The Reasoning-and-Wage Dataset: 739 Occupations, Full Method and Open Download
The result comes from a single, deliberately narrow index. We took three O*NET reasoning abilities, averaged them into one demand score per job, and regressed that score against 2024 federal wage data. No proprietary weighting, no hidden model. The inputs are public, the code is archived, and the full table is below.
Key Takeaways
- A reasoning-only demand index explains 49% of the variance in occupational median wages [IQ Career Lab analysis, 2026] across 739 jobs, a linear R-squared of 0.485 (95% CI 0.453 to 0.522)
- The log-wage fit, standard for right-skewed pay, reaches 0.605 (95% CI 0.575 to 0.638), with a Spearman rank correlation of 0.795 [IQ Career Lab analysis, 2026]
- The slope is about $37,490 in median annual pay per one-point climb on the O*NET reasoning scale of 0 to 7 [IQ Career Lab analysis, 2026]
- This is an occupation-level figure, not an individual one. Your own IQ tracks only about 4% of income differences between people, r near 0.20 (Strenze, 2007). Conflating the two is the ecological fallacy
- Everything here is downloadable. The 739-row table ships as an open CSV and a full JSON with the statistics block
What the Index Measures
The demand score for each job is the mean of just three ONET 29.1 ability levels: Inductive Reasoning (element 1.A.1.b.5), Deductive Reasoning (1.A.1.b.4), and Mathematical Reasoning (1.A.1.c.1), each rated on the ONET Level scale of 0 to 7 [IQ Career Lab analysis, 2026]. We chose these three because they are the closest labor-market proxy for the fluid reasoning that most IQ tests lean on hardest, and nothing else.
That narrowness is the point. These three abilities are a deliberate slice of O*NET's 52 ability descriptors, not an IQ test for jobs and not a full cognitive-demand model. Isolating them lets their explanatory power be read cleanly against the 739 wage points [IQ Career Lab analysis, 2026] rather than buried inside a larger composite.
The Method, Stated Plainly
We pulled ability levels from the ONET 29.1 Abilities file (db_29_1_text/Abilities.txt, the LV rows) and median wages from the BLS Occupational Employment and Wage Statistics national file for May 2024 (oesm24nat/national_M2024_dl.xlsx, the detailed-group A_MEDIAN column). We joined the two on the six-digit SOC code, averaging ONET detailed occupations unweighted within each SOC, which left 739 matched occupations [IQ Career Lab analysis, 2026]. Twelve top-coded medians above the BLS ceiling of $239,200 were set to that ceiling.
The demand score is mean(inductive, deductive, mathematical). R-squared is the squared Pearson correlation against the median wage in dollars (the linear fit) and against the natural log of the median wage (the log fit, which is standard because pay is right-skewed) [IQ Career Lab analysis, 2026]. The rebuild script and every input file are archived alongside the open dataset.
The linear fit lands at an R-squared of 0.485, so reasoning demand alone accounts for about 49% [IQ Career Lab analysis, 2026] of the variance in occupational median wages, within a 95% confidence interval of 0.453 to 0.522. The log-wage form, corrected for right skew, reaches 0.605 [IQ Career Lab analysis, 2026]. The slope is about $37,490 in added median pay for each one-point climb on the reasoning scale [IQ Career Lab analysis, 2026].
Does the Result Survive Stress Tests?
Yes, on every check we ran [IQ Career Lab analysis, 2026]. Dropping the 12 top-coded salaries moves the log fit by almost nothing, from 0.605 to 0.601. Weighting each occupation by its employment count lifts it to 0.723. A 2,000-sample bootstrap puts every confidence interval far above the 4% individual-level figure (Strenze, 2007).
None of this is an artifact of a few thin, unusual jobs. It is a genuine, robust between-occupation pattern that holds under every reweighting we tried [IQ Career Lab analysis, 2026].
Not all three abilities pull the same weight. On its own, inductive reasoning leads at a log-wage R-squared of 0.631, with deductive close behind at 0.612 and mathematical reasoning weaker at 0.396 [IQ Career Lab analysis, 2026]. The chart shows each ability's standalone share of occupational wage variance.
Does the Ranking Pass a Face-Validity Check?
Yes. The occupations sort the way you would hope: physicists, mathematicians, and biochemists sit at the top of the reasoning scale and are paid accordingly, while refuse collectors and textile pressers sit at the bottom of both [IQ Career Lab analysis, 2026]. The climb from one end to the other is smooth rather than jumpy.
The table below shows the extremes. Reasoning index runs on the O*NET scale of 0 to 7; wages are BLS OEWS May 2024 medians, with the one top-coded row flagged. The full ordering of all 739 jobs is in the open CSV, so you can check any occupation you care about.
For the reader-facing version of what this means for a career, we walk through how much IQ affects your income and rank the cognitive demand of high-paying roles with the same index.
| Reasoning index (0 to 7) | BLS 2024 median wage | |
|---|---|---|
| Physicists | 5.63 | $166,290 |
| Mathematicians | 5.42 | $121,680 |
| Biochemists and Biophysicists | 5.00 | $103,650 |
| Statisticians | 4.94 | $103,300 |
| Refuse collectors | 1.37 | $48,350 |
| Textile and garment pressers | 1.62 | $33,880 |
Why This Is Not an IQ-to-Income Number
The index also controls for nothing else, so treat 49% as a correlational ceiling, not proof that reasoning sets the wage [IQ Career Lab analysis, 2026]. Licensure and credentialing barriers, unionization, industry, seniority, and long hours all pad the pay of high-tier jobs, and a demanding role may simply be one that screened hard on a degree or a license. Cognitive difficulty can mark the tier without paying for it, which is why the log-scaled fit near 0.605 [IQ Career Lab analysis, 2026] is the more honest number to quote.
For scale, published work sets the ceiling. When Dey and Loewenstein modelled the full bundle of O*NET job requirements against wages, the fit reached an R-squared of 0.933 [Dey and Loewenstein, 2019]. Our reasoning-alone slice recovers close to half of that on the linear fit and about 60% on the log fit [IQ Career Lab analysis, 2026], which is what you would expect from three abilities standing in for a much larger recipe.
How to Cite and Reproduce
Cite this as the IQ Career Lab reasoning-demand and occupational-wage dataset, 2026, built from O*NET 29.1 Abilities and BLS OEWS May 2024. Both source datasets are public and named in the method above, and the full JSON carries the exact statistics, confidence intervals, and robustness runs used here. To reproduce the figure, average the three named ability levels per SOC, join to the BLS median, and square the Pearson correlation against wage and against log wage.
If you want to know where your own reasoning profile would land on this scale before reading too much into any single job, you can take a free cognitive assessment built from the same O*NET-aligned abilities.
See Which Cognitive Tier Fits Your Reasoning
The number is only as trustworthy as its method, and now the method is yours to check. Download the 739-row table, rerun the fit against the 2024 BLS medians, and hold the figure to account.
