2026-08-22 Weekly Reads: Worrying About a Debt Crisis, Yet Barely Acting

Why This One

Recent WeChat forwards kept circling back to financial consumers, positions and leverage, blowups and exits, and why ordinary people stay at the table even when they know the risks. Those pieces tend to pin the problem on greed, ignorance, or wishful thinking. This August 2026 working paper by Ricardo Delao and Wenhao Li of USC Marshall offers a finer-grained breakdown: a person can simultaneously worry about a debt crisis, believe debt is still propped up by multiple forces, and not think that risk is enough to change their current vote or investment.

The paper studies U.S. government debt, not P2P, stocks, or personal borrowing. But the question it raises travels well: the fact that an asset has buyers today and a stable price doesn't mean everyone shares the same explanation for why it has value; demand that looks identical in calm times may not be the same kind of absorption capacity under stress.

Core Argument of the Paper

The authors don't offer options upfront. They first ask people, open-ended, what they think backs U.S. government debt, then organize the answers and existing literature into six sources: future fiscal surpluses, global demand for safe assets, the ability to keep rolling over debt, central bank support, real assets the government can sell or mobilize, and financial repression that channels savings into government debt through policy and regulation. Then the study has 985 screened government bondholders, 1,001 registered voters, and 247 people with graduate training in economics or finance allocate 100 points across these six sources.

The average rankings of the three groups are surprisingly close. Global safe-asset demand ranks first in every group, averaging 26 to 30 points; textbook-style future fiscal surpluses get only 14 to 16 points. Safe-asset demand plus continued rollover together get 43 to 46 points, close to half of all value, roughly three times the surplus share. Even among the economics and finance graduate group, future surpluses get only 16 points; professional training doesn't significantly change the ranking.

But averages hide divergence. Each respondent tends to pile heavy weight on the single source they believe in most, with the largest individual allocation averaging around 60 points. In other words, the group mean isn't a shared "synthesis theory" — it's the average of many different value frameworks. The open-ended answers still surface the same main channels, with safe-asset demand still prominent, so the ranking isn't just an artifact of the menu. It also differs noticeably from national newspaper coverage: the press emphasizes fiscal surpluses far more often and says very little about permanent rollover.

Why does this divergence matter? Because prices in normal times only tell us that total demand is sufficient to absorb existing supply — not who will still be willing to absorb it when stress hits. Two investors may hold the same Treasury at the same price today, one believing the world will always need safe assets, the other believing the government will generate future surpluses; if fiscal conditions deteriorate, their reactions could be completely different. Future surpluses depend on policies that haven't happened yet and can't be immediately handed to creditors; real government assets can be sold, and central bank resources can be deployed without adding the same kind of fiscal commitment. The same apparent debt size and price can therefore correspond to different crisis resilience.

The paper then separates value beliefs from risk concerns. All three groups put the average probability of a U.S. debt crisis in the next decade near 50%, broadly consistent with two other independent questions — the most likely timing of a crisis and the maximum sustainable debt-to-GDP ratio — suggesting the concern isn't an artifact of one particular phrasing. Yet among those who express concern, 91% of voters say debt didn't play a decisive role in their vote, and 72% of investors haven't made specific position changes because of it. Which value-belief someone holds also does little to explain who worries more or who acts.

A randomized information experiment throws this gap into sharper relief. The study shows some bond investors current debt levels and the Congressional Budget Office's long-term projections; their reported ten-year crisis probability rises by 14.9 percentage points. The share planning to reduce Treasury or bond fund positions rises by only 4.2 percentage points, and the estimate is imprecise. The information significantly shifts risk perception without producing a commensurate adjustment in risk exposure.

What's Worth Questioning

First, the paper measures stated beliefs, intentions, and past behavior — not actual trades, voting records, bond prices, or real reactions during a crisis. People can say they're worried in a survey, or treat 50% over ten years as a vague "that's dangerous," without actually pricing things that way. The authors also explicitly note the study doesn't observe the marginal investors who set market prices.

Second, allocating 100 points across six sources forces a cleaner breakdown than people would naturally produce in daily thinking. The open-ended responses reproduce the main channels, reducing menu-induction concerns, but don't eliminate the distance between survey expression and real decisions. The randomized information treatment raises risk judgments, but planned position cuts move only slightly and imprecisely — which can't be turned around to prove people are rationally choosing not to act.

Third, U.S. Treasuries simultaneously serve as global reserve assets, collateral, and dollar liquidity vehicles, so the conclusions can't be directly transplanted to local government debt, corporate bonds, P2P, or ordinary wealth-management products. The model explains how the same price can hide different stress tolerance, but the survey itself can't structurally identify market depth or refinancing capacity, nor tell us whether any particular belief is right. The working paper is still in peer discussion; numbers and interpretations may shift in later versions.

Takeaways for Recent Concerns

The most valuable thing to take from this research isn't "will U.S. debt blow up" — it's separating risk judgment from value support. For any asset or project, ask separately: who is absorbing today, and is the absorption driven by cash flow, scarcity, institutional support, someone else continuing to buy, or some implicit guarantee? All of these reasons can sustain prices in normal times, but they have very different stability under stress. Just looking at "it hasn't fallen yet" or "it's always been able to roll over" is exactly what hides the most important compositional differences.

The recent materials' point that "if you hold a position, don't leverage; if you leverage, don't hold a position" being more executable than grand forecasts also gets explained here. People can't necessarily judge which crisis narrative is correct, let alone predict the turning point accurately; but they can decide in advance how much exposure they can tolerate when concern can't be converted into reliable probabilities, under what conditions they must cut positions, and which funds can't enter high-volatility scenarios. Behavioral rules don't require resolving the entire macro debate first.

The paper also reveals the distance between "knowing a risk" and "letting that risk into decisions." Even a risk rated very high can be crowded out by other goals: voters also care about jobs, taxes, and public services; investors also care about liquidity, benchmark allocation, and alternative assets. Real risk management isn't collecting one more scary piece of information — it's writing down in advance what evidence would change action, by how much, and if you do nothing, which value-support you're effectively betting on continuing to hold.

The same applies to personal projects. A system can keep running on real paying users, your own time patching holes, external platform free tiers, or manual firefighting after every failure. The surface fact that "the service is still online" doesn't distinguish these supports. Listing the support sources, then asking which one would still work under stress, gets closer to whether the system is actually sustainable than just looking at current results.

How to Read It

Start with the authors' public non-technical overview, putting the six value supports and the three samples' numbers on one page; focus on distinguishing the group average weights from the divergence where each individual puts about 60 points on a single source. Then look at the open-ended responses, the allocation experiment, and the randomized information treatment in the paper, confirming which conclusions come from descriptive statistics and which from randomized comparisons. When reading the model section, hold onto one main thread: identical normal-time prices don't mean the same absorbing parties and available resources under stress. Finally, return to the limitations, always distinguishing stated beliefs, planned actions, and real market behavior.

Author overview: Wenhao Li: What People Think Backs U.S. Government Debt

Full paper: Delao and Li: Beliefs About Government Debt Valuation and Sustainability