where will you sit in the income distribution when your career peaks?
A multi-factor model that combines parental rank-rank persistence, education premiums, geographic mobility, and career trajectory into a probabilistic forecast of your peak-career income percentile. Grounded in Chetty's Opportunity Insights research, BLS earnings data, and the SCF wealth survey.
1. core architecture
The model is a layered prediction in three stages: (i) a background-driven prior, (ii) a current-income signal projected to peak career, and (iii) a Bayesian-style blend weighted by age. The final percentile is bounded between the 1st and 99th and is interpreted as your expected rank in the US household income distribution at peak earnings (≈ age 47).
2. the background prior (42 predictors)
Starting from Chetty et al.'s rank-rank persistence equation↗:
child_rank = 33.6 + 0.341 × parent_rank
Conditional adjustments are then added on top, additive in percentile points. The major contributors:
| predictor group | range (pts) | empirical source |
|---|---|---|
| education attainment | −10 to +30 | BLS earnings by attainment↗ |
| institution selectivity | −3 to +12 | Chetty Mobility Report Cards↗ |
| field of study | −5 to +9 | BLS occupational earnings (OES)↗ |
| current geography | −4 to +7 | Opportunity Atlas commuting zones↗ |
| upbringing geography | −6 to +5 | Chetty Land of Opportunity (2014)↗ |
| family structure | −4 to +3 | Fading American Dream (Chetty 2017)↗ |
| marriage | −3 to +8 | SCF dual-earner premium↗ |
| industry premium | −5 to +10 | BLS industry / NAICS earnings↗ |
| role level | −3 to +25 | Compensation surveys (Mercer)↗ |
| company type | −3 to +8 | BLS National Compensation Survey↗ |
| race (conditional on parent rank) | −8 to +4 | Chetty Race & Economic Opportunity (2018)↗ |
| parental wealth | −2 to +12 | SCF intergenerational wealth↗ |
| current wealth signal | −3 to +10 | SCF age-adjusted net worth↗ |
| savings rate / financial literacy | −2 to +5 | FINRA NFCS / capital compounding↗ |
| network strength | −2 to +5 | Chetty Social Capital (2022)↗ |
| salary negotiation ▸ behavioral | −3 to +5 | Babcock & Laschever, Women Don't Ask↗ |
| employer-change cadence ▸ behavioral | −1 to +5 | ADP Pay Insights — job-changer premium↗ |
| continuous upskilling ▸ behavioral | −2 to +5 | BLS skill premium & reskilling returns↗ |
| professional credentials ▸ behavioral | 0 to +6 | BLS certification & license premiums↗ |
| relocation flexibility ▸ behavioral | −2 to +5 | Opportunity Atlas mover effects↗ |
| AI leverage vs automation risk ▸ behavioral | −7 to +6 | Acemoglu & Restrepo — task automation↗ |
| work intensity ▸ behavioral | −2 to +3 | BLS CPS hours–earnings data↗ |
| sponsorship / mentorship ▸ behavioral | −1 to +5 | Ibarra — sponsorship research↗ |
| high-interest debt load ▸ behavioral | −6 to +1 | FINRA NFCS — debt drag↗ |
| retirement / match capture ▸ behavioral | −2 to +4 | SCF retirement & tax-advantaged data↗ |
3. bayesian blend with current income
Younger workers have unreliable current income signals (they haven't peaked). Older workers' background is fully realized. The model uses an age-based weight:
| current age | weight on current income | weight on background prior |
|---|---|---|
| < 25 | 15% | 85% |
| 25 – 30 | 30% | 70% |
| 30 – 35 | 50% | 50% |
| 35 – 40 | 65% | 35% |
| 40 – 45 | 78% | 22% |
| 45 – 50 | 88% | 12% |
| 50+ | 95% | 5% |
4. career projection & behavioral velocity
Current income is projected forward to age 47 (the empirical peak) using a roughly +0.6 percentile per remaining year for college-educated workers, with a slight decline post-55. Trajectory modifier (steady promotions, fast track, meteoric) shifts the slope.
On top of that, the ten behavioral predictors in category g — negotiation, employer-change cadence, upskilling, relocation flexibility, work intensity, and AI exposure — reshape the curvature of the growth phase, not just its endpoint. Aggressive strategy choices lower the growth exponent, bending the arc upward earlier (you reach peak sooner); an automation-exposed role raises it, flattening and delaying the climb. This is why changing a single choice in category g visibly moves the trajectory curve in section iii, not only the headline percentile.
5. probability distribution
Bands of probability are computed by integrating a normal CDF with σ ≈ 14 percentile points (residual variance from the rank-rank regression, tightened by the additional predictors). This is the bar chart you see in section ii.
6. income percentile mapping
Percentile-to-dollar conversion uses the 2024 US household income distribution from Census income data↗ — roughly: 50th = $75K, 80th = $160K, 90th = $234K, 95th = $315K, 99th = $700K+.
7. what the model doesn't capture
- Tail events: founding a successful company, marrying into wealth, or a disability event can move you ±20 percentile points and aren't modeled.
- Local variation: the geographic buckets are coarse — your specific commuting zone matters, not just whether you're in a "top metro."
- Tax / take-home: all numbers are gross household income, not after-tax cash flow or net worth growth.
- Cohort effects: wage stagnation, AI displacement, and asset price changes that will shape your generation aren't projected forward.
- Causation: the coefficients are conditional associations, not levers you can pull mechanically. Some are sticky (race, parental rank); others are genuinely actionable (education, geography, industry).