Key takeaways:

• Misclassification of workers as independent contractors is a pervasive and widespread problem. AAPI workers are overrepresented in three of the 11 commonly misclassified occupations: manicurists and pedicurists, home health aides, and personal care aides. Vietnamese, Bangladeshi, Filipino, Samoan, and other Pacific Islander workers are overrepresented within these occupations.
• Groups with lower median hourly wages also have larger shares of their working populations in the 11 commonly misclassified occupations.
• Federal protections against misclassification are limited and currently under attack by the Trump administration. The state and local landscape for curbing misclassification is varied, which leaves some workers less protected than others.

In March, EPI published updated research highlighting the cost to workers of being misclassified as an independent contractor for 11 commonly misclassified occupations. Asian American and Pacific Islander (AAPI) workers were overrepresented in three of those occupations—manicurists and pedicurists, home health aides, and personal care aides—relative to their share of the overall workforce.

Most federal, state, and local labor laws apply only to employees and not to independent contractors, so misclassification strips workers of key protections such as minimum wage laws or qualifying for employer-provided health insurance and retirement benefits. Additionally, both misclassified workers and social insurance funds lose out on income: the report conservatively estimates that for the three jobs in which AAPI workers are overrepresented, misclassification costs workers at least $7,000 annually and costs social insurance programs $600 to $800 per worker each year.

With the understanding that the umbrella term “AAPI” encompasses an immensely diverse population both in ethnic origin but also in economic outcomes, this piece goes beyond the narrow view that all AAPI workers are high-wage earners. Below, we provide more detail on which groups of AAPI workers are most likely to be employed in lower-wage commonly misclassified occupations.

Disaggregated data shed light on particular AAPI communities that may be vulnerable to misclassification

Across all occupations, AAPI workers comprise approximately 8% of the total workforce. For three of the 11 occupations highlighted in the report—manicurists and pedicurists, home health aides, and personal care aides—AAPI workers make up 67%, 13%, and 10% of employment, respectively, according to Current Population Survey (CPS) data.

Table 1 provides a detailed breakdown of the composition of the AAPI workforce for the three occupations in which AAPI workers are overrepresented. Here, we use the American Community Survey (ACS) as it offers detailed race definitions which the CPS does not offer due to sample size restrictions.

Asian Indian and Chinese populations combined make up over 40% of the working-age AAPI population, thus their relatively large shares of the AAPI workforce in these occupations are not surprising. However, several groups are disproportionately represented across these occupations compared with their share of the overall AAPI workforce.

For example, Bangladeshi workers make up 5.1% of AAPI workers employed as home health aides while only constituting 1.1% of the total AAPI workforce. Chinese workers represent almost half (47.7%) of AAPI home health aides while representing just over one-fifth of the overall AAPI workforce (20.9%). AAPI employment among manicurists and pedicurists is largely held by those of Vietnamese origin (71.4%).

Finally, a majority of AAPI personal care aides are either Filipino (32.8%) or Chinese (20.8%). Filipino workers, however, are overrepresented by twice their share of the overall workforce. While Samoans and other Pacific Islanders comprised a much smaller share of personal care aide employment, they are also overrepresented in this occupation by more than twice their share of the overall workforce.

Table 1

Figure A provides a more comprehensive picture of the share of each detailed group employed across all 11 commonly misclassified occupations, revealing that smaller communities—often overlooked because of their size relative to the aggregate AAPI workforce—may be among the most vulnerable to misclassification. Workers belonging to seven of those groups are more likely than the average U.S. worker to be employed in one of those occupations. Almost 20% of Vietnamese workers are employed in one of those occupations, with over half concentrated as manicurists and pedicurists.

Samoan, Hawaiian, and other Pacific Islanders have the next highest shares working in the 11 occupations, making up 15% or more of their total working-age population. These groups also earn lower median hourly wages than the national median and the aggregate AAPI median hourly wage. Their disproportionate representation in commonly misclassified occupations further exposes these workers to wage suppression due to misclassification.

(function(){function e(){window.addEventListener(`message`,function(e){if(e.data[`datawrapper-height`]!==void 0){var t=document.querySelectorAll(`iframe`);for(var n in e.data[`datawrapper-height`])for(var r=0,i;i=t[r];r++)if(i.contentWindow===e.source){var a=e.data[`datawrapper-height`][n]+`px`;i.style.height=a}}})}e()})();

Misclassification enforcement varies by state—meaning different AAPI populations can be disproportionately impacted

Federal protections from misclassification are limited and are currently under attack by the Trump administration, which has proposed a rule to weaken standards to determine worker classification under the Fair Labor Standards Act, the Family and Medical Leave Act, and the Migrant and Seasonal Agricultural Protection Act. The proposed rule narrows the definition of who is a covered employee under these statutes, encouraging employer schemes to reclassify their employees as independent contractors to evade those obligations.

Broadly, the Trump administration has been actively dismantling long-standing federal worker protections, leaving states to bear the responsibility of ensuring workers are given rights and protections and that they can exercise them. For most states, labor and employment protections only apply to workers classified as employees, meaning workers misclassified as independent contractors are denied their legal rights and protections.

EPI’s 2026 misclassification report outlines state and federal policy recommendations that ensure proper enforcement mechanisms to curb misclassification. One of the recommendations includes implementing the ABC test. Unlike the six-part “economic reality” test or the “common law” test, the ABC test presumes that a worker is an employee unless they can demonstrate they are an independent contractor based on three criteria. Placing the onus on the employer to determine the employment status of a worker provides protections against misclassification and extends proper protections to workers. Many states have adopted the ABC test for unemployment insurance programs and, to a lesser extent, for wage and hour orders and other employment applications.

As shown in Figure B, The AAPI population is highly concentrated across a handful of states. Almost half of the prime-age working Asian population is concentrated in California, New York, and Texas, and a majority of the Pacific Islander population resides in California, Hawaii, and Washington. Overall, 21 states have significant numbers of AAPI residents, and some are home to large shares of specific AAPI communities. For example, the Hmong community in Minnesota and the Burmese community in Indiana are concentrated in states that have smaller total AAPI populations.

The current landscape for state policy protections against misclassification is quite varied. For example, among the states with the largest AAPI populations, California is the only state to adopt the ABC test for both unemployment insurance and employment law, although certain occupations are exempt from the test—including app-based drivers. California also institutes penalties for misclassifying a worker, which can include restitution payments and, if the misclassification was willful, a penalty between $5,000 to $25,000 per violation.

Texas, on the other hand, has significantly less state enforcement. Apart from using the common law test for its unemployment insurance program and providing a definition of an independent contractor for workers’ compensation, Texas mainly relies on federal law for classifying workers as employees. In the last 15 years, Texas lawmakers have introduced several bills that would create penalties for misclassifying workers in the construction industry, but all have stalled or failed.

Figure B Comprehensive protections are needed to protect workers from misclassification

AAPI workers are facing multi-pronged attacks from the Trump administration through the degradation of federal protections for workers, immigration, and equity. Occupational segregation and other labor market disparities lead women, people of color, and immigrants to be disproportionately represented in occupations that are commonly misclassified. These factors—in addition to historical and current geopolitical relations that shape the flow of labor to the U.S., immigration and citizenship status, and English language proficiency—can contribute to the concentration of AAPI workers in these occupations. Disaggregated data further identify which specific AAPI communities are overrepresented, revealing that smaller, less economically secure groups are often most exposed to the costs of misclassification. Strong policies at the federal, state, and local levels are needed to combat misclassification and to ensure workers can exercise their rights.

Key takeaways:

• The vast majority (85%) of young college graduates work in occupations that have seen strong employment growth in recent years.

• Young college graduates, like college graduates in general, are more likely to work in AI-exposed occupations than the overall workforce—and considerably more likely than young noncollege workers.
• But both young college graduates and young noncollege workers have experienced rising unemployment over the last three years, suggesting AI is not likely to be driving labor market weakness.

In the first blog post of our Class of 2026 series, we showed that the strong labor market for young college graduates of the early 2020s had begun softening in recent years. A growing share of young college graduates are seeking employment, but because their employment rates have not kept up with this job search, their unemployment rate has risen faster than the overall rate. The second blog post in the series discussed the industries where young college graduates worked. We found that recent graduates work in growing industries, but are forced to enter a weakened labor market with less job turnover, deteriorating their ability to break in. Young college graduates work in the tech sector at a similar rate to college graduates, and there is no clear evidence that tech sector employment is significantly decreased despite warnings about the advancement of AI.

In this blog post, we delve deeper into the occupations where young college graduates are likely to work.1 We examine whether it has been relatively more difficult to secure employment in these fields as the labor market has weakened. We also scour the data for signs that exposure to AI-related occupations may disproportionately affect the prospects for young college graduates as they enter the labor market.

Most young college graduates work in occupations with strong growth

Over 60% of young college graduates work in professional and related occupations or management, business, and financial occupations. Figure A displays the share of employment in each occupation or occupation grouping for young college graduates ages 22 to 27, all college graduates, and young workers without a four-year college degree. Occupations in the figure appear in order of the share of young college graduates employed in each, from largest to smallest. Over half (62.8%) of young college graduates work in professional, management, business, and financial occupations. Workers of any age with a college degree are slightly more likely to work in those two occupations (64.5%), though more likely in management occupations than professional occupations. On the other hand, nearly half (48.3%) of young noncollege workers are in service occupations or farming, construction, installation, and production occupations.

Figure A

Figure B shows the change in employment in each occupation between 2019 and 2026 and between 2023 and 2026, arranged in the same order as Figure A for comparison. Since 2019, management, business, and financial occupations and transportation and material moving occupations experienced the most growth, followed by professional and related occupations.

The top four occupations for job growth since 2023 account for 85% of young college graduate employment. The occupations with employment losses over the last three years were more likely to employ young noncollege workers than college graduates. It doesn’t appear that the occupations where young college graduates tend to work have been hit particularly hard in the last couple of years.

Figure B

While there has been job growth among occupations that tend to be filled by young college graduates, some worry about an increase in labor market underutilization, i.e., when workers with a college degree wind up working in jobs that typically don’t require one. Using O*NET data2, the New York Federal Reserve tracks this type of labor market underutilization. While the share of recent college graduates working at a job that doesn’t require a college degree has ticked up slightly over the last three years, it remains lower than it was for workers who graduated in the aftermath of the Great Recession. Even as late as 2017, young college graduates were working at these noncollege jobs at higher rates than they are today.

While college-educated workers are in more AI-exposed occupations, this does not appear to be driving labor market weakness

Much has been written in the last few years about AI exposure and its impact on the labor market. Using data from ADP, a large payroll processing company, Brynjolfsson, Chandar, and Chen find that entry-level workers in AI-exposed occupations—particularly AI uses that automate, not augment their work—have experienced an employment decline larger than that of older workers in the same occupations and all workers in less exposed occupations, explaining some of their stagnant overall employment growth. Atkinson and Yamco also find that declines in AI-exposed occupations are tied to lack of hiring rather than layoffs, hitting harder for young people attempting to enter the labor market. The second blog post in our series noted an across-the-board slowdown in hiring—which hurts the job prospects of all young workers, not only those in the industries most affected by AI.

On the other hand, researchers at the Yale Budget Lab argue that there has only been a slight increase in the shift in the occupation mix of employment, which would be evidence of AI automating jobs. They find that high AI-exposed occupations—determined by the top quintile of AI exposure—have yet to show declining employment, so no “dissimilarity” between young and older college graduates in terms of occupation mix has materialized. Raderman also finds that there isn’t strong evidence that AI is responsible for weaker labor market outcomes for recent college graduates, using evidence from Tillerman on college majors paired with change in unemployment.

Given the variation in assessments, we wanted to take a look at the data ourselves. Gimbel, Kendall, and Kulsakdinun have done an admirable job of summarizing the literature that attempts to classify AI exposure and propose a weighted aggregate measure of AI exposure.3 We employ this measure to investigate whether young college graduates may be more likely to be at risk in AI-exposed occupations than other workers.

In Figure C, we display the AI exposure of occupations weighted by the share of the entire workforce in each occupation. Moving from the left to the right on the figure increases AI intensity. For instance, professional and office & administrative support occupations are more AI exposed (to the right), while production, transportation, and service occupations are less AI exposed (to the left). Overall, the mean AI exposure score is 0.23.4

Figure C

In Figure D, we show the distribution of select demographic groups by occupation and AI exposure. As with earlier analysis, we compare young college graduates with all college graduates and young noncollege workers, in separate panels in the figure.

According to the aggregate measure, college graduates do have higher AI exposure in the labor market than the overall workforce. It is clear there is more mass in the direction of higher exposure (to the right) and their mean exposure is 1.07, higher than that of workers writ large. But the AI exposure of young college graduates isn’t any higher than that of college graduates in general. Mean AI exposure among young college graduates is 1.00.

Figure D

What is striking is that the AI exposure among young college graduates (1.00) is considerably higher than that of young noncollege workers (-0.61). If AI was driving labor market outcomes, we’d expect young college graduates to fare worse in today’s economy, e.g., see larger declines in employment or faster increases in unemployment. But, when we compare unemployment rates as we did in the first blog post of this series, both groups experienced similar increases in unemployment over the last two to three years. Trends in employment rates were also consistent across these groups.

Since the weakening labor market is hitting both young college and noncollege workers alike, it’s hard to argue that AI is uniquely causing job losses for new labor market entrants graduating from college now or in recent years. These findings are consistent with the literature, as there is currently no consensus about the effects of working in AI-exposed occupations on employment thus far.

1. Throughout this brief, we define young college graduates as people between the ages of 22 and 27 with only a four-year college degree. Unlike similar analyses of young workers, we do not exclude young college graduates who are currently enrolled in school, but the results here are robust either way. Unless otherwise noted, data for 2026 represent a 12-month average from April 2025 through March 2026 for the most up to date and reliable estimates, which removes seasonality and increases sample sizes.

2. O*NET or the Occupational Information Network provides the largest up-to-date database of information about workers sorted into detailed occupations. Information provided is about skills, abilities, education, training, and more.

3. We use an updated summary AI exposure PCA score (principal component analysis weighted standardized z-score) provided by the authors, May 13, 2026.

4. The PCA score scale is centered at 0, the unweighted mean across occupations.