traditional employment forms, which is contrary to the goals of policymakers. 

• The implementation of the ABC test led to a 6.43% decline in self-employment. 

• The implementation of the ABC test caused a 4.79% reduction in overall employment in the state.

These numbers represent changes in employment in ABC test states relative to our control states (states that had the common law test). Occupations with high shares of independent contractors experienced the largest reductions in employment. 

Below we have event study graphs that depict the negative effects of the ABC test on traditional (W-2) employment, self-employment, and overall employment over time, which also support our findings from our main DiD approach. The vertical dashed lines in those graphs represent the time the ABC test went into effect, and the y-axis coefficients show the estimated change in employment relative to the pre-implementation period (e.g. before the ABC test went into effect). 

1. ABC Test Impact on Traditional (W-2) Employment

2. ABC Test Impact on Self-Employment 

3. ABC Test Impact on Overall Employment

Both overall employment and traditional (W-2) employment exhibit a significant and sustained decline after adoption, worsening notably over the subsequent 5–15 years. In contrast, self-employment shows an immediate sharp drop following implementation, followed by a slight short-term rebound before experiencing a prolonged and pronounced decline. 

These results highlight the ABC test's substantial negative effects on employment outcomes, with the largest negative impact on self-employment. While the decline in self-employment aligns with expectations that the ABC test makes it more difficult to work as an independent contractor, the reduction in traditional (W-2) employment was a surprise. 

There are several reasons why an ABC test would cause traditional (W-2) employment to fall, though we could not inquire into these mechanisms given the nature of our data. One possible explanation is that businesses that had workers on payroll and also relied heavily on independent contractors shut down because of the ABC test (therefore causing W-2 employment to also fall). This is supported by anecdotal evidence coming out of California where employers started closing down their offices that had both W-2 employees and contractors  (e.g., some of these stories). Therefore, businesses that had both W-2 employees and independent contractors moved to other states that don’t have an ABC test. Over time, there is a growth in W-2 employment in non-ABC test states relative to ABC test states. 

Finally, we employed several robustness checks and all results remained consistent. Additionally, we excluded states that passed an ABC test immediately before, during, or after the financial crisis and the pandemic. Our findings are consistent even when excluding the financial crisis and pandemic environment.

Conclusion

Worker classification policies, such as the ABC test, have emerged as a key response to concerns about misclassification of gig workers and independent contractors. These policies aim to reclassify more workers as traditional (W-2) employees, granting them access to benefits and protections. 

Our study provides the first causal analysis of the ABC test to measure its impact on labor market outcomes in the United States. 

Our findings show that the ABC test does not lead to more workers becoming W-2 employees. Instead, an ABC test caused significant declines in employment: traditional (W2) employment fell by 4.73%, self-employment fell by 6.43%, and overall employment fell by 4.79%. Occupations with high shares of independent contractors experienced the largest reductions in W-2 employment. 

These results suggest that ABC tests are not altering the worker composition, with more workers becoming traditional W-2 employees as intended, but they are reducing employment for both W-2 employees and self-employed workers.

Earlier this month, the Bureau of Labor Statistics published the first news release on the much-anticipated survey on the gig economy. The Contingent and Alternative Worker Supplement (referred to as the CWS), which was last released in 2018, provides an official estimate on the size and characteristics of the independent workforce.

As part of her role on the Data Users Advisory Committee for the BLS, Liya had the opportunity to provide recommendations on updating the survey to better capture the evolving nature of independent work. Similarly, in 2019, the Committee on National Statistics appointed an expert panel, chaired by Susan Houseman (Upjohn Institute), to evaluate the shortcomings of the CWS and propose recommendations for its modernization.

Fortunately, the BLS implemented several of these suggestions. Most notably, for the first time in the survey’s history, they included questions about workers using alternative work arrangements as a secondary or supplementary job. Historically, the CWS focused exclusively on workers in their main or sole jobs.

Other surveys and administrative data sources show that the largest share of independent contractors are supplemental earners, especially those who use apps such as Uber, Lyft, DoorDash, and Instacart. App-based work was not included in this CWS release. The BLS intends to publish these additional estimates in the coming months.

In this post, we highlight some of the key findings on independent workers from the CWS survey. The results show that independent work as a primary job is growing, and that Professional and Business Services continues to be a top industry for these workers.

Some top-level results:

• Approximately 11.9 million workers, or 7.4% of total U.S. employment, identified as independent contractors on their sole or main job, an increase from 6.9% in 2017.

  • However: This is understating the true size of the independent workforce since it’s only reporting on workers who use independent work as their primary source of income. The size of the independent workforce likely ranges from 11.9 million on the low end to 65 million on the high end (see Page 6, Table 1). Also, in an older WSJ article, Liya discussed similar issues with the 2018 CWS.

• As in prior surveys, independent contractors overwhelmingly preferred their alternative work arrangements over W-2 employment (80.3 percent), whereas only 8.3 percent would prefer traditional, W-2 employment arrangements.

• In 2023, 8.4 million people reported holding multiple jobs, representing about 5.2% of total employment.

  • However: This figure likely understates the true size of this group. Below we explain why the survey was unable to fully capture this data.

• 74.2 percent of independent contractors had health insurance, compared to 84.9 percent of workers in traditional W-2 arrangements.

Trends Over Time: Growth in Independent Contracting

The contingent and alternative work arrangements survey is a supplement to the Current Population Survey (CPS) and has been conducted in 1995, 1997, 1999, 2001, 2005, 2017, and most recently in 2023. The CWS has two key parts of the survey—the first on contingent work (a temporary job, like a seasonal tour guide), and the second part on alternative work arrangements, which include independent contracting (e.g. freelancing, gig work, consulting), on-call work, temporary help, and contract firm employment.

The chart highlights the steady rise in independent contracting as a main or sole job, which increased from 6.7% of total employment in 1995 to 7.4% in 2023. Meanwhile, on-call work has remained relatively stable at around 1.7%, and temporary help agency work has declined from 1.0% in 1995 to 0.6% in 2023. Contract firm employment has fluctuated slightly but continues to represent a small segment of the workforce.

It’s unlikely that the rise in app-based delivery and transportation work is the reason for the growth in independent work as a primary job because app-based delivery and transportation workers tend to be supplemental earners (see Myth #2 here). An IRS tax data report also studied this question and concluded, “we find that the exponential growth in labor [online platform economy] work is driven by individuals whose primary annual income derives from traditional jobs and who supplement that income with platform-mediated work.”  Furthermore, another IRS tax data report on independent contracting found that the significant growth in independent contracting since 2001 pre-dates the rise of the gig economy—similar to what the CWS data shows here.

Readers should be aware that CWS data is frequently misunderstood by the media, with many mistakenly equating independent contracting or alternative work arrangements with modern app-based gig work, such as ridesharing and delivery. However, app-based gig work represents less than 10% of independent contracting (see Myth #1 here). Additionally, the CWS has yet to release specific data on app-based gig economy jobs.

Lastly, 11.9 million workers is an understatement of the size of the independent workforce, as this figure only captures workers engaged in independent contracting as their primary job. Evidence from tax data suggests that the majority of independent contractors in the U.S. are supplemental earners, rather than primary income earners.

Industry Insights

The survey highlights the share of all independent contractors working in specific industries, here are the top sectors:

1. Professional and Business Services: This industry remains the largest for independent contractors. In 2023, 24.1% of all independent contractors were working in Professional and Business Services, up from 21.3% in 2005.

   1. This is consistent with IRS tax data sources which found that ‘Professional, Scientific, and Technical services,’ a sub-industry of Professional and Business Services, is the sector with the most independent contractors and has seen the greatest growth in independent contracting since 2001.

2. Construction: Construction continues to be a key industry for independent contracting, with 17.8% of all independent contractors in Construction in 2023, however, this represents a decline from 19.3% in 2017 and a more significant drop from 22.0% in 2005, suggesting a gradual shift away from this industry in terms of independent contracting.

   1. We have not see a formal assessment on this yet, but it’s possible that this may partially be a reflection of state laws that have made it more difficult to be classified as an independent contractor in construction.

3. Financial Activities: 10% of independent contractors were working in financial activities in 2023, which has remained at about the same share in previous years.

Another way to analyze industry trends is to examine the percentage of workers in each sector who are independent contractors. From this perspective, the leading industry for independent contracting is Real Estate, Rental, and Leasing, where 24.2% of all workers are independent contractors (see chart below). In other words, nearly one in four workers in this sector is an independent contractor. With this new perspective, Construction (18.5%) and Professional and Business Services (13.5%) are still among the top 5 industries.

Demographic Trends: Independent Work Opportunities for Older Worker

The CWS data shows a significant upward trend in the employment share of independent contractors among men and women aged 55 and older from 1995 to 2023. Independent contracting as a main or sole job for older workers has increased by approximately 50 percentage points since 1995, while W-2 arrangements for older workers has increased by roughly 20 percentage points in the same time period.

Independent work can be valuable as a main source of income for retired or older workers who have moved past the “9-5” routine but remain open to transitioning to part-time or short-term flexible work. A 2020 study published in the American Economic Journal: Macroeconomics found that for many older workers, labor force participation near or after normal retirement age is limited more by a lack of acceptable job opportunities or low expectations about finding them than by unwillingness to work longer. The study found that about 60 percent of nonworking older respondents said they would be willing to return to work with a flexible schedule.

The CWS data on older workers is consistent with previous research—for example, a 2019 NBER study found that the incidence of self-employment among older workers was even higher than was previously measured. Conditional on working, the share of individuals with their main job as self-employment is less than 20 percent for the age groups 18-29 and 30- 49. But the share of individuals with self-employment income as a main job is 45 percent for 65-69 and almost 60 percent for those aged 70-74.

For many older workers, independent work can be an attractive way to ease into retirement or earn income when they are no longer part of the full-time, employment workforce. This is especially appealing to older individuals who already have healthcare coverage.

The Majority of Independent Workers Prefer Their Work Arrangements Over W-2 Employment

The 2023 data shows that 80.3% of independent contractors prefer their current arrangement over traditional, W-2 employment, an increase from 79% in 2017. In other surveys, workers cited dependent care obligations, personal circumstances, or a strong preference for job flexibility (over job stability) as the primary reasons. In particular, working mothers tend to prefer independent work arrangements because flexible work makes it easier for moms to juggle both parenting and their careers (see my previous posts on ‘The Future of Flexible Work is Female’ and ‘Celebrate Moms by Increasing the Availability of Flexible Work’).

The Puzzling Story about Multiple Job Holders

For the first time, the CWS surveyed multiple job holders: in 2023, 8.4 million people reported holding multiple jobs, representing about 5.2% of total employment. The most common alternative arrangement for a second job was working as an independent contractor, representing 22.8 percent of multiple jobholders.

The number of multiple job holders seems unusually low. Moreover, almost all other data sources indicate that independent contractors working as supplemental earners make up the majority of the independent workforce.

Susan Houseman, a member of the BLS Technical Advisory Committee, discussed this issue in a public BLS meeting, noting that the CWS survey asks workers about their jobs in the week prior. However, workers who have side jobs may not work those jobs every week; rather, they tend to work sporadically, especially those in the app-based gig economy. Additionally, the CPS generally identifies a much smaller population of secondary job holders, whereas other official data sources report higher numbers. For example, the U.S. Census Bureau’s Longitudinal Employer-Household Dynamics (LEHD) program finds that the multiple jobholding rate in 2018 was 7.8 percent. In the past, the Census Bureau’s Survey of Income and Program Participation (SIPP) also reported that there were 13 million U.S. workers holding more than one job, representing about 8.3% of workers.

We will dedicate a future Substack post to further analyzing the multiple-job holding component of the CWS once the BLS releases the raw data.

Looking Forward: Flexible Benefits for a Flexible Workforce

A 2020 Journal of Economic Perspectives study found that about 80% of self-employed workers indicated that they desire portable benefits—benefits that are not tied to one particular employer and can travel with the worker.

States are already experimenting with various portable benefits models. Last year, Liya testified before the Utah Legislature on a portable benefits bill that eliminates the presence of benefits as a factor in worker classification tests. This bill was passed and went into effect in May 2023. As a result, Target’s Shipt launched a pilot benefits program in Utah in partnership with the benefits company Stride. Similarly, with the support of Pennsylvania’s Governor, DoorDash introduced a first-of-its-kind pilot portable benefits program, also managed by Stride. Other portable benefits models, such as those in California and Washington state, mandate a specific set of benefits for contractors in app-based delivery and transportation sectors.

Congressional lawmakers are increasingly showing interest in these ideas. Earlier this year, Liya testified before the House Committee on Education and the Workforce about creating a safe harbor for states to experiment with portable benefits models. She also submitted a list of recommendations in response to Senator Cassidy’s request for information on portable benefits.

The key concept behind flexible benefits reforms is the recognition that current policies force workers to choose between independent work and access to fringe benefits. Flexible and portable benefits reforms aim to bridge this gap, enabling true independent workers and self-employed individuals to gain access to benefits.

To support independent workers in accessing flexible and portable benefits, Liya and Jonathan Wolfson recently published both a Federal and a State Policy Guide. These guides build on Liya’s previous research, including her 2023 policy brief and 2020 study on barriers to portable benefits solutions.

In general, labor unions are meant to be a force for good, providing a much-needed voice and representation for workers in negotiations with their employers. However, challenges arise when bargaining demands become unreasonable or unfair–perhaps the demands from the labor union are too high or the company’s offer is unreasonably low. 

Labor unions can turn into “villains” when their demands become excessive and extortive. 

This is why the International Longshoremen Association (ILA), the only labor union controlling all ports across the entirety of the east coast, is facing a public backlash for demanding “a total ban on the automation of cranes, gates and container movements that are used in the loading or loading of freight at 36 US ports.” This demand is in addition to other requests on compensation – for example, significantly higher wages and benefits packages. 

As many have pointed out, companies have already agreed to offer reasonably higher compensation packages to port workers–increasing wages by nearly 50%, tripling employer contributions to employee retirement plans, strengthening health care options, and even retaining the current language around automation and semi-automation. 

However, in order to force companies to meet their unreasonable demands on the use of technology and even higher compensation packages, the ILA is threatening the country by locking down 36 ports on the east coast with more than 45,000 workers on strike starting today. Economists have already discussed the severe negative effects consequences this will have on supply chains, prices, and even jobs for Americans in other sectors. But perhaps the best illustration of how American workers will be worse off during these port strikes comes from the ILA union President:

“When my men hit the streets from Maine to Texas, every single port locked down. You know what's going to happen? I'll tell you. First week, be all over the news every night, boom, boom, second week. Guys who sell cars can't sell cars, because the cars ain't coming in off the ships. They get laid off. Third week, malls are closing down. They can't get the goods from China. They can't sell clothes. They can't do this. Everything in the United States comes on a ship. They go out of business. Construction workers get laid off because the materials aren't coming in. The steel's not coming in. The lumber's not coming in. They lose their job.”

This interview highlights the trade-offs that any pro-worker agenda must grapple with (as I’ve discussed in a previous post). When the ILA threatens to cripple other jobs in the economy, it is illustrating that not all labor unions are synonymous with being ‘pro-worker.’ Indeed, the ILA is okay with other workers in America losing their jobs in order to pay for the excessive demands of this select group of workers. This is hardly an inspiring message to join the American labor movement. 

Of course, companies are right to reject the excessive demands, especially on a *total* ban of automation. Making U.S. ports worse-off is not a winning strategy for workers, and especially not for U.S. port workers. 

If there is a total ban on automation, cargos will be redirected to more efficient and lower cost ports as east coast ports become less efficient and more costly compared to other ports that welcome technological advancements. Over time, the 36 east coast ports that are stuck in the Stone Age will see less business, slower employment growth, more lay-offs, and some may even close down completely. 

Blocking technological advancements is not only bad for the East Coast ‘Stone Age’ port workers, but it’s also bad for America’s competitive edge. Asian and Middle Eastern ports which do welcome technology will continue to grow faster in the long run compared to the U.S. ports. For those who sympathize with MAGA goals, banning technological advancements on U.S. ports is a great “American Last” strategy. 

For those who may seem unconvinced of this future reality, look no further than the case study of the Rust Belt’s economic deterioration in the second half of the 20th century. 

A Warning to Port Workers: How Excessive Labor Union Demands Ruined the Rust Belt

A new study published in Journal of Political Economy showed that excessive demands from strong unions, which led to a high frequency of strikes in the Rust Belt region between 1950-2020, are one of the main reasons for the decline in manufacturing employment in the Rest Belt during that time period. 

Rust Belt manufacturing industries had, by far, the highest rates of work stoppages from strikes in the United States. According to CPS data from the Bureau of Labor Statistics, between the years 1973 and 1980, 48.1% of workers were union members in Rust Belt manufacturing compared to 28.4% among manufacturing workers in the rest of the country. As the economists note, “While unionization rates in manufacturing were around twice as high in the Rust Belt as outside, rates of work stoppages were about seven times as high in the Rust Belt. In other words, labor relations were particularly fraught among unions concentrated in the Rust Belt.”

During this time, the compensation package demands were also higher than usual as compared to the demands in other regions. 

So what happened? Excessive union demands, strikes, and work stoppages led to lower average rates of investment and productivity growth by Rust Belt firms relative to those in the rest of the country. Over time, employment moved from the Rust Belt to the rest of the country. 

This is common sense. Rust Belt firms faced increasing labor costs and reduced productivity, forcing many to either downsize or relocate to regions with less union influence and lower labor costs. 

Overall, these labor union conflicts accounted for 55% of the decline in the Rust Belt's share of U.S. manufacturing employment, leading to its economic deterioration throughout the second half of the 20th century. 

Many might sensibly ask, “but what about the impact of China and globalization on the Rust Belt?” The study accounted for this as well, finding that trade had a secondary, negative effect on the Rust Belt. As they write (pg. 2783):

Rising imports have virtually no employment effect until the mid-1970s, and the losses are concentrated in the 1980s and 1990s. This suggests that international forces at best played a supporting role in the Rust Belt’s decline in the latter part of the time period and likely had little to do with the large secular decline in the region’s employment share that occurred in the first 3 decades after the end of World War II.

The Rust Belt case study is a perfect example of when unions go too far, leading to worse worker outcomes in the long-run. 

Other studies also show that unions which moderate demands to avoid pushing firms out of business or reducing employment are more likely to persist over time. This moderation creates a balance between gains for workers and ensuring the firm’s survival, benefiting both parties in the long run.

Port Automation: Yes, Some Port Jobs Will be Lost, but Fear Not

Economists generally agree that real wage growth is feasible when there are technological developments that make workers more productive, like how our generation experienced the ‘computerization’ of white collar jobs, or when the construction industry started using construction vehicles and other machines. These developments are welcomed because they enhance the skills of workers, making them more productive and allowing them to increase their output for any given hour. 

Take a look at this video on how automation in Chinese ports have made port workers more productive, transforming the nature of their tasks from physically opening and closing gates to sitting behind a computer screen to direct port activity. 

The concern that automation might mean certain types of jobs will be lost is real. What this typically means is that some or many of the port workers will go into completely different types of roles in the same industry (e.g. the Chinese case where they are behind the computer screen), or they will enter into other expanding industries where their skill sets are in demand. 

It’s not uncommon for entire types of tasks or roles to be lost. What many neglect to understand is that technological advancements will either expand the same industries and/or create entirely new jobs that never existed before. For example, there are no more workers in the video tape industry today, but there are thousands in the streaming industry. Economists call this “creative destruction”--old jobs are destroyed, new jobs are created. 

While outdated, I’ve always liked this piece on how to understand job destruction and job creation from a historical perspective. Here is a job destruction list of some jobs that have been significantly reduced or completely eliminated since the start of the 20th century. 

And here is a job creation list showcasing the expansion and creation of new jobs since the 20th century (at the time of this report, the authors used jobs in 2002). Notice, for example, there were zero electricians, auto mechanics, and airplane pilots and mechanics. Today, there are hundreds of thousands of them. 

While some jobs were eliminated, more jobs were created and the economy grew overall. This is all part of a healthy, dynamic process in our economy. 

So yes, some port jobs in the future may be reduced, but that’s not bad for workers in the long-run. The economy expands with technological advancements, enabling the creation of new jobs and industries in the process. 

Today’s guest post is by Susannah Barnes-Petitt, a graduate student in economics at George Mason University, and the Program Manager for both Labor Policy and Fiscal Studies at the Mercatus Center. 

Subscribe now

The question of how and whether to support parents through child care subsidies or tax credits is emerging as a key policy issue. Both Presidential candidates have supported expanding the pandemic-era Child Tax Credit in hopes of reducing child poverty and making it easier for families to balance the demands of raising a child and working. While details on expanding the Credit vary between the candidates, both would like to make the American Rescue Plan’s expansion of the Child Tax Credit permanent as a starting point.

The expansion under the American Rescue plan gave most households with children up to $300 per child under 6 and $250 per child between 6 and 17. The first payment in July of 2021 was allocated to approximately 59 million children, while later payments were distributed to over 61 million children. The expansion lasted for five months, before ending in December 2021. Assessments of the program’s efficacy are mixed – proponents of the expansion argue that the expanded credit reduced food insufficiency and child poverty, while critics of the expansion argue that the program could create dependency on welfare for the lowest income groups by reducing the incentive to work. 

New research published in the Journal of Public Economics by Elizabeth Ananat, Benjamin Glasner, Christal Hamilton, Zachary Parolin and Clemente Pignatti looks at the latter claim and seeks to test the employment effects of the expansion under the American Rescue Plan. The authors did not find that the expanded credit had a statistically significant and robust impact on employment or hours worked. 

Regardless of household income, size of payments, or changes in work arrangements, the 2021 child tax credit expansion did not appear to affect employment or labor force participation. 

These results are a good thing for those concerned about reduced employment following the expansion, especially because the lowest income groups who gained the most from the expansion and whose labor supply is likely more elastic did not see significant changes in employment. However, as the authors point out, these results may be driven by the temporary nature of the expansion or as a result of the labor market conditions of the time. 

No Employment Effects is a Mixed Bag

It is worth noting that “no employment effects” can be both positive and negative. If you are concerned about disincentives to work, seeing no employment effects is a good thing. But if you are interested in welfare programs that can boost employment, seeing no effect is worrisome. 

Creating attachments to the labor force and having parents who work is one of the most effective ways to improve social mobility. A study by the Center for Poverty and Inequality Research at UC Davis found that employment for 30 weeks out of the year instead of 20 is associated with a 16 percent higher chance that someone will rise above the poverty line and lowers the chances of reentering poverty by one third. Our current welfare system is not designed to boost social mobility. While child poverty has declined over the past several decades, social mobility has stagnated. Many children are no better off than their parents. 

The most effective welfare programs would help families balance work and demands of raising children. This also means protecting the flexible and diverse forms of work, including non-traditional work arrangements like self-employment, freelancing, and gig work.

Behavior Differs for Temporary and Permanent Programs

The recent data seems to suggest that the pandemic-era Child Tax Credit was an effective tool to help families during times of need. 

But extrapolating the data from a temporary expansion during times of economic distress to present day may create poorly crafted policy. Behavior in the short-run and the long-run differs. When given cash transfers without an end date families may act differently than they did during a temporary expansion. The end date on the expansion may have served as a warning against drastic changes in behavior like limiting hours worked or quitting jobs. It may be a different story during a permanent expansion, especially for those receiving other forms of government assistance.

Indeed, some research does indicate that in countries where welfare programs for women and children are expansive (like Scandinavian countries), women work fewer hours and experience greater gender gaps in earnings. This result is especially pronounced for mothers, who experience “motherhood penalties” that reduce their participation in the labor force and their hours worked. 

One year ago today, I launched this Substack. I’m so grateful to all of you for your interest in my work and for your continued support. Given the emphasis in this presidential election on ‘pro-worker’ policies, I’ve revived my post from last year, 5 ways America can pursue a pro-worker agenda, for all my new subscribers.

Subscribe now

I was born in the Soviet Union, sometimes associated as a “worker power” paradise, where all aspects of work were regulated under the principle that every Soviet citizen must have a good, meaningful job that pays decent wages. Loosely, that meant a job was guaranteed by the government, there were rigid pay standards for each profession, it was impossible to get fired (which inevitably meant there was less job mobility), and the pursuit of income from self-employment was strictly prohibited. This latter prohibition on work is what sent my grandfather to jail for nearly a decade when he tried to earn more income by making and selling hats.

Three decades later, and several thousand miles west, I am witnessing America’s labor movement at the brink of a pivotal moment. Conservatives are now joining progressives in embracing the vital role of “pro-worker” policies (often contrasted with “pro-business” policies) for advancing the future of American society that, above all else, prioritizes worker power. Rhetoric from both the traditional left and the New Right are capitalizing on this moment. The Biden Administration’s Office of the U.S. Trade Representative just issued a proposal to investigate new regulations to pursue “worker-centered trade policy.”  The conservative Heritage Foundation recently reversed some of its labor policy positions to now “reclaim the role of each American worker as the protagonist in his or her own life.”

The key question now is what policies are “pro-worker”? How is this defined? Are pro-worker policies ones that ensure job security for every worker? Or are they the policies that try to maximize job opportunities for workers? Are pro-worker labor unions ones that have greater political power or ones that can best serve employees in their respective workplaces? Do pro-worker policies also focus on workers as consumers who can access goods and services and new technologies at lower prices—thereby increasing their standards of living?

At the heart of the challenge in pursuing “pro-worker” policies is that there are trade-offs in how we empower the American worker.

Greater Job Security or More Work Opportunities?

While you may think of Italy as the land of endless beautiful landscapes, food, and art, it is not a country known for its thriving labor market. Indeed, my husband and almost all his well-educated cohort from Bocconi University, the highest-ranked economics program in Italy, were forced to seek employment abroad upon graduation. Less than ten years ago, Italy’s youth unemployment rate was at a staggering 45 percent, and even now at 25 percent, Italy’s youth unemployment rate is one of the highest in the developed world.  

It is no coincidence that Italy is also known for having one of the most “pro-worker” labor policies among its Western counterparts—one that provides extensive job security and decent wages for current workers and makes it almost impossible for businesses to fire employees. But, as a result, employers are hesitant to hire workers, especially inexperienced youth, older workers, or where it’s unclear how the worker will perform on the job. There is little labor market mobility or turnover—so if you’re going to search for a job in Italy, you will have better luck finding charming cafes on the cliffside.

In fact, a set of studies found that when Italy even slightly reduced protective employment regulations, it led to increased employment and more jobs for younger workers. But Italy is not a stand-alone case. Through case studies around the world, economists have long documented how more-restrictive labor regulations, especially regarding job security, lead to less labor mobility, fewer job opportunities and higher unemployment rates.

In essence, countries with more rigid labor markets tend to have much higher unemployment rates and fewer job opportunities—a trade-off that a pro-worker policy agenda must grapple with. Yes, employment protection regulations can help with job security and job stability for existing workers who are already in those positions, but it comes at the expense of harming new workers, labor market dynamism, and reducing job creation.

Ensuring Good Jobs and High Wages: But for Whom?

Small business establishments make-up the backbone of American society. Creating a small business is often the source of income for underprivileged and minority workers. While there are some small businesses that make high profits, many small business owners, especially those with fewer than five employees and typical of your town’s “Main Street” establishment, can earn an average salary of $29,000 per year. That’s far below the national average mean wage.

What that means is when organizations like the American Compass or the Economic Policy Institute call for new policies to force businesses to increase wages, they are in practice calling for a transfer from low-income small business earners to low-income wage earners. 

Beyond the fact that some small business owners are also low-income earners, they are the worst low-income earners to make unemployed. Why? Because they are low-income workers that employ hundreds of thousands of other low-income workers. If they become unemployed because a regulatory change makes it too expensive for them to operate their business, all their employees become unemployed too.

Some may argue that if the small business cannot sustain decent wages for workers, then it shouldn’t exist at all. But that again illustrates a trade-off in “pro-worker” policy proposals since the small-business owner is a worker who employs many other workers and, like those other workers, aims to make a living and sustain their family.

For example, existing rhetoric on the minimum wage points to the “big corporations” that can afford the minimum wage hikes (a “pro-worker” policy), but is virtually silent on the impact it would have on small business owners, many of whom are also low-income earners. 

A recent study uses one of the most comprehensive tax datasets to examine the question of how the minimum wage impacts the owners of businesses who pay the minimum wage. The study, conducted in Israel, found that while minimum wage hikes reduced profits of companies (with minimum-wage intensive companies bearing the bulk of the cost and adjusting their workforces more aggressively), the profits declined more for lower-income business owners. Indeed, owners of businesses with higher shares of minimum-wage workers ranked at the bottom of the income distribution of business owners.

Again, this highlights the trade-offs in the current proposals for “pro-worker” policies from both the right and left. A rural household where the main income earner is a barber at his own shop and employs three others perhaps should not be the collateral damage of these “pro-worker” policies—he, too, is a worker we must consider in this dimension.

A Vision for a Pro-Worker Agenda 

When my family and I moved to rural America in the mid-1990s, we found it stood true to its symbol as the land of opportunity and abundance. Work was readily available, and my parents also found plenty of opportunities to pursue other sources of income through self-employment, contracting, or starting their own businesses. Within two decades, we experienced the income and intergenerational mobility that should be at the heart of every pro-worker policy agenda. My family was privileged to enter an economy with a lot of job turnover and open doors, largely because the United States was one the world’s most entrepreneurial, dynamic and flexible economies, which created a plethora of work opportunities for millions of people.

Generally, we would want a pro-worker agenda to maximize this promising aspect of American life—a flourishing and dynamic entrepreneurial environment that constantly creates better jobs, destroys worse ones, and allows for greater labor market mobility. At the same time, a pro-worker policy agenda should create better infrastructure to support workers and low-income earners without hindering the beneficial aspects of a thriving and dynamic labor market. What does this support look like?

1) Promoting a vibrant entrepreneurial environment

In a recent piece, I outline how we can promote an abundance of work opportunities through a vibrant entrepreneurial environment. Unfortunately, the rate of startup entry and the pace of business and employment dynamism in the U.S. has been falling over recent decades and patterns suggest that the incentive for entrepreneurs to start new firms has diminished over time. Countless studies have documented that too much red tape for businesses impedes economic activity, startup entry, entrepreneurship and job creation. This is especially true for early-stage entrepreneurs who are often resource- and cash-constrained. Indeed, one recent analysis of U.S. regulations across industries finds that greater industry regulations are associated with lower firm startup rates and fewer new hires, especially among smaller businesses.

Generating an abundance of work opportunities begins with creating a welcoming environment for existing businesses and for new entrepreneurs to build, scale and hire workers. For that process to unfold, we need to ease the regulatory shackles and revive our once vibrant and dynamic entrepreneurial environment.

2) Welcoming new and diverse forms of income-generating opportunities

Each one of us can provide a job for other members of our society when we hire a music teacher or a tutor, when we order customized T-shirts on Etsy, when we request a ride from the airport, or when we have an electrician inspect our homes. We should toss-out outdated notions that jobs can only be provided by established companies. 

There are more than 60 million workers earning income—either as a primary or secondary source—through these nontraditional means. The main beneficiaries of this type of work are those who place the greatest premium on flexibility—often these are women who are primary caregivers or workers with other personal circumstances that prevent them from working in traditional employment. While some aspects of this type of work are not ideal, prohibiting non-traditional work arrangements will surely reduce income opportunities for workers.

3) Redesigning labor unions to be an effective voice for all workers

In my piece for Newsweek on Labor Day, I outline how redesigning rules around organized labor can unlock the future of an innovative and worker-centric labor movement. The most important avenue is to reform the National Labor Relations Act (NLRA) of 1935 to promote a more diverse union landscape in which workers can choose the one that best serves them. Currently, the NLRA prohibits any formal cooperation between workers and the employer that is outside of the government-granted monopoly union, thereby banning other mechanisms to enhance a worker’s voice.

In an extensive and high-profile study entitled What Workers Want,economist Richard Freeman and law professor Joel Rogers concluded that our current system had not delivered to American workers the diverse set of institutions they sought in the workplace. Instead, it offered a single choice—collective bargaining through the winner-takes-all union—or no independent representation or participation in the workplace. To meet the wishes of the American workforce and enhance workers options, the NRLA must be reformed to encourage a more cooperative, competitive, and diverse environment for worker representation.

4) Reducing regulations and rigidities in the labor market

As discussed above, several types of labor market regulations are counterproductive for a pro-worker agenda. Reducing these rigidities would help lower long-term unemployment, increase labor market mobility, and expand job opportunities for workers. In a forthcoming piece, I will unpack these regulations and how they impact our labor markets.

5) Pursuing more constructive safety-net policies to address low wages and poverty while creating fewer labor market distortions

For several decades, countless economists and scholars have argued for safety-net policies that are free from the type of labor market distortions discussed in this piece. This means, for example, that instead of calling for increasing the minimum wage, there should be more policy solutions such as reforms to the earned-income tax credits to better address low wages and poverty. The key insight from these researchers is ensuring an effective safety-net that protects and empowers workers without constraining and impeding on our labor markets.

America does indeed need a new labor movement—but we need one that generates a thriving labor market with an abundance of work opportunities, while at the same time offering reasonable safety-net measures that do not impede the beneficial aspects of our labor markets and the future economy. Our path forward should rebuild resilience in our labor markets to technological, cultural, or global adaptations and one that benefits all types of workers, businesses, and the broader economy.