The College ROI Divide: Why Your Major Matters More Than Your College

Stop asking whether college is worth it. Start asking which program is worth it. Our Scorecard field-of-study ranking shows typical pay one year after a bachelor’s can differ by about 3.95× —evaluate programs on your shortlist before you compare campus brands.

A college ROI divide has opened between majors: some bachelor’s pathways still produce strong returns while others struggle on typical pay, debt, completion, and labor demand. The split is about program economics, not whether you enrolled.

This issue opens The Future of Higher Education and Work—a four-part framework for evaluating credentials on typical terms. Later installments cover when credentials fail to pay, AI and the new premium, and pathways beyond a four-year default.

We ranked bachelor’s programs across four ROI drivers using May 2026 Scorecard program data merged with BLS occupation demand, Census earnings, and an O*NET AI exposure index—cuts readers cannot get from a generic college list. Typical pay means the national median (half above, half below), not a guarantee about your offer.

Preview: Our analysis of College Scorecard data shows 1-year program medians can differ by about 3.95× between fields in our pull—before debt, completion, or demand enter the frame.

Pair this framework with EDsmart’s guide on whether college is worth it and our analysis of degrees with weak ROI and high AI exposure.

The four drivers of college ROI

Before you rank campuses, run the same four questions on every program on your shortlist. Every section below maps to one driver—not a separate topic.

Earnings is typical pay after you finish—not sticker-price optimism. Debt is what you borrow relative to that pay. Completion is whether students actually earn the credential. Demand is whether employers and licensure boards still need the skill.

Every section below maps to one driver. Use the matrix as the lens; the tables that follow supply the numbers.

If you only read three things

• Major choice can move typical pay by about 3.95× (~4×). Computer Science and Fine Arts are not interchangeable because both say “bachelor’s degree.”
• About 40.4 million adults have some college and no credential. Completion is often the hidden variable between weak and strong outcomes.
• Debt stress is a ratio, not a headline number. Fine Arts program medians run about 1.6× debt to one year’s typical pay; Computer Science families near 0.4× in our Scorecard build.
• Program beats brand. A selective campus can still host weak programs; a less selective one can host strong ones—verify the field, not only the name.

Start here: Degrees tied to strong labor demand and lighter debt are gaining ground on ROI. Degrees with weak early pay, heavy loans, or low completion are losing ground, even when the campus brand sounds prestigious. The split is about program economics, not whether college “works” on average.

How to use this article

What it’s for: Showing why higher education outcomes have become uneven—and how to evaluate a specific program before you commit.

• Four drivers → The framework behind every section
• Driver 1: Earnings → Typical pay by field
• Driver 2: Debt → When loans outrun the first paycheck
• Driver 3: Completion → The credential you never finished
• Driver 4: Demand → Labor-market shifts (including AI context)
• Checklist → Questions to ask about your program
• Charts & tables → debt-to-earnings by major, AI exposure by major, college payoff (2026)

Why “is college worth it?” is the wrong question

For decades, “go to college” was treated as universal advice. National averages still show a bachelor’s premium—but they blend every major together.

Cross-checking College Board Education Pays 2026 (Census CPS earnings), typical full-time pay for bachelor’s workers runs near $81,800 vs. $50,600 for high school graduates—a 61.7% premium. Our analysis of College Scorecard data shows much wider spreads when you compare fields directly. The surprise is not that some majors earn more—it is how much more, and how often debt and non-completion erase the gain.

Driver 1: Earnings—where the pay ladder breaks apart

Why this matters: If typical early pay is low, every other part of ROI starts behind—even before loans and completion enter the picture.

Our analysis of College Scorecard data shows 1-year program medians reaching about $173,344 at top reporting institutions for Computer Science vs. $43,834 for Computational Science—roughly a 3.95× gap (about 4× in headlines). Typical program medians run about $71,458 (computer science) and $43,834 (computational science); Scorecard does not publish national 1-year medians for either field. Computational Science (CIP 30.30) is a separate field from Computer Science (11.07); the spread compares program-level figures, not two paths inside one major family.

Degrees gaining ground (typical early pay and demand)

• Computing and engineering fields lead our Scorecard early-career ladder (see table).
• Registered nursing (CIP 51.38) lands in the top 50 with 1-year program medians above $100,000 at top reporting institutions.
• STEM occupations still pay more than non-STEM jobs in BLS data, even when individual majors vary.

Our analysis of BLS May 2024 OEWS data puts typical pay near $103,580 in STEM occupations vs. $48,000 in non-STEM occupations. Individual majors still vary inside those buckets.

Computer science: valuable on the ladder, tougher at the door

Computer Science leads 1-year program medians in our Scorecard pull ($173,344 at top reporting institutions; typical program median about $71,458). That tracks program-level cohorts, not every graduate’s first offer.

At the same time, Georgetown CEW has reported rising unemployment among recent computer science graduates as enrollment grew. Hiring cycles and automation of entry-level tasks can make the first job harder even when typical program medians stay high. PSEO national data still show typical pay rising from year 1 to year 5 for CS graduates (about 46% in our pull). Think of CS as high upside with more competition at entry, not a single “easy” path.

Driver 2: Debt—when loans outrun the first paycheck

Why this matters: A degree is not automatically valuable if debt overwhelms early-career income. ROI often loses ground when loans eat the first paycheck—not only when pay is low.

Low early pay is not the same as a “bad major.” Some fields face wage compression, uneven program quality, graduate-school dependency, or weak local demand. The debt ratio makes the stress visible.

Degrees under pressure (weak early pay or heavy debt)

• Education and liberal arts families show typical debt at 63% and 76% of one year’s typical pay in our Scorecard build.
• Fine Arts shows the heaviest stress in our family table: debt near 1.6× one year’s typical pay.
• Cross-checking Education Pays 2026, performing-arts worker medians run near $44,000 and computer science near $87,000 before loans.

Debt is a ratio, not a balance. A loan that is manageable at one year’s typical pay can fail when early earnings are weak—compare debt ÷ pay, not debt alone.

For an interactive cut by major, see debt-to-earnings by major on EDsmartData.com.

In our May 2026 Scorecard pull, typical pay for dentistry and taxation master’s programs differs by nearly 3×. Many licensed careers still require graduate school even when bachelor’s medians look modest.

External researchers (including FREOPP-style program ROI work) have estimated that about 23% of bachelor’s programs could show weak returns once completion and forgone wages are counted, while well under 1% of listed institutions show negative 10-year ROI in our campus-level Scorecard build on EDsmartData.com. Headlines that mix program and campus definitions can talk past each other.

Driver 3: Completion—the credential you never finished

Why this matters: Completion turns out to be the hidden variable. Starting college without finishing often lands closer to “some college, no credential” outcomes than to a completed bachelor’s.

“Some college, no credential” is its own category in government statistics. It is not halfway between high school and a finished degree.

National Student Clearinghouse figures cited on our college dropout rates page put the group at about 40.4 million adults. BLS Employment Projections (2021) show typical weekly pay near $935 for some college with no degree vs. $1,432 for bachelor’s or higher.

Cross-checking Education Pays 2026, typical worker medians run near $50,600 for workers with a high school diploma, $61,300 for associate holders, and $81,800 for bachelor’s holders. Our Census ACS earnings pull shows similar spreads for adults 25+: about $40,153 for high school graduates, $47,260 for some college or associate credentials, and $70,044 for bachelor’s holders.

Completion multiplies value. Starting without finishing often lands closer to “some college, no credential” outcomes than to a completed bachelor’s.

Among bachelor’s-granting campuses in our Scorecard build, six-year completion rates and typical pay 10 years after entry move together:

Typical pay differs by about $34,077 between the lowest and highest completion bands—$44,130 where six-year completion is below 40% vs. $78,207 where completion is 80% or higher. That gap can exceed the typical associate–bachelor’s spread in Education Pays.

Driver 4: Demand—what the labor market rewards now

Why this matters: Earnings, debt, and completion still sit inside a job market that rewards some skills and licenses more than others.

Our analysis of BLS Employment Projections 2024–34 shows how far apart demand can be for occupations tied to common major families. Openings include growth and replacements for workers who retire or change fields.

Software developers average about 115,200 openings per year with +15.8% projected employment growth—much faster than the all-occupation average. Registered nurses add about 189,100 openings annually despite slower percentage growth because the occupation is enormous. Elementary teachers and fine artists show negative projected employment change with far fewer openings—thin demand at the national level even when individual graduates succeed.

STEM occupations still pay more than non-STEM jobs in BLS wage data: typical pay near $103,580 vs. $48,000. That is a level signal; the openings table shows whether new jobs are expected.

• Graduate earnings climb faster in some fields than others. Census LEHD PSEO shows typical pay from year 1 to year 5 rising about 46% for Computer Science vs. about 57% for visual and performing arts (national medians, different cohorts).
• Licensed paths still need credentials. Our analysis of College Scorecard data lists typical pay about $142,745 four years after completion for Law vs. $88,472 for Insurance (four-year program medians).
• Task change and AI. Tool use on campus rose quickly; pay still rewards scarce technical and regulated skills. Issue 3 goes deeper on AI, skills, and the credential premium.

We mapped O*NET technology skills to major families—an original EDsmart exposure index, not a job-loss forecast:

High exposure does not mean a major is worthless—computer science ranks high on both exposure and pay in our Scorecard pull. See the AI exposure by major chart for the full ranking.

How we built this view

We merged College Scorecard program and campus fields (May 2026 build), BLS occupational wages (May 2024), Education Pays worker medians, Census ACS and LEHD graduate outcomes, O*NET technology skills, and Clearinghouse completion counts. Derived rankings, debt-to-earnings cuts, and automation exposure scores are original EDsmart analysis on EDsmartData.com, with cohort rules on the College Scorecard ROI methodology page.

Across 3,886 campuses with both debt and earnings fields, median loan debt is about 0.34× one year of typical pay 10 years after entry; about 16.9% of those campuses show debt above half of one year’s earnings.

Program vs. campus: Compare program medians when you are choosing a major; use campus ROI when you are comparing net price and completion on a shortlist. A prestigious college can still list programs with weak typical pay; a less selective campus can host strong ones.

Paths that are not a four-year degree

Certificates, skilled trades, community college, and apprenticeships can beat selected bachelor’s fields on typical pay with less debt. They belong in the same decision frame: earnings, debt, completion, and demand—not a separate debate about whether “college” is worth it.

Trade ROI pages for electricians, HVAC, and nursing are on EDsmartData.com. College Scorecard also lists non-degree credentials, though a systematic national certificate ranking is still in progress.

Program evaluation checklist

Replace “Is this a good school?” with the four drivers applied to your program—whether the path is two years, four years, or a certificate.

1. Earnings: What is typical pay one and four years after finishing this program in our Scorecard pull?
2. Debt: What is typical loan debt among completers compared with one year of typical pay?
3. Completion: What is the completion rate for students with a background like yours?
4. Demand: Where do graduates work, and how exposed are those tasks to automation in our O*NET index? Use exposure as context, not a verdict.
5. Campus vs. program: Does this specific major at this campus beat alternatives on your shortlist—or is prestige doing the selling?

If a program fails two or more drivers against the alternatives on your list, treat that as a signal—not a moral verdict on the field.

At the campus level, our Scorecard ROI build on EDsmartData.com shows a typical 10-year surplus near $143,847 overall, with about 0.8% of listed institutions negative after average net price. Typical surplus near $169,141 at public colleges vs. $107,804 at for-profits reminds you that sector still matters—after you have evaluated programs.

What to remember

The real message: Stop asking whether college is worth it. Start asking which program is worth it—for your price, your completion odds, and the jobs that field feeds. The series continues with why credentials fail, how AI reframes demand, and pathways beyond a bachelor’s default.

1. Earnings: Major choice can move 1-year program medians by about 3.95× (~4×) in our Scorecard field-of-study pull.
2. Debt: Loan stress rises when debt is a large share of one year’s typical graduate pay (Fine Arts near 1.6× in our family table; CS families near 0.4×).
3. Completion: About 40.4 million adults have some college and no credential; high-completion campuses cluster at much higher typical pay 10 years after entry than low-completion peers.
4. Demand: STEM and licensed fields still lead typical pay ladders, but entry-level competition and task change (including AI) can widen spreads over time.
5. Program > brand: Verify the same field on every campus you are considering—institution prestige does not guarantee program ROI.

What we still cannot measure cleanly

Public data are strong on typical pay, debt, and completion, but several headline questions still lack a single national series. Until those land, treat gaps as unknowns, not zeros.

• Scorecard 1-year figures use national medians when published; otherwise the highest institutional program median in our pull (computer science has no published national 1-year median—typical program median about $71,458).
• We do not yet have a national underemployment rate by major (e.g., NY Fed SCE). This report uses Scorecard “not working 1 year” as a program-level proxy only.
• ACS table B20004 groups some college and associate credentials; SCNC-specific earnings still come best from BLS/CPS tabulations.
• We have not yet pulled every certificate and apprenticeship program from Scorecard in a systematic way.
• We do not yet publish a multi-year Scorecard-by-CIP trend file in this build; the “what’s changing” section uses cross-section levels plus graduate-earnings trajectories (PSEO) and external labor-market reports.

Sources

We merged College Scorecard program and campus fields (May 2026), BLS wages and employment projections, Education Pays worker medians, Census ACS and PSEO outcomes, O*NET technology skills, and Clearinghouse completion counts. Derived rankings and exposure scores are original EDsmart analysis.

Interactive tables, charts, and cohort notes are on EDsmartData.com. Definitions: College Scorecard ROI methodology.

Pay-spread note: The ~4× headline uses Scorecard 1-year program medians in our pull ($173,344 at top reporting institutions for Computer Science vs. $43,834 for Computational Science; typical program medians about $71,458 and $43,834) = 3.95× before rounding.

1. EDsmartData.com: institution- and program-level ROI builds, debt-to-earnings by major, AI exposure by major, and related pages (May 2026 build).
2. U.S. Department of Education College Scorecard (program and campus fields).
3. Bureau of Labor Statistics Occupational Employment and Wage Statistics (May 2024 national file).
4. College Board Education Pays 2026 and U.S. Census Bureau Current Population Survey.
5. Federal Reserve Survey of Consumer Finances (2022 wave).
6. NCES and National Student Clearinghouse (completion and some-college-no-credential counts).
7. U.S. Census Bureau LEHD Post-Secondary Employment Outcomes.
8. Bureau of Labor Statistics Employment Projections program (2024–34 national occupation tables).
9. O*NET 28.0 Technology Skills (automation exposure index).