Run the median four-year degree through an honest after-tax model and you can get two very different answers, and the distance between them is a single number.

At a 3 percent discount rate, the median degree, at the net price families actually pay, clears about $210,000 over going straight to work. At my 7.8 percent rate, the same degree, with the same costs, the same earnings, and everything else held fixed, comes in about $17,500 behind. Nothing changed but the rate.[1] That one input decides whether the median graduate comes out ahead, which is all the headline debate has ever been about.

So it is worth asking plainly which rate is right. The answer is not a matter of taste. A near-risk-free rate is the wrong tool for this asset, for three reasons.

The discount rate is not a technicality

A discount rate converts future dollars into today’s dollars. A college degree pays late. The student gives up most of four or five years of full-time earnings, then earns more, with the largest gains arriving decades out. A low rate treats those distant dollars as nearly as valuable as cash in hand today. A high rate says a dollar you might earn at 50 is worth much less than a dollar you could earn and invest at 22.

Because the payoff is so backloaded, the rate you choose does more to the answer than any other assumption I have tested. Lower it far enough and most degrees look good. Some of the cheerful studies go that low, or lower. Georgetown University’s Center on Education and the Workforce left the headline figures of its original College Payoff undiscounted, with a 2.5 percent version only in an appendix, then used 2 percent in its college rankings and zero in its 2025 update.[2] Each lower rate would make college look better on the same facts.

Reason one: three percent is below what the family is paying

Start with the cost of the money. The rate should reflect what the capital is actually worth, and by each of the benchmarks that follow, that is well above 3 percent. Federal undergraduate loans currently carry about 6.5 percent and Parent PLUS loans about 9.1 percent.[3] The stock market has returned about 10 percent a year over the long run, or nearly 7 percent after inflation.[4] The 7.8 percent I use sits below the Parent PLUS rate and near the federal government’s own benchmark, the 7 percent real rate the Office of Management and Budget uses to approximate the return on private investment.[5]

Three percent is lower than the interest on the loan the family is taking out to pay the bill. It is below the yield on a long-term Treasury bond, which has run well above 4 percent this year. A degree is not a Treasury bond. Discounting it at the risk-free rate prices it as if the payoff were guaranteed by the United States government, which it is not.

Reason two: more than a third of buyers do not finish where they bought

More than a third of students who start a four-year degree do not finish at the college where they started within six years.[6] Consider what that means as an investment. You can buy this asset, pay in money and years, and still face more than a one-in-three chance of not finishing there within six years.

Finance prices risk. Nowhere does it give an asset with a 36 percent non-completion rate a risk-free rate without charging for that risk somewhere else. The riskier the payoff, the higher the return an investor demands, which is the same as saying the higher the rate the cash flows are discounted at. A startup is not discounted like a savings account. A degree that more than a third of buyers do not finish where they bought it is not discounted like a Treasury.

That risk has to show up somewhere in the math. You can put it in the discount rate, the way I do, or you can model it directly by weighting the outcome of the student who finishes against the one who does not. What you cannot do is leave it out of both. Some of the headline figures I have examined do exactly that. They report the outcome of the student who finishes and discount it at a rate that assumes no risk at all. In those, the non-completion rate appears nowhere.

Reason three: the later the dollar, the less it belongs to the degree

There is one more thing a high rate gets right. The premium is backloaded, so heavy discounting gives the least weight to the years when the wage gap is largest. That looks like stacking the deck against college. It is the opposite.

In my view, the late-career years are the years when other forces overwhelm the credential. By 55, a graduate’s income rests on promotions, industry conditions, the company’s fortunes, and three decades of their own performance. The college chosen at 18 is one factor among many, and probably not the dominant one.

I think of a peak salary at 50 as mostly something a career produced, not something a diploma caused, and a model that credits the whole late-career gap to the degree is overstating what the degree did. As with the completion risk, you could model this directly, trimming the late-career premium year by year. The rate carries both corrections in one number. It lets the near years, when the credential is doing its real lifting, carry the answer, and lets the far years count for less, which is where the credit belongs anyway.

The rate moves the count, not the conclusion

Even if you grant the lowest defensible rate, the failures do not go away.

At my 7.8 percent rate, nearly four in five four-year colleges come in below the high school path at full price. Run the model at 3 percent, below even the risk-free rate, and it is still nearly one in three at full price, and a fifth at the net price families actually pay. The discount rate does not decide whether degrees fail. It decides how many. That is 525 institutions where the median graduate, at full price, comes out behind even at that generous rate.[1]

So the discount rate is worth getting right, but it is not where the decision lives. A family is not choosing the national median. It is choosing one school, at one price, and at hundreds of those schools the math is negative at every rate anyone serious would use. The rate debate decides the headline. It does not save the family that picked one of the 525.

And the institution count understates it, because it only flags schools whose median graduate fails. Three more problems sit underneath, and not one of them moves with the discount rate. About 36 percent of students do not finish at the college where they started, paying for a credential that may never arrive.[6] About 42 percent of recent graduates who did finish are working in jobs that typically do not require a degree.[7] And roughly a quarter of the students who enroll at four-year colleges, finishers and non-finishers together, earn less than a typical high school worker six years after they start, an earnings comparison with no discounting in it at all, and most of them attend schools where most students clear that bar.[1] Count those and far more students fall short than attend the 525 schools, and the discount rate touches none of it.

Where a low rate fits

There is a fair use for a low rate. A policymaker asking whether college in the aggregate, across millions of students, returns more than the government’s borrowing cost can defend a low time-value rate. That is a portfolio question, and a portfolio diversifies away much of the individual risk.

A family is not a portfolio. A family makes one bet, once, with one child, at one price, with a real chance it does not pay off. For that decision the rate has to carry the risk the family is actually taking. At a rate that does, the median degree is not a sure thing. It is close to a coin flip, and it loses even at the net price families actually pay.

That is the whole rate dispute, and it comes down to one number. But the dispute only ever decides one thing, whether the median clears the bar. Saying college is worth it and saying college is a scam are the same mistake in opposite directions, a verdict on a median no one actually attends. Anyone who tells you college pays should have to say what discount rate they used, and whether that rate could survive more than a third of students not finishing where they started. In my experience, most cannot.

Run the math at any rate you like at collegeroi.org/roi.

Companion Pieces

• The Doubt Everyone Explained Away

Reference Sources

1. Shivamber, Leon. “College ROI Model and Calculator.” collegeroi.org, 2026. Accessed June 18, 2026. The after-tax net present value model across roughly 6,000 institutions behind every model figure in this piece: the median college NPV against the high school baseline (negative $17,510 at median net cost and a 7.8 percent discount rate, positive $210,602 at 3 percent), the share of four-year institutions with negative returns (79.2 percent at full cost falling to 31.3 percent when the rate drops to 3 percent, 57.6 to 21.4 percent at each college’s own net price), the share of four-year-college undergraduates enrolled at the 525 institutions still behind at 3 percent and full cost, about 9 percent, and the enrollment-weighted share of four-year-college former students, completers and non-completers together, earning below the high school baseline six years after entry, about 25 percent. Figures recomputed from the model’s own methodology by changing only the discount rate. The 7.8 percent run reproduces the distribution published in my book exactly.
2. Georgetown University Center on Education and the Workforce. “The College Payoff: Education, Occupations, Lifetime Earnings.” CEW Georgetown, 2011, and “A First Try at ROI: Ranking 4,500 Colleges.” CEW Georgetown, 2019 (its 2025 update says its earlier reports used this 2 percent rate). Accessed June 18, 2026. The 2011 College Payoff’s headline figures are undiscounted. Its technical appendix states “We chose 2.5% because this represents the real interest rate of long term government bonds,” the ROI rankings state “we use a 2 percent discount rate,” and the 2025 ROI update removed the discount rate entirely. Each lower rate raises the present value of backloaded college earnings.
3. Federal Student Aid, U.S. Department of Education. “Interest Rates and Fees for Federal Student Loans.” Federal Student Aid, 2026, and “Interest Rates for Federal Direct Loans First Disbursed Between July 1, 2026 and June 30, 2027.” Electronic Announcement GENERAL-26-33, June 4, 2026. Accessed July 24 and September 25, 2026. Undergraduate Direct loans at 6.52 percent and Parent PLUS loans at 9.07 percent for loans first disbursed July 1, 2026 through June 30, 2027. The prior award year, which ran through June 30, 2026, carried 6.39 percent and 8.94 percent (Electronic Announcement DL-25-03, May 30, 2025).
4. Damodaran, Aswath. “Historical Returns on Stocks, Bonds and Bills.” NYU Stern, Damodaran Online, 2026. Accessed June 18, 2026. The long-run annualized return on U.S. equities, roughly 10 percent nominal and nearly 7 percent real (geometric averages for 1928-2025 in the page’s data workbook: 10.0 and 6.8 percent).
5. Office of Management and Budget. “Circular A-94: Guidelines and Discount Rates for Benefit-Cost Analysis of Federal Programs.” The White House (archived), the 1992 version, reinstated by OMB Memorandum M-25-23 (April 8, 2025), which revoked the 2023 revision. Accessed June 18, 2026. Specifies a 7 percent real discount rate that “approximates the marginal pretax rate of return on an average investment in the private sector in recent years.” My 7.8 percent personal rate, the one my book uses, sits near this benchmark. It is set above the risk-free rate to price completion and earnings risk, below the Parent PLUS rate.
6. National Center for Education Statistics. “Undergraduate Retention and Graduation Rates.” NCES, Condition of Education, 2022. Accessed June 18, 2026. The six-year graduation rate for first-time, full-time bachelor’s-seeking students who started in fall 2014 was 64 percent, counting only degrees finished at the college where they started, so about 36 percent did not finish there within six years.
7. Federal Reserve Bank of New York. “The Labor Market for Recent College Graduates.” Federal Reserve Bank of New York, 2026. Accessed September 24, 2026. A graduate working in a job that typically does not require a college degree is counted as underemployed. The underemployment rate for recent graduates was about 42 percent in the second quarter of 2026 (41.95 percent in June, in the data updated August 6, 2026), up from 41.5 percent in the first quarter.