The STEM-versus-liberal-arts debate is one of the most common framings in choosing what to study, and it rests on a flawed premise: that the two categories are the right unit of comparison. They are not. Earnings vary roughly threefold within each category, which means the spread inside STEM and inside liberal arts is larger than the average gap between them. Comparing the categories produces a tidy headline and a misleading conclusion. The useful comparison happens at the level of specific programs and careers. This guide replaces the category argument with the data, as an analytical aid within How to Choose a Major.
The framing persists because it is easy to repeat and hard to check. "STEM pays, liberal arts doesn't" fits in a sentence, lines up with a familiar story about practicality, and matches a few real data points at the extremes. What it cannot survive is contact with the full distribution. Once you look at the actual range of outcomes inside each category, the clean line between them dissolves into two heavily overlapping clouds. The goal here is not to defend liberal arts or to talk anyone out of STEM. It is to move the decision down to the level where the numbers actually mean something, so that the choice is made on a real forecast rather than on a slogan.
The Category Is the Wrong Unit
The debate assumes STEM and liberal arts are each coherent enough to carry a single earnings verdict. Neither is.
STEM is a mixture
It spans high-paying engineering and computer science alongside lower-paying laboratory and field-science roles. The "STEM pays well" headline is carried by a few fields; others sit far below the category average. STEM is not one earnings story.
Liberal arts is a mixture
It spans modest-paying paths alongside economics and tracks that feed law, business, and management, which pay well. The "liberal arts pays poorly" headline ignores its high-earning end. Liberal arts is not one earnings story either.
Because each category is a mixture, the average for each is a blend that describes no actual graduate. A student does not major in "STEM" or "liberal arts"; they major in a specific program that leads to specific careers, and that program sits somewhere in a wide range. Arguing the categories is arguing about averages that no individual experiences.
It is worth being precise about what "liberal arts" even means here, because the word does double duty and that ambiguity fuels the debate. Sometimes it names a set of fields: the humanities and social sciences, languages, history, philosophy, the arts. Sometimes it names a kind of school: a small residential college that teaches across many fields and admits to the college rather than to a program. The earnings argument almost always means the first sense, the fields, but it borrows emotional weight from the second, the schools. The site organizes study by field, mapping all of higher education into federally defined majors, which is the only version of "liberal arts" that has earnings data attached to it. Keeping the two senses separate is the first step toward a comparison that holds together, because a graduate's outcome is driven by the field they studied and the career it leads to, not by whether the campus called itself a liberal arts college.
The same precision applies to STEM. Science, technology, engineering, and mathematics is an acronym of convenience, not a single labor market. Engineering and computer science behave like one kind of field in the wage data. Many of the pure sciences behave like another, closer in early earnings to the social sciences than to engineering. Folding them all under one banner is what lets the high-paying corner pull the whole category's reputation upward, and it is exactly the move that breaks down the moment you read the majors archive field by field.
The Ranges Overlap More Than the Headlines Admit
The decisive fact the category debate hides is the overlap between the two distributions.
Earnings within both STEM and liberal arts vary by roughly a factor of three from the lower to the upper percentiles. When two ranges that wide sit next to each other, they overlap substantially: the high end of the liberal arts range reaches well into the STEM range, and the low end of the STEM range drops well into the liberal arts range. A high-earning liberal arts path out-earns a low-earning STEM path, routinely. The category averages can differ while the individual outcomes overlap heavily, which is exactly the situation Reading Earnings Data Honestly warns about: the median hides the range, and here the range is the whole story.
A useful way to picture it: imagine the two categories not as two points on a number line but as two overlapping bell curves drawn on top of each other. The STEM curve is shifted to the right, so its peak sits higher. But both curves are wide and flat, not tall and narrow, and the right tail of the liberal arts curve sits squarely under the left half of the STEM curve. The gap between the two peaks is what the headline reports. The overlap between the two bodies is what the headline omits, and the overlap is far larger than the gap. When someone says STEM out-earns liberal arts, they are describing the distance between the peaks. When you choose a major, you are placing one bet somewhere inside one curve, and where you land inside it matters more than which curve you chose.
This is why the "average" is such a treacherous number in this debate. An average is the right tool when a group is uniform and the wrong tool when it is not. Both categories are about as far from uniform as a group can get. The honest comparison is not average to average but range to range, and the moment you lay the ranges side by side, the question stops being "which category pays more" and becomes "where in each range does my specific program sit." That second question has an answer. The first one never did.
The Early-Versus-Mid-Career Pattern
The headlines also lean on a snapshot, and the snapshot favors STEM in a way that overstates the lifetime difference.
STEM tends to lead in starting salary, which is the number most comparisons use because it is the easiest to measure. But the trajectory matters as much as the starting point, the same lesson as Job Growth Projections. Several liberal arts paths, particularly those that feed management, law, and business, narrow or close the early gap by mid-career as graduates move into higher-paying senior roles. A field that starts lower but climbs steeply can match or pass a field that starts higher and plateaus. The early-career snapshot, by definition, cannot see this, so it systematically overstates STEM's lifetime advantage for the paths that climb.
There is a structural reason some liberal arts paths climb. Many of them are pre-professional in disguise: they are the undergraduate on-ramp to a graduate or professional credential that does the real earnings work later. A political science or history major who goes to law school, an economics major who moves into finance or consulting, a psychology major who becomes a licensed clinician, all carry an undergraduate earnings figure in their early twenties that says almost nothing about where they end up. The category data captures them at the bottom of a staircase they are about to climb. A site that reports earnings at a single point in time, as nearly all of them do, will always undercount the fields whose payoff is back-loaded behind further study.
The reverse caution applies to STEM. Some technical fields pay well at the start precisely because they are front-loaded: the degree is the credential, the starting salary is high, and the curve is flatter from there. That is not a criticism. A strong, steady salary from year one is a real advantage, especially against debt. But it does mean a flat-but-high STEM path and a low-but-steep liberal arts path can cross somewhere in the middle of a career, and which one "wins" depends entirely on the horizon you measure over. Compare them at year one and STEM wins clearly. Compare them at year twenty and the answer is contested. The honest version of the comparison names the time horizon out loud, because the same two paths can swap places depending on it. The ROI Calculator lets you stretch the horizon out rather than freeze it at graduation, which is where the back-loaded paths finally show their value.
A Worked Example: Two Real Fields, Not Two Categories
The argument is easy to nod along with and easy to lose the moment a real choice appears. So put two actual fields on the table, one from each camp, and walk them through the comparison the right way.
Take Computer Science as the STEM contender and economics as the liberal arts contender. The reflex says CS wins, full stop. The data says: it depends entirely on which CS path and which economics path, at which net cost, over what horizon.
Start at the top, with the major-level medians. Computer science shows a higher midpoint. If that number were the whole story, the comparison would be over. But the midpoint is the center of a wide band, and the band is the part you actually live in. Drop one level and computer science splits into a theory-leaning track, a systems and IT track, and a software-engineering track, and those tracks pull apart in both earnings and completion. Economics splits too, between a quantitative, finance-feeding track and a more general policy-and-analysis track, and its top end reaches into territory the headline never credits to "liberal arts."
Now lay the ranges side by side instead of the medians. The high end of the economics range overlaps the middle of the computer science range. A graduate of a strong, quantitatively rigorous economics program who enters finance or data analysis can out-earn a graduate of a general computer science program who enters mid-tier IT support. The fields traded places, and nothing about the "STEM versus liberal arts" frame predicted it. What predicted it was the specific program and the specific career each graduate entered, visible on the careers data the two paths feed into.
Finally, add cost and horizon. Suppose the economics option comes at a low in-state net price and the computer science option at a high out-of-state one. The ROI gap narrows further or reverses, because the return is earnings over cost, not earnings alone. Stretch the horizon to mid-career and the economics path's finance and management destinations may keep climbing while a flatter technical path levels off. None of this makes economics the better choice in general. It makes the general question unanswerable and the specific one answerable: this CS program, at this college, at this net price, against this economics program, at that college, at that net price. That comparison has a winner. "STEM versus liberal arts" never did.
Compare Programs, Not Categories
The replacement for the category debate is a program-level comparison, which is the only level where the numbers are meaningful.
To compare a STEM option against a liberal arts option honestly:
| Step | What to do |
|---|---|
| 1. Drop to the program | Compare the specific programs, not "STEM" vs "liberal arts" |
| 2. Read the ranges | Use the 25th-to-75th-percentile bands, not the medians alone |
| 3. Factor cost | Weigh each degree's net cost; a cheaper degree lifts ROI regardless of category |
| 4. Add growth and interest | Include job growth and your own interest, which drive trajectory and completion |
The ROI Calculator does the cost-versus-earnings math at the program level, and the Career Path Explorer shows where each program leads. A specific computer science program and a specific economics program can be compared meaningfully; "STEM" and "liberal arts" cannot.
The reason cost belongs in the comparison and not as an afterthought is that ROI is a ratio, not a salary. A degree's return is what it earns divided by what it cost to get, and the cost side moves as much as the earnings side, often more. A liberal arts degree taken at a low net price, with aid or in-state tuition, can post a stronger return than a higher-earning STEM degree taken at full sticker price with borrowing. The earnings headline only ever looks at the numerator. The whole point of computing ROI at the program level is to put the denominator back in, because two students who study the identical field at different net costs do not get the identical return. Net price varies enormously college to college for the same program, which is why the colleges archive and the per-college net-price figures matter as much to this decision as the field does.
The Mistakes This Debate Causes
Treating the category as the unit of comparison is not a harmless simplification. It produces specific, costly decisions, and the same few recur.
The first is picking the category and stopping there. A student decides "STEM" is the safe answer, enrolls in whichever STEM program admission and aptitude allow, and assumes the category's reputation will carry them. But the category average is earned by its top fields, and a student who lands in a lower-paying corner of STEM has bought the label without the outcome. The fix is to refuse to commit at the category level. Drop to the specific program on the majors archive and read its earnings range before treating it as the safe choice.
The second is reading the median as a personal promise. Both categories are so internally varied that the median describes almost no one. A student who chooses on the midpoint has quietly assumed they will land in the exact middle of a band that may be three times as wide as the median suggests. The fix is to read the 25th-to-75th-percentile range, not the median alone, and to ask honestly where in that range a realistic version of yourself lands, a habit Reading Earnings Data Honestly builds out in full.
The third is ignoring cost on the way in. A student compares two fields on earnings, picks the higher one, and never notices it costs far more to obtain at the colleges they are actually considering. ROI is earnings over cost, and the higher-earning field can post the weaker return once the price is in. The fix is to run both options through the ROI Calculator at their real net prices, not their sticker prices, before declaring a winner.
The fourth is freezing the comparison at graduation. The early-career snapshot is the easiest number to find and the most misleading one to decide on, because it cannot see the back-loaded paths that climb later. The fix is to name a horizon and compare over it, so a path that starts lower and rises is not penalized for a starting line nobody stays at.
Every one of these mistakes is the same mistake wearing a different hat: mistaking a broad average for a personal forecast. Keeping the comparison at the program level, at the real net cost, over a stated horizon, is what turns the data back into a forecast you can act on.
What the Earnings Number Leaves Out
Even a clean, program-level earnings comparison is not the whole decision, because a salary figure is silent about several things that have real financial weight.
The first is risk and variance. Two programs can share a median and differ sharply in how spread out their outcomes are. A field with a tight range offers a more predictable result; a field with a wide range offers a higher ceiling and a lower floor. Neither is strictly better. A risk-tolerant student might prefer the wide field for its upside; a debt-loaded student might prefer the narrow one for its floor. The median hides this entirely, which is why the percentile band on each major profile is as important as the midpoint.
The second is job growth and durability. A field's current wage says what graduates earn now; its projected growth says whether the door stays open. A high-paying field with shrinking demand and a moderate-paying field with expanding demand can point in opposite directions over a career. This is the Job Growth Projections argument applied to the category debate, and it cuts across the STEM-liberal-arts line in both directions: some STEM fields are growing fast, some are flat, and the same is true on the liberal arts side.
The third is flexibility. A broad liberal arts field that feeds many careers carries an option value a narrow technical credential does not. If one destination closes, a wide field has others; a tightly credential-gated field is committed. That optionality does not appear as a dollar figure anywhere, but it is a real form of return, especially in a labor market that reshapes itself faster than a four-year degree takes to finish. The relationship between a field of study and the careers it can lead to is exactly what Major vs Program vs Career maps, and the Career Path Explorer lets you see how many destinations a given field actually opens.
None of these turns the earnings number into the wrong number. They make it one input among several. A complete comparison reads the median, then the range around it, then the growth behind it, then the flexibility it preserves, and only then decides. The STEM-versus-liberal-arts headline collapses all of that into a single contest the data was never shaped to settle.
Interest Belongs in the Comparison
A purely financial framing of STEM versus liberal arts misses the factor that often decides the actual outcome: interest, and its effect on finishing and excelling.
Choosing STEM purely for the earnings headline, without interest or aptitude, risks landing at the low end of the STEM range or not finishing at all, in which case the category advantage evaporates. A liberal arts field of genuine interest, pursued to a high level and kept affordable, can outperform a disengaged run at STEM. This is the Passion vs Paycheck logic applied to the category debate: interest has financial value, and the category averages assume an engaged, completing student that the disengaged chooser may not be.
The completion point deserves weight, because it is where the financial argument quietly turns against the cynical chooser. The published earnings for a field belong to its graduates. They say nothing about the students who started the program and left, and a degree not finished is the worst financial outcome available: the cost without the credential. Fields chosen against genuine aptitude or interest have higher attrition, and a student who switches out of a high-earning STEM program after two years, or stops short of a degree entirely, never touches the number that justified the choice. Interest is not a soft factor competing with the hard financial one. It is a direct input to the hardest financial fact in the whole decision, which is whether you finish at all. A field you will actually complete, at a cost you can actually carry, beats a higher-paying field you abandon, every time the math is run.
Where This Fits
This guide reframes a debate that recurs throughout the choosing-what-to-study cluster. It draws on Reading Earnings Data Honestly for the range argument, Job Growth Projections for the trajectory argument, and Passion vs Paycheck for the interest argument, and it feeds the data-validation steps of How to Choose a Major. The conclusion is that STEM versus liberal arts is the wrong question. The right one compares specific programs at specific net costs, where the answer is far more individual than the headline allows.
