How to analyze any stablecoin in 30 minutes

Three questions that filter the rest

Chapter 1 split "1 USDC = 1 USD" into a technological claim, a legal claim and an economic claim, then left three questions about the loop: who can pass, what it costs, and when it is open. This chapter keeps those, and restates them as a filter for whether the other thirteen questions are worth the remaining minutes.

Whose legal obligation is the token. The answer is a named legal entity with a jurisdiction and an address, not a brand. Possession of USDC does not, by itself, create a right to walk up to Circle and be paid. Circle Mint is the desk. Tether Limited runs the same structure for USDT. If the name cannot be written down, the later questions about reserves are questions about an unnamed book.

Who has the right to redeem at par. Direct redemption is a primary-market trade for a verified client. A holder who bought on an exchange has a secondary-market exit at the screen price. Mixing those two is the most common error in a pitch deck. USDG is described in chapter 3 as fully redeemable 1:1, a stronger retail term than USDT's and USDC's institutional desks. USDe has no formal par window.

What happens when the arbitrage loop stops for 48 hours. A partner bank fails on a Friday, a hedge venue suspends withdrawals, a regulator pauses mint. Either a backup path exists, or the price drifts without a bound. Circle's loop was closed operationally for about sixty hours in March 2023. The reserve still existed. The print went to about 0.87 USD on 11 March because the desk was shut.

Those three answers set the vocabulary for questions 2 through 6. Question 4, who controls the collateral, often fills question 13, regulation, without a second document: an OCC trust charter, a NYDFS BitLicense, or a blank. Question 6, redemption, predicts question 14, liquidity, because the arbitrage-threshold arithmetic only runs while someone can actually pass through the desk.

Sixteen questions in thirty minutes

The sixteen questions have a fixed thirty-minute budget. Four of them, collateral type, mint mechanics, customer identity, and the cycle-survival synthesis, take sixty seconds each. Two of them, risks and regulation, take nearly five minutes each because they rarely fit in one document. The other ten share what is left. The split is the author's allocation of that outline, not a stopwatch reading. If the offering document is missing, the clock overruns, and that overrun is itself a result.

Order is not decorative. Question 1 names the mechanism family, so questions 2 to 6 are asked in that family's language: a cash desk, a vault, or a hedge book. Asking "is it fully reserved" of USDe is a category error. Asking "who is the futures counterparty" of USDC is the same error in reverse. USDe is a synthetic, not a payment stablecoin.

1. Peg mechanism. Redemption arbitrage, over-collateralization with liquidation, or a delta-neutral hedge. "Fully backed" is not a mechanism. USDC at 0.87 USD in March 2023 was fully reserved on the technological and legal layers while the loop was closed.
2. Collateral type. Cash and T-bills, over-posted crypto, a hedged book, metal, or nothing. Sixty seconds. If the type cannot be named, stop.
3. Mint mechanics. Who may create tokens, against what, and whether that permission is a contract role or a desk. Sixty seconds.
4. Who controls the collateral. A licensed entity with an address, a DAO, or a book of futures at venues. This is the question that often answers regulation for free.
5. The attestation or reserve page. Named attestor, snapshot date, standard, and what the letter does not cover. How to read that page sits one section down.
6. How redemption works. Who may demand one dollar for one token at the issuer, on what ticket, on which days. Not "I can always sell for a dollar."
7. Who the customer is. Primary-market client, secondary holder, or a user who showed up for a reward. Sixty seconds. PYUSD's supply tracked the reward, not transactional use.
8. Freeze and blacklist. USDT and USDC ship a blacklist. USDS added a freeze DAI never had.
9. Yield and wrappers. The GENIUS Act bars a payment-stablecoin issuer from paying interest directly. Wrappers remain an open regulatory question as of September 2026.
10. Risks. Nearly five minutes. If official documentation has no risk section, that is a red flag on its own.
11. Depeg history. Depth does not predict return. USDC fell 13% and came back. USDX fell more than 40% and did not. Chapter 2 keeps the episode table. A depeg with no post-mortem is disqualifying.
12. Distribution. DefiLlama and RWA.xyz against the token's own documentation, dated. A few percent of aggregator disagreement is normal.
13. Regulation. Nearly five minutes. A licensed-status claim absent from every public register is disqualifying. Question 4 often pre-fills this cell.
14. Liquidity. Whether 50 mn USD can exit without moving the price more than a few basis points. Question 6 predicts the answer.
15. Named legal entity. No named party responsible for redemption is a red flag on its own.
16. Cycle-survival synthesis. Sixty seconds. Fold questions 1, 4, 6 and 10 into one sentence: what breaks first, and who is on the other side of it.

The parallel habit for a tokenized fund, with a different asset class and a thirteen-question card, is How to analyze any RWA in 30 minutes.

Four questions in five minutes

The five-minute version answers only questions 1, 4, 6 and 10: peg mechanism, who controls collateral, how redemption works, risks. Every historical disaster named in this series, UST, deUSD, xUSD, USDX, FDUSD, had a visible weakness in exactly one of those four points. The Polish closing chapter lists FDUSD in that set; the specific print is not restated in the sources used for this chapter, so the name is kept and the depth is not invented. Distribution and regulation matter. They are rarely what killed the token by themselves.

The Polish closing chapter asked the same four, with the 48-hour loop stop sitting inside the risks cell.

The five-minute pass produces four sentences and a list of holes, not a safe/unsafe stamp. The thirty-minute pass fills the other twelve so that a hole is named rather than skipped.

The scorecard and its decision rule

Each of the sixteen questions has five columns: the question, a one-sentence answer, a source with a date, a confidence mark, and a red-flag box.

A red flag is a gap or a fact that, on its own or in a stack, means the product cannot be used in a business decision until someone has done full due diligence. It is not a moral label. It is a stop.

Confidence is the reader's mark on how sure that one-sentence answer is, given the source level. High means a primary document dated by the reader. Medium means a dashboard or an issuer page. Low means a secondary portal, a missing document, or an inference. Low confidence on questions 1, 4, 6 or 10 is treated as a red flag even when the red-flag box is empty, because those four are the five-minute filter.

The decision rule is mechanical. Four or more red flags, or low confidence on any of questions 1, 4, 6 or 10, triggers full due diligence before any business decision.

BDO, Deloitte and KPMG are the attestor names that themselves carry a signal. Circle's USDC page is a monthly Deloitte attestation plus weekly disclosures. Tether's USDT page is quarterly BDO Italia, under ISAE 3000, which is an assurance opinion on management's assertion for a stated date, not a year-round test of controls. Paxos moved PYUSD's attestor to KPMG. Naming the firm is part of question 5. It is not a substitute for reading what the letter covers.

Gaps that are normal, and gaps that disqualify

Some missing numbers are ordinary. A private issuer's exact operating-cost breakdown is not public, and Tether does not file a 10-K. An undisclosed multisig signer count, beyond a headline number of signers, is common. A few percent of disagreement between DefiLlama, RWA.xyz and the token's own page is a method gap, not evidence that the reserve is short. Circle's attestation showed 77.17 bn USD of USDC on 30 April 2026, net of unissued and denied tokens; DefiLlama's raw on-chain figure was 73.71 bn USD on 29 June 2026. Different dates, different cuts. They are not averaged. A shallow depeg with no precise recovery timestamp in the press is annoying and not, by itself, disqualifying.

Disqualifying gaps are red flags on their own:

• the issuer publishes no attestation or audit at all
• no named legal entity is responsible for redemption
• official documentation contains no risk disclosure
• a depeg exists with no published post-mortem
• a claimed regulatory status cannot be verified in any public register

An attestation, recalled from chapter 3, is an accountant's opinion on management's assertion for a stated date. It does not confirm reserve composition in every line, encumbrances, or internal controls. An audit, in the reserve-specific sense that chapter uses, is a full examination of the reserve over a period. As of 2026, neither USDT nor USDC has a published full reserve-specific annual audit in these sources. Circle's financial statements as a public company are a different object from a reserve audit of USDC. A proof of reserves is a third document: a moment-in-time check that some assets exist. It does not, by itself, say they are unencumbered or that they cover the whole liability.

Reading an attestation, in practice, is five lines on the scorecard. Who signed it (BDO, Deloitte, KPMG, or nobody). Which standard (ISAE 3000, an AICPA attestation standard, or an unnamed letter). Which date the snapshot covers, written in the same cell, because a quarterly letter does not speak for the weeks after it. What management asserted (usually that circulating tokens are covered by the fair value of reserves). What the accountant did not test (composition, liens, the weekend the bank is closed, the desk hours). If any of those five lines is blank, confidence on question 5 is low. If the first line is blank, question 5 is a red flag.

A working source hierarchy for this asset class, from closest to a binding statement to furthest:

1. Redemption terms and the attestation. They bind the issuer, or they are the accountant's opinion on what the issuer asserted.
2. A regulator register or filing. GENIUS, MiCA EMT, NYDFS, OCC charter. Question 13 lives here.
3. The issuer's own transparency page. True and selective.
4. A dashboard. DefiLlama, RWA.xyz. Dated, filters usually hidden.
5. A secondary portal or an academic preprint that repeats a dashboard. A lead, not a number to act on.

Most circulating supply figures sit on levels four and five. They need a fresh read and the reader's own date. If that read cannot be done, the number is citable with a caveat, which is a medium or low confidence mark, not a silent copy.

What the papers cover, and what they miss

OpenAlex is a research index. The counts below are OpenAlex's API read on 5 September 2026: 1,710 records from 2023 through that day with "stablecoin" in the title or abstract. Annual counts went 213, then 190, then 440, then 867. 2026 is 8 months and 5 days, not a full year. A full-year 2026 total is not stated here.

Raw 2026 growth reads as +97.0% over 2025. 238 of the 867 records are automated data deposits, mostly one Zenodo hourly-dataset series, all zero-citation. Excluding them, 2026 falls to 629 papers and growth drops to +43.0%. The preprint and repository share of the raw 2026 pile is 553 of 867. Peer-reviewed journals are 171 of those 867, down from 206 in the whole of 2025, which is expected: review takes months, and 2026 is incomplete. Central-bank and major-institution authorship (BIS, ECB, Fed, NBER, IMF) is 67 of 1,710 papers, 3.9%, growing 9, then 8, then 22, then 28 by year, with the market rather than with the automated deposits.

A peer-reviewed paper is a text that has passed an editor and referees at a journal with an ISSN. A preprint is a working paper posted to a repository, SSRN, arXiv, Zenodo, before or instead of that process. Of the 40 most-cited papers in the 1,710, 33 (82.5%) are peer-reviewed, though peer-reviewed journals are only 33.0% of the whole corpus (565 of 1,710). Average citations are 4.18 for a peer-reviewed paper against 0.41 for a preprint or repository entry, counting the whole set, which favours older papers that have had time to be cited.

The literature skews toward what is easy to measure from a price series: correlations, named events, regulation. DeFi has 243 of 1,710 titles (14.2%). Regulation, law and compliance has 242 (14.2%). Redemption mechanics have 6 papers, all 2025–2026, zero citations each. Admin-key, multisig and custody-risk words in the title: zero papers out of 1,710. That is not a finding that nobody studies those risks. It is a finding that the data a regressor would need, daily redemptions by channel, signer lists beyond a count, listing-promotion contracts, are not public.

Terra/UST has 27 papers with those names in the title. SVB/USDC has 7. The highest-cited paper in the set is Silicon Valley Bank bankruptcy and Stablecoins stability (International Review of Financial Analysis, 2023, 61 citations). The mechanics that decided whether those pegs returned, the open or closed loop, are the six uncited papers.

A practical cut for whether a record is worth opening: an author with a university or central-bank affiliation, a title that states a research question rather than a figure filename, and either a citation or a named publisher (SSRN under a university, NBER, IMF, a journal with an ISSN). Absence of peer review does not disqualify. SSRN and arXiv are ordinary early channels in economics. Absence of any method to check does.