Kent Smetters, Boettner Chair Professor at the University of Pennsylvania's Wharton School, has written an important article on “Why Debt Matters.” In the August issue of the Journal of Policy Analysis and Management, he makes a blunt claim: Washington’s preferred models cannot see how debt actually hurts the economy or why bond markets sometimes snap rather than drift. The issue isn’t just that federal borrowing slowly crowds out private capital. It’s that the modeling toolkit inside the Beltway filters out the mechanisms, confidence, intergenerational tradeoffs and fiscal limits that determine how and when a debt problem becomes a crisis.
Smetters’ starting point is disarmingly simple. If your retirement portfolio didn’t hold Treasuries, much more of your “safe” savings would end up funding productive corporate investment. With federal debt, however, a large share of savings is redirected to finance current consumption. Over time, that means less capital per worker, lower wages and higher borrowing costs. So far, that’s standard crowding out. But his deeper point is about dynamics: Debt risks are nonlinear. When investors stop believing they’ll be repaid in real terms, yields don’t climb politely; they jump.
What makes Smetters’ article especially valuable is not just the warning, but the diagnosis of why the warning isn’t landing. Much of official analysis still leans on the Ramsey “infinitely lived agent” model (often dressed up with a couple of reduced-form rules). In that world, debt is mostly a time-shift with little real bite; the math is convenient and the policy comfort is soothing.
The overlapping-generations (OLG) framework tells a different story. OLG forces us to model households, firms and government explicitly across time, with prices and quantities determined in general equilibrium. In that setting, crowding out, fiscal capacity and the possibility of self-fulfilling shifts in market beliefs appear naturally. It’s also the only framework that can treat explicit Treasury debt and the “implicit debt” of pay-as-you-go entitlements on the same economic footing.
Using the right model matters for how we read the evidence. Reduced-form regressions that mix recessions (which raise both the supply of government paper and the demand for safe assets) with secular debt accumulation will tend to understate the long-run debt effect on rates. Smetters demonstrates this by simulating data from an OLG model in which the true debt–rate relationship is strong and nonlinear: when the usual regressions are applied to that data, they misleadingly recover only a small contemporaneous elasticity. In other words, the empirical comfort is illusory as it arises because the measurement strategy blinds us to the underlying mechanism.
That critique dovetails with my own work on the debt sensitivity of interest rates. In a new policy brief, I estimate the relationship between interest rates and debt using quarterly U.S. data from 1985 to 2024 with a broad structural set of controls. The main result is straightforward: Every 1-percentage-point rise in the debt-to-GDP ratio lifts the 10-year Treasury yield by about 4.6 basis points (bps) in baseline, and 5–6 bps once you address leverage points and lags.
Those estimates sit squarely in the middle of the modern literature and exceed the Congressional Budget Office’s (CBO’s) current 2 bps parameter by 2–3 bps. That difference sounds small until you remember that long-term budget math compounds over decades: A higher debt–rate link feeds larger interest costs, which feed higher debt, which feeds higher rates. Pretending that link is half as strong as it is doesn’t make the burden go away; it just pushes the recognition into the future.
Smetters also highlights two omissions that make official projections look rosier than they should. First, official projections omit the fact that implicit debt from Social Security and Medicare is economically equivalent to explicit debt dollar-for-dollar in how it shifts resources across generations and crowds out capital; ignoring implicit debt inflates our apparent fiscal capacity.
Second, most reduced-form policy models, like those used in official projections, never truly close the budget. They roll debt forward indefinitely instead of forcing the eventual tax hikes, spending cuts, inflation or default that real-world closure requires. Markets won’t indulge that assumption forever. When beliefs flip from “repay in real terms” to “won’t,” yields reprice fast.
Further, Smetters surveys the range of institutional forecasts and shows how even small changes in assumptions transform the long-run debt picture. The Government Accountability Office (GAO) is already more pessimistic than the CBO: On GAO’s standard assumptions, debt climbs to 229% of GDP by 2054, and if borrowing costs rise just one percentage point above baseline, the trajectory accelerates to 275%. By contrast, if rates fall by one point, debt still hits 192%. The lesson is that debt paths are highly sensitive to seemingly modest shifts in rates or crowding out.
Even within CBO’s own framework, doubling the assumed crowd-out of private capital from 33% to 66% produces a grim projection: By 2055, federal debt held by the public would exceed 250% of GDP, with per capita output nearly a fifth lower than the extended baseline.
But the critical thresholds emerge most clearly in the Penn Wharton Budget Model simulations. Under the most optimistic assumption—that markets keep believing Congress will eventually stabilize fiscal policy—U.S. debt cannot sustainably exceed 200% of GDP. In reality, the danger zone is lower, somewhere in the 175–190% range, beyond which markets are likely to doubt repayment in real terms. Once that belief shifts, yields can jump abruptly. The implication is stark: The U.S. has less fiscal space than official projections suggest, and market confidence is the only fragile barrier between today’s trajectory and tomorrow’s crisis.
Where does that leave us? With three practical steps.
1) Use OLG as the default for long-horizon fiscal analysis; it’s the only framework that matches the economics of intergenerational policy.
2) Incorporate implicit liabilities and impose real closure in projections.
3) Update the debt-sensitivity parameter to at least 4 bps supported by empirical literature, rather than the current 2 bps comfort value used by the CBO.
If policymakers continue budgeting with models that don’t account for secular crowding out or sudden regime shifts, we’ll go on mistaking silence for safety until the bond market corrects us in a single step.