How Shifts in Labor Supply and Demand Shape Outcomes for Young Workers

This is the second blog post in a three-part series that explores labor market challenges and opportunities for young adults.

In our previous blog post, we showed that since April 2023, when the U.S. labor market was at its strongest, young adults have experienced a noticeable erosion in employment opportunities. The pattern of this erosion is consistent with a “low-hire, low-fire” economy in which firms hold on to existing workers but scale back hiring. The question that naturally follows is: What forces in the economy explain the data on declining opportunities for young workers, especially recent college graduates?

To answer it, we focused on a set of labor supply and labor demand factors that have featured in public discussion as potential explanations for changing labor market conditions. Some emphasize broad cyclical forces, such as declining job openings, while others point to shifts in hiring practices, industry composition, public sector employment or workforce availability. Rather than presuming any of these explanations to be decisive, we examined them within a single empirical framework to assess their relative association with changes in young adults’ labor market outcomes.

The factors we examined are these:

• A decline in overall labor demand
• Rising demand for artificial intelligence (AI) jobs
• Contraction in manufacturing employment
• A decline in the foreign-born population
• Changes in women’s federal employment

Using individual-level data from the Current Population Survey (CPS),The CPS is sponsored jointly by the U.S. Census Bureau and U.S. Bureau of Labor Statistics. we related employment, unemployment and labor force participation outcomes both to worker characteristics and measures of labor demand that vary by state, month and year. General labor demand was captured using job openings data from the Job Openings and Labor Turnover Survey (JOLTS),JOLTS is conducted by the U.S. Bureau of Labor Statistics. while demand for AI work came from detailed job-posting data.

The Business Cycle Still Does Most of the Work

The figure below summarizes the relative contribution of each factor to changes in the employment-to-population ratio, unemployment rate and labor force participation rate for workers ages 18 to 24 and for older workers. The key observation is clear: The decline in overall labor demand, as measured by job openings, accounts for most of the deterioration in young workers’ outcomes.

SOURCES: CPS, JOLTS and authors’ calculations.

NOTES: From April 2023 to December 2025, the actual changes in 18- to 24-year-olds’ employment-to-population ratio, unemployment rate and labor force participation rate were -2.24 percentage points, 3.51 percentage points and -0.35 percentage points, respectively. The actual changes in 25- to 64-year-olds’ employment-to-population ratio, unemployment rate and labor force participation rate were 0.15 percentage points, 0.63 percentage points and 0.68 percentage points, respectively.

Nationally, the employment-to-population ratio among 18- to 24-year-olds fell by more than 2 percentage points between April 2023 and December 2025. Most of this decline reflected weaker overall labor demand as hiring slowed and job openings became scarcer. Importantly, the deterioration appeared primarily as higher unemployment rather than as labor force exits, indicating that young workers were still searching for jobs but with fewer opportunities available.

By contrast, once overall labor demand is accounted for, there was no comparable deterioration for workers ages 25 to 64. Their employment outcomes remained largely stable. This reinforces a central theme of our work: Hiring slowdowns show up first and most clearly among young and inexperienced workers.

Where AI Enters the Story

Although the business cycle dominates, AI is not irrelevant. The figure shows that rising demand for AI jobs independently explains a meaningful share of the recent erosion in young workers’ outcomes. This demand growth can increase unemployment for young adults because it raises skill requirements for entry-level jobs, reducing access to employment for workers who rely on initial hiring to enter the labor market.

To measure AI job demand, we used job-posting data from Lightcast, which includes the full text of job descriptions by state, month and year. A posting was classified as an AI job if it required skills from one of 10 AI skill clusters,For more details on AI-related skills, see the 2026 Artificial Intelligence Index Report (PDF), an initiative of Stanford University’s Institute for Human-Centered Artificial Intelligence. including areas such as generative AI, machine learning and neural networks. This approach captures how job content (the skills and tasks employers list in job postings) is changing, not whether firms are using AI to automate jobs or to screen applicants.

Nationally, for workers ages 18 to 24, increased demand for AI jobs accounted for roughly:

• Forty-five percent of the decline in their employment-to-population ratio
• One-third of the increase in their unemployment rate

In Eighth Federal Reserve District states,Headquartered in St. Louis, the Eighth District covers all of Arkansas, most of Missouri, and parts of Illinois, Indiana, Kentucky, Mississippi and Tennessee. AI job demand accounted for nearly half of the increase in young workers’ unemployment rate. However, their overall employment-to-population ratio rose because of offsetting changes in labor force participation.

AI Effects Are Concentrated at Labor Market Entry

The following table focuses on three groups of labor market new entrants: (1) individuals with no more than a high school diploma, (2) Black new entrants with no more than a high school diploma, and (3) recent college graduates. These groups allow us to compare outcomes for young workers entering the labor market with different levels of education and exposure to labor market risk. Because Black young adults historically have exhibited greater sensitivity to changes in labor market conditions, including this group illustrates how widely employment outcomes vary among new entrants.

Three clear patterns emerged during this period. First, falling overall job openings remained the dominant force behind worsening outcomes for every new-entrant group shown. For example, looking specifically at unemployment among recent college graduates, declining job openings explained more of the increase than any of the other four factors.

Second, demand for AI jobs, at the technology’s current pace of adoption and diffusion, appeared to be displacing some new entrants, particularly college graduates. Although recent graduates still had relatively high employment-to-population ratios—79% nationally and 82% in Eighth District states in the first quarter of 2026—their lack of experience and concentration in entry-level roles place them at risk as job requirements shift toward AI tasks.

Third, other factors—changes in the foreign-born population, contraction of the manufacturing sector and shifts in women’s federal employment—often explained little to none of the recent erosion in young workers’ outcomes. These factors either moved too late in the period or affected groups not concentrated at the point of labor market entry.

What Have We Learned?

The recent employment experiences of young workers fit squarely within a low-hire, low-fire labor market. Since April 2023, hiring has slowed, and young workers, especially new entrants, have borne the brunt of that softening. AI adds an additional headwind at the point of labor market entry, particularly for recent college graduates, but its effects remain smaller than those of the broader decline in job openings.

Even at this stage of adoption and diffusion, AI matters, but in a narrow, early and age-specific way. AI is not eliminating jobs across the economy. Instead, it is raising the bar for young workers trying to secure their first foothold in the labor market, an area in which employers feel less of a need to advertise openings, make offers and hire.

Our next blog post will examine how “disconnected” young adults—those not in school or working—are spending their time.

Notes

1. The CPS is sponsored jointly by the U.S. Census Bureau and U.S. Bureau of Labor Statistics.
2. JOLTS is conducted by the U.S. Bureau of Labor Statistics.
3. For more details on AI-related skills, see the 2026 Artificial Intelligence Index Report (PDF), an initiative of Stanford University’s Institute for Human-Centered Artificial Intelligence.
4. Headquartered in St. Louis, the Eighth District covers all of Arkansas, most of Missouri, and parts of Illinois, Indiana, Kentucky, Mississippi and Tennessee.