Follow your dreams! (Or not?)
Oskar's hypothesis was blunt: passion careers don't pay. Do what you love and you take a pay cut for the privilege, reliably enough to call it a rule.
It holds, but only for one kind of passion, and only after you survive three decisions about how to measure it. Artistic work does pay less — about 18% less at the most artistic end of the professional workforce than at the least. Work built around teaching and helping people looked like it paid less too, until accounting for how many women do those jobs made that penalty disappear. And curiosity-driven work pays a premium.
The setup
The earnings come from the American Community Survey: 2.16 million full-time workers aged 25–64 with a bachelor's degree or better. The passion measure comes from O*NET, the Labor Department's database that scores every occupation on six interest dimensions — artistic, investigative, social, enterprising, realistic, conventional — each from 1 to 7. Graphic designers score 7.0 on artistic. Software developers score 2.3.
For each occupation I took the median wage and asked whether the interest scores predict it. A correlation of 0 means the interest score tells you nothing about pay. A negative correlation means the more of that interest an occupation involves, the less it tends to pay — loosely, on average, with plenty of exceptions.
Three choices in that setup change the answer: which variable stands in for passion, which workers count, and which other factors you account for.
Choice one: which passion variable
O*NET publishes six "work values" alongside the interest scores, and one is called Working Conditions. It is the tempting passion proxy — interests describe what kind of work a job involves, work values describe how rewarding people find it, and "rewarding" sounds closer to what following your dreams means.
Working Conditions correlates 0.80 with actual median earnings. Read the definition and you find out why: good pay is one of the things the score is built from. Use it and you get a confident, well-powered finding that passion pays — produced entirely by having earnings on both sides of the equation. The interest scores contain no pay information, so the analysis runs on those instead.
Choice two: which workers count
Include every bachelor's-degree holder and artistic interest predicts nothing about pay at all (correlation 0.04, which is statistical noise). Restrict the sample to professional occupations and a real penalty appears (−0.25). Same data, same method, opposite conclusions.
The reason is who gets excluded. Degree-holders working as retail supervisors, receptionists and administrative assistants sit at the bottom of the all-workers pay distribution, and their jobs score low on artistic interest. They pile up in the low-artistic, low-pay corner, and that pile drags the fitted line flat, cancelling out the downward slope the artistic professions create.
Both samples answer a real question. They answer different ones. "Do people who followed a calling earn less than people who chased money" assumes someone who had that choice, which argues for the professional restriction — but that is an argument made before any statistics run, and it decides the sign of the result.
Choice three: what else explains it
Inside professional occupations, two interests carried penalties. Artistic, and social — the dimension scoring how much a job involves teaching, helping and caring for people. Nurses, teachers, social workers, therapists.
Social interest correlates 0.62 with the share of an occupation that is female. And occupations lose pay as they feminize, independent of the skill they require — Levanon, England and Allison tracked this across a century of census data.
Adding percent-female to the model tests whether the social penalty is really about social work or really about gender. Controlling for something means asking what the relationship looks like when you compare occupations at the same level of that thing — female-heavy occupations against other female-heavy occupations, rather than against male-heavy ones. Done that way, the social penalty collapses to zero. Percent-female absorbs all of it. The social penalty was pay discrimination against women's work, measured through an interest inventory.
That leaves the other four interest dimensions, and one of them behaves unlike anything above. Investigative interest — how much a job runs on research, analysis and figuring things out — predicts higher pay, not lower, and survives both tests unchanged.
The artistic penalty survives the identical test. It also survives a harsher one: rebuild every occupation's median wage using only its male workers. That strips out any gap between what men and women earn inside the same job, so what remains is a penalty attached to the occupation itself. The artistic penalty is still there — slightly larger, in fact. Artistic interest is nearly uncorrelated with percent-female (0.11), so the two were never measuring the same thing.
What's left
The hypothesis half-survives. Artistic passion carries a real pay penalty. Investigative passion is paid a premium. The social professions' penalty belongs to gender discrimination rather than to passion.
Graphic designers take a median of $58,000. Software developers take $120,000. The artistic penalty accounts for about a fifth of that gap — the rest is everything else that separates the two jobs.
| effect per scale point | with % female controlled | male-only wages | |
|---|---|---|---|
| Artistic | −3.7% | −3.3% (t = −2.27) | −4.1% (t = −2.61) |
| Social | −5.1% | −0.3% (t = −0.18) | −0.5% (t = −0.30) |
| Investigative | +7.5% | +6.5% (t = 4.48) | +5.9% (t = 3.79) |
| % female | — | −60% across the range (t = −5.09) | −49% (t = −3.89) |
196 professional occupations, ACS 2019–2023 five-year sample. t-statistics beyond about ±2 are conventionally read as unlikely to be chance.