↖︎ Vishal Singh
The ZIP Code Destiny·Data Story № 37

The Toll
Booth

Higher education is sold as the great equalizer — the one institution that can override an unlucky birthplace. But read its ledgers and it looks less like an escalator than a toll booth: which college you can reach, what it charges a poor family, and what it pays out decide whether a degree is a mobility engine or debt without a diploma.

The ZIP Code Destiny · U.S. Dept. of Education College Scorecard 2025 · IPEDS 2023 · ADI · USDA RUCA · June 2026
$30.0k
for-profit median 10-yr earnings
62%
net price ÷ income, poorest public-4yr tier
2,200
counties with no four-year college
11 → 29
miles to nearest 4-yr, least → most deprived

For a century the American story about college has been a story about escape. The land-grant act, the GI Bill, the Pell Grant — each was sold as a toll-free road out of wherever you happened to be born. And in the aggregate the road is real: a bachelor's degree still buys a large lifetime earnings premium. But "college" is not one thing. It is 6,400 separate institutions charging different prices, admitting different students, and paying out wildly different futures — and which one a teenager can actually reach is shaped, like everything else in this book, by an address.

Raj Chetty, John Friedman and their co-authors made the point unforgettable in their 2017 "Mobility Report Cards": a handful of mid-tier public colleges and city universities move far more poor kids into the upper-middle class than the Ivies do, because they actually enroll poor kids and graduate them into decent jobs. David Deming and colleagues documented the dark mirror — for-profit colleges that recruit the disadvantaged, load them with debt, and leave them no better off in the labor market. Caroline Hoxby and Christopher Avery showed that thousands of high-achieving low-income students never even apply to the selective schools that would cost them almost nothing. This plate maps the toll booth using what the federal government now publishes for every institution: its earnings, its net price by family income, its completion rate, and — joined through IPEDS — its physical location. Three gates, one address.

01

What the gate pays out

Start at the exit. Ten years after enrolling, the median student who took federal aid at a public college earns about $42,400; at a private non-profit, $51,700. At a for-profit college the median is $30,000 — below the other two distributions almost entirely, and a $12,300 shortfall against even the public median. For-profits enroll the most Pell-eligible students of any sector — a median Pell share of 55%, against 31% at public and 34% at private non-profit colleges — and pay them the least. That is the for-profit earnings penalty Deming and colleagues estimated experimentally; here it is in the raw distribution of every Title-IV institution in the country.

Distribution of median 10-year earnings of federally-aided students, by institutional control. Each curve is the share of institutions in $5k earnings bins; the tick marks are sector medians. These are medians of Title-IV-aided students, not earnings by parental income. Source: U.S. Dept. of Education College Scorecard 2025 (md_earn_wne_p10). Cf. Deming, Goldin & Katz 2012.
02

A progressive price, a regressive bill

Now the price of admission. Net price — what a family actually pays after grants and scholarships — is published by family-income tier, and on paper the schedule looks fair: at the median public four-year college a family earning under $30k pays about $9,300, while a family over $110k pays $18,900. The sticker scales with means. But a price is only meaningful against an income. Divide each tier's net price by the midpoint of its income band and the schedule inverts: the poorest tier is handed a bill equal to roughly 62% of its income, the richest tier one worth about 13%. The progressive discount is real and still not nearly enough. The same shape, more expensive at every rung, appears at private colleges.

Median net price by family-income tier (left) and net price as a share of tier-midpoint income (right), public vs private four-year colleges. Burden uses tier midpoints ($15k / $39k / $61.5k / $92.5k / $150k). Net price is the average for aid recipients in each band; not all families pay the median. Source: College Scorecard 2025 (npt41–npt45). Cf. Hoxby & Avery, BPEA 2013.
03

Debt without a degree

Put cost on one axis and payout on the other and the colleges sort into quadrants. Most expensive schools do pay off — cost and earnings are positively correlated (r = 0.61), and completion tracks earnings too (r = 0.56), so this is not a story of cost being irrelevant. The danger is concentrated in one corner: institutions that combine low completion rates with low earnings, where students borrow and leave without the credential that would have repaid the loan. 108 institutions sit in that debt-without-a-degree quadrant — a sliver of for-profits among them, but mostly open-access two-year and certificate programs whose students were poor going in and poorer coming out. The toll was paid; the gate never opened.

Median 10-year earnings vs cost of attendance, one dot per institution, colored by control and sized by undergraduate enrollment. The shaded box flags the debt-without-a-degree quadrant (completion below 40% and earnings below $32k). A fixed-seed sample of ≤4,500 institutions is drawn. Source: College Scorecard 2025 (costt4_a, md_earn_wne_p10, c150). Cf. Deming et al., AER 2016.
04

The mobility engines

The toll booth is not destiny. Among the third of four-year colleges that enroll the most Pell students, some both take poor kids and lift them — high Pell share, real post-college earnings, and a near-zero net price for the poorest families. Rank them on all three and the list is almost a roll-call of urban public systems: CUNY's Baruch, City, and Hunter colleges; the California State campuses; and Berea College, the tuition-free Appalachian school that admits only low-income students. These are exactly the institutions Chetty's mobility report cards crowned as the country's strongest engines. The contrast at the bottom is just as instructive: schools that enroll just as many poor students but graduate them into earnings barely above the high-school line.

The fifteen strongest mobility engines among high-Pell four-year colleges, scored on a composite of Pell share, earnings, and (low) net price for the poorest families. Restricted to U.S. institutions with ≥1,000 undergraduates in the top third of Pell share. The composite is descriptive, not a validated mobility estimate. Source: College Scorecard 2025. Cf. Chetty, Friedman, Saez, Turner & Yagan, NBER 2017.
05

The deserts

Before a teenager weighs price or payout, the campus has to be reachable at all. Of the country's 3,144 counties, only 944 contain a four-year college; 2,200 have none. The map shades each college-less county by the great-circle distance to its nearest four-year institution — gray where a college sits inside the county line. The pale eastern seaboard and the dense Midwest hide their colleges close at hand. The deep reds are the Great Plains, the Mountain West, the rural South, and above all Alaska, where the nearest four-year campus can be hundreds of miles and a flight away. For a low-income family without the savings to send a child to board, distance is itself a price.

Distance to the nearest four-year college, by county. Darker = farther (miles); gray = a four-year college lies within the county. Distance is measured from county internal points to institution coordinates as the crow flies. Hover any county for the distance and nearest campus. Sources: IPEDS 2023 (institution locations); Census counties_2020 internal points. Distance is supply geography, not enrollment.
06

Farthest from those who need it most

The deserts are not randomly placed. Sort counties into ten deprivation deciles and the distance to a four-year college climbs almost monotonically: from about 12 miles in the least-deprived tenth to 29 miles in the most. Rurality multiplies it again — in rural counties the average resident is 34 miles from a four-year campus against 11 miles for urban residents, a roughly three-fold gap, and 9 in 10 rural counties have no four-year college at all. The toll booth is literally placed farthest from the families with the least ability to pay the toll. This is the supply landscape that the rest of the chapter's choices — what to charge, whom to admit — then operate on top of.

Population-weighted average distance to nearest four-year college, by county deprivation decile (left) and by rurality (right), with the share of counties having no four-year college. Deprivation is the ADI national rank; rural = majority of residents in USDA RUCA codes 7–10. Connecticut's renumbered counties lack RUCA and drop from the rural split. Sources: IPEDS 2023; ADI via tract atlas; USDA RUCA via dim_geo.
07

Reading it honestly

The single largest caveat is geographic. Every college on these maps is pinned to its own street address, so what we have measured is the supply landscape — where colleges are — not where any given neighborhood's children actually enroll. A teenager in a college desert may board hundreds of miles away; one next door to a flagship may never set foot in it. The gold standard for the true childhood-tract-to-college link is the Chetty–Friedman mobility report-card data, which traces individual students from the ZIP codes they grew up in to the colleges they attended and the incomes they reached. That dataset is not in this project's staged lake, and we do not pretend to reproduce it; the distance maps here are a coarse proxy for access, useful for showing the shape of the problem and nothing finer.

The earnings figures carry their own asterisks. College Scorecard reports median earnings of Title-IV-aided students only — not all graduates, and not broken out by the income a student started from — so the for-profit penalty conflates the sector's effect with the disadvantage of the students it recruits; the experimental work of Deming and colleagues is what isolates the former. Small cohorts are suppressed, completion rates mix two- and four-year programs, and the mobility-engine composite is our own arithmetic, not a causal estimate. Connecticut's planning-region FIPS mismatch nulls it out of the deprivation split. What the data here can establish is the descriptive architecture of the toll booth: a real for-profit earnings gap, a net-price schedule that is progressive in dollars and regressive in burden, and a supply of four-year colleges distributed farthest from the poor and the rural. What it cannot do is follow a single child through the gate. For that, the chapter leans on the causal studies named below.

Notes & data