Computer science degrees grew fast—recent grad unemployment rose too

The CS boom—and a tougher entry-level market

Georgetown CEW The Major Payoff (2025) documents a supply surge and a hiring slowdown at the same time. Degree production in computers, statistics, and mathematics rose 159% from 2009 to 2023. Humanities and liberal arts fell 33%. In 2022, nearly one-third of recent bachelor’s holders majored in STEM; a similar share majored in business and communications.

Unemployment among recent graduates

Since May 2022, unemployment among recent college graduates consistently exceeded the rate for all workers. In March 2025, the average rate was 5.8% for recent graduates, 4.0% for all workers, and 2.7% for all college graduates ages 22–65 (Figure 12).

Among recent graduates pooled 2021–2023, unemployment was highest in humanities and the arts (7.7%) and social sciences (6.3%), and lowest in education and public service (3.2%) and healthcare (3.4%). Computers, statistics, and mathematics: 6.8%—up from 5.3% in 2014 for computer science recent grads specifically (7.2% in 2021–2023).

Earnings when employed still look strong

Recent graduates in computers, statistics, and mathematics reported about $79,000 median full-time earnings when employed—while education recent grads reported about $47,000 with 2.8% unemployment. High pay and high joblessness can coexist when many graduates are employed but more are searching than in calmer years.

CEW cites hiring caution in technology, finance, and media and automation pressure on some entry tasks—not a permanent verdict on CS careers.

Also important: National trends do not predict every local labor market or every employer’s hiring plan.

Humanities enrollment declines may reflect student risk aversion after seeing tech layoffs, not only earnings math—many humanities graduates still report strong satisfaction and mid-career mobility in non-STEM occupations. Bootcamp graduates entering software roles in 2021 may have different 2025 outcomes than 2024 CS bachelor’s graduates competing for the same junior titles.

Colleges adding CS capacity should pair expansion with career services, project portfolios, and co-op contracts—not only lecture seats.

STEM vs non-STEM outcomes · debt-to-earnings by major · Major Payoff section.

Unemployment by field among recent graduates

CEW Figure 13 lists unemployment for recent graduates pooled 2021–2023: humanities and the arts 7.7%, social sciences 6.3%, computers/statistics/mathematics 6.8%, education and public service 3.2%, healthcare 3.4%. Arts detailed majors reach 8.9% unemployment among recent grads while education detailed majors are near 2.8%.

Degree production vs hiring

Colleges expanded CS and related degrees faster than many entry-level tech hiring pipelines in some years. Students should seek internships, project portfolios, and regional labor market data—not only national STEM medians.

Prime-age STEM workers still earn far more than many non-STEM fields in CEW and BLS data. The tension is between long-run field pay and short-run job market timing for new graduates.

What colleges are doing about CS unemployment

CEW notes institutions expanding computer science offerings while entry-level tech hiring slowed in some markets. Career services and co-op programs partially offset weak entry-level hiring, but students should verify placement support before enrolling based on historical medians alone.

Prime-age STEM earnings remain strong in CEW and BLS data—students who persist through early-career volatility may still reach high mid-career medians. The policy question is how to reduce friction in the first five years after graduation, not whether technology fields pay over a lifetime.

Bootcamps, certificates, and CS degrees

Coding bootcamps promised fast entry into software jobs; outcomes vary by provider and hiring cycle. Traditional CS degrees still dominate leadership pipelines in large tech firms, but hiring managers increasingly accept portfolio-based evidence for some roles. CEW notes enrollment surges in computing while entry-level postings cooled in parts of 2023–2025.

Skills employers still reward

Algorithms, systems design, security, and data engineering remain scarce at senior levels even when junior generalist roles shrink. Students should align electives with measurable skill demand—cloud platforms, databases, statistics—not only core CS theory. Co-op programs that place students before senior year reduce unemployment risk after graduation.

Alternating hiring cycles and major choice

Technology hiring tightened for some entry-level software roles while demand remained strong for cybersecurity, data engineering, and health informatics. Students should read BLS occupational projections and local job boards, not only 2021–2023 unemployment pools in CEW tables.

Double majors (CS plus statistics, or CS plus domain expertise in biotech) often age better than generic CS degrees in crowded markets.

CS applicant playbook

Secure internships before senior year.

Build a public portfolio (repos, capstone, measurable projects).

Monitor local job boards, not only national unemployment pools.

How this article fits The College Payoff, 2026

This piece is one of twenty companion articles to The College Payoff, 2026. The parent report synthesizes College Board Education Pays, Georgetown CEW, Census, BLS, College Scorecard, Federal Reserve SCF wealth tables, and EDsmart institution-level findings. Each breakout answers one question journalists and families ask when those sources are quoted side by side.

When federal sources update on different calendars, start from the parent report’s definitions section, then return here for the topic-specific FAQ. Interactive tables and state pages on EDsmart Data—tuition trends, debt-to-earnings by major, attainment by state, STEM outcomes—let you filter national benchmarks beyond what a single government table provides.

If a number here differs from a chart in the parent report, use the parent report and the underlying federal publication as the authority. Breakouts explain how to read those sources; they are not a separate dataset.