The record · six times it broke, and how long the answer took

The record

Pipes: step 1 of four, Holding. World: Storm.

Step 1 of four — 3 ahead

History: six times the pipes broke and an answer came

These are not predictions. They are six real episodes, each date checkable against an official source, listed oldest to newest. Read them for one thing only: how long the gap was between the first visible trouble and the official answer — and how long until it was actually over.

Why six, and not three or twelve

Six is the smallest number that covers the range of speeds actually observed — from 5 days to 850 days — without teaching the same lesson twice.

Why not three: the three easiest to name (2007–08, 2020, 2022) all got an answer inside four months. Three cases would quietly teach that answers always come fast, and would hide the 124-day, 203-day and 850-day cases that make this dashboard's own forecast honest: 24.0% within 12 months, even odds only by March 2029 — a percentile, not a date.

Why not twelve: past six, a reader stops holding the cases in his head, the extra cases repeat mechanisms already shown, and older ones are harder to date to the month — which is exactly where invented history creeps in.

The six episodes, oldest to newest

1When every bank in the country shut for a week

United States
First troubleNovember 1930 — on 7 November the Bank of Tennessee in Nashville closed, the first of the Caldwell & Company chain to go; scores of banks followed as depositors pulled cash town by town
Official answer6 March 1933 — the national bank holiday, ordered by Proclamation 2039 at 1 a.m. that Monday. The Emergency Banking Act was passed on 9 March and it let the Reconstruction Finance Corporation buy bank stock with public money; banks reopened from 13 March
850 days from first trouble to the answer — the slowest on this page.
What it felt likeIf your bank shut, the money in it was simply gone — there was no government deposit insurance until the Banking Act of 1933 later that year. The downturn had already been running 43 months.
Until it was overThe runs stopped inside a week of the holiday and the banks reopened on 13 March 1933. The Depression itself did not end until the war.
Sources: Federal Reserve History, “Banking Panics of 1930–31” (Caldwell, 7 November; Bank of United States, 11 December) — federalreservehistory.org/essays/banking-panics-1930-31; “Bank Holiday of 1933” (1 a.m., 6 March, reopening 13 March) — federalreservehistory.org/essays/bank-holiday-of-1933; “Emergency Banking Act of 1933” (9 March; RFC stock purchase) — federalreservehistory.org/essays/emergency-banking-act-of-1933; “The Great Depression” (deposit insurance created in the 1933 reforms; bottom in March 1933; full recovery in the war) — federalreservehistory.org/essays/great-depression; NBER business-cycle table (contraction August 1929–March 1933, 43 months) — nber.org/research/data/us-business-cycle-expansions-and-contractions.

2Four banks failed in one month

Japan
First trouble26 November 1997 — the failure of Tokuyo City Bank was announced, the fourth failure that month after Sanyo Securities, Hokkaido Takushoku Bank and Yamaichi Securities. The Bank for International Settlements calls it “probably the day that Japan’s financial system was closest to a systemic collapse”
Official answer30 March 1998 — the first injection of public money into Japanese banks: ¥1.8 trillion of capital into the major banks, a thing that had never been done before
124 days from first trouble to the answer.
What it felt likeBank failures were on the news one after another that month; depositors queued, and no Japanese government had ever put public money into a bank.
Until it was overNot 124 days. Long-Term Credit Bank was nationalised on 23 October 1998, and a second injection four times larger — ¥7.5 trillion — came in March 1999, a year after the first.
Source: Nakaso, The financial crisis in Japan during the 1990s, BIS Papers No 6 (2001) — bis.org/publ/bppdf/bispap06.pdf. §1.3.2 (pp. 8–11) for the successive failures of Sanyo Securities, Hokkaido Takushoku Bank, Yamaichi Securities and Tokuyo City Bank in November 1997, the 26 November announcement and the “closest to a systemic collapse” line; §1.3.2.5, “The introduction of public funds and the first capital injection” (p. 11) for the ¥1.8 trillion injection of 30 March 1998; §1.4.1 (p. 12) for the nationalisation on 23 October 1998; §1.5.2 (p. 15) for the ¥7.5 trillion injection of March 1999, “more than four times as large” as March 1998.

3The funds nobody could put a price on

United States
First trouble9 August 2007 — BNP Paribas suspended three funds because it was “impossible to value certain assets” in US mortgage debt, and the market where banks lend to each other overnight froze. The junk-bond spread had gone from 2.41% on 1 June 2007 to 4.28% by 27 July
Official answer12 December 2007 — the Term Auction Facility (TAF), announced the same day by the Fed, the ECB, the Bank of England, the Bank of Canada and the Swiss National Bank. Then the Primary Dealer Credit Facility on 16 March 2008, and large-scale bond buying on 25 November 2008
125 days from first trouble to the first facility.
What it felt likeJobs went: the unemployment rate reached 10.2% in October 2009, and the central bank cut its policy rate all the way to zero.
Until it was overThe recession ran from December 2007 to June 2009 — 18 months, and 22 months after the first trouble. The junk-bond spread peaked at 21.82% on 15 December 2008 and was back to 6.39% on 31 December 2009.
Sources: BBC News, 9 August 2007, on the BNP Paribas funds and its statement that it was “impossible to value certain assets” — news.bbc.co.uk/1/hi/business/6938425.stm; Federal Reserve, TAF announcement, 12 December 2007 — federalreserve.gov/monetarypolicy/20071212a.htm; PDCF, 16 March 2008 — federalreserve.gov/newsevents/pressreleases/monetary20080316a.htm; bond purchases, 25 November 2008 — federalreserve.gov/newsevents/pressreleases/monetary20081125b.htm; NBER, 20 September 2010 (peak December 2007, trough June 2009) — nber.org/news/business-cycle-dating-committee-announcement-september-20-2010; BLS, October 2009 release (10.2%) — bls.gov/news.release/archives/empsit_11062009.htm; all spreads: ICE BofA US High Yield Index option-adjusted spread, series BAMLH0A0HYM2 — fred.stlouisfed.org/series/BAMLH0A0HYM2 (full daily history read from an archived copy of the series download: archived 3 Nov 2025 snapshot).

4The promise with no money behind it

Euro area
First troubleOctober 2009 — on 19 October the new Greek finance minister told the Eurogroup that the 2009 deficit could reach 12.5% of GDP, against the 3.7% the previous government had notified in the spring. Greek borrowing costs over Germany’s then climbed for two years
Official answer10 May 2010 — the Securities Markets Programme (SMP), the ECB’s first purchase programme for government bonds. Then two three-year bank loans (LTROs) announced on 8 December 2011 and first paid out on 21 December 2011; and Outright Monetary Transactions (OMT) announced on 6 September 2012
203 days from first trouble to the first facility; 1,053 days to the answer that actually ended it.
What it felt likeA government that had just admitted its published figures were wrong was borrowing at rates that kept climbing, and the money lent to it came with conditions attached — a memorandum of budget cuts and reforms signed in exchange for the loans.
Until it was overBorrowing costs turned after Mario Draghi’s “whatever it takes” remarks on 26 July 2012 and the OMT announcement on 6 September 2012 — 35 months after the first trouble. The lending did not stop there: a third Greek programme was signed only in August 2015.
Sources: European Commission, A new chapter for Greece — timeline of events (19 October 2009, 12.5% of GDP; loans of 2 May 2010; second programme of 14 March 2012; third programme signed 20 August 2015) — economy-finance.ec.europa.eu/system/files/2018-08/05_timeline_of_events_2.pdf; European Commission, Report on Greek government deficit and debt statistics, COM(2010) 1 final, 8 January 2010 (the 3.7% to 12.5% revision) — ec.europa.eu/eurostat/…/COM_2010_report_greek; ECB, 10 May 2010 — ecb.europa.eu/press/pr/date/2010/html/pr100510.en.html; ECB, 8 December 2011 — ecb.europa.eu/press/pr/date/2011/html/pr111208_1.en.html; Draghi verbatim, 26 July 2012 — ecb.europa.eu/press/key/date/2012/html/sp120726.en.html; OMT, 6 September 2012 — ecb.europa.eu/press/pr/date/2012/html/pr120906_1.en.html.

5The fortnight that did not become 2008

United States
First trouble9 March 2020 — the S&P 500 fell 7% in the opening minutes and trading was halted market-wide for 15 minutes; the junk-bond spread went from 3.38% on 20 January 2020 to 10.87% on 23 March
Official answer17 March 2020 — the Commercial Paper Funding Facility (CPFF) and the Primary Dealer Credit Facility; then on 23 March 2020 the open-ended promise to buy Treasury and mortgage bonds “in the amounts needed”
8 days from first trouble to the facilities; 14 days to the open-ended bond buying.
What it felt likeWhole industries shut within days and the unemployment rate went to 14.7% in April 2020.
Until it was overTwo months: the recession ran from February to April 2020. The next problem was inflation, not deflation.
Sources: Reuters, 9 March 2020, on the 7% fall and the 15-minute market-wide trading halt — reuters.com/graphics/USA-MARKETS/0100B5L144C (archived copy: web.archive.org/web/20261011033602/reuters.com/graphics/USA-MARKETS/0100B5L144C); Federal Reserve, CPFF, 17 March 2020 — federalreserve.gov/newsevents/pressreleases/monetary20200317a.htm; PDCF, 17 March 2020 — federalreserve.gov/newsevents/pressreleases/monetary20200317b.htm; FOMC statement, 23 March 2020 — federalreserve.gov/newsevents/pressreleases/monetary20200323a.htm; BLS, April 2020 release (14.7%) — bls.gov/news.release/archives/empsit_05082020.htm; NBER (peak February 2020, trough April 2020) — nber.org/research/business-cycle-dating; spreads via ICE BofA BAMLH0A0HYM2 as above.

6The pension funds that nearly blew up in five days

United Kingdom
First trouble23 September 2022 — the government’s fiscal statement of unfunded tax cuts. By 28 September the yield on 30-year gilts had risen 130 basis points in three trading days, “a move three times larger than any other historical move over a similar period”, and the funds that hold pension money were being forced to sell
Official answer28 September 2022 — the Bank of England began temporary purchases of long-dated government bonds “on whatever scale is necessary”, fully indemnified by the Treasury
5 days from first trouble to the answer — the fastest on this page.
What it felt likeThe schemes that pay retired people — the safest money in the country — were hours away from having to sell everything at once, into a market with no buyers.
Until it was over21 days of buying: the operations ran from 28 September to 14 October 2022 and bought £19.3 billion, and the last of it was sold back by 12 January 2023. No bank failed.
Sources: Bank of England news release, 28 September 2022 — bankofengland.co.uk/news/2022/september/bank-of-england-announces-gilt-market-operation; Bank of England Quarterly Bulletin, “Financial stability buy/sell tools: a gilt market case study”, 20 November 2023 (130bp in three days; three times larger than any other move; the 23 September fiscal statement; £19.3bn bought 28 September–14 October 2022; unwound by 12 January 2023) — bankofengland.co.uk/quarterly-bulletin/2023/2023/financial-stability-buy-sell-tools-a-gilt-market-case-study.

And sometimes nothing comes

Every episode above ends with a dated announcement, because that is what gets written down. But an announcement is not guaranteed. On this dashboard’s own count, a stress line was past its firing level in 163 of 287 month-ends since 2003 — breaks are common — and an official answer eventually arrived in 6 of 10 armed episodes. Four in ten, it did not. Two of those are real, datable, and had no facility at all:

The corporate default wave, 2000–03. Measured month by month, the US junk-bond spread was above 5% — the level this dashboard treats as a break for that line — at every single month-end from February 2000 to September 2003 — 44 months in a row, peaking at 10.59% on 31 October 2002. It came back below 5% on 31 October 2003, at 4.73%. There is no facility to name for those 44 months, because none was created.

The energy and commodity squeeze, 2015–16. The same spread was above 5% at every month-end from July 2015 to August 2016 — 14 months in a row, peaking at 8.87% on 11 February 2016. Nothing was announced. The Fed went the other way: it raised its policy rate on 16 December 2015, while the spread was still above the level.

What made those two different from the six? What was broken. In all six answered episodes a payment pipe had stopped working — banks that could not pay depositors, banks that would not lend to each other overnight, a bond market with no buyers. A central bank has a tool that fits that exactly: lend against collateral, or get a law passed that guarantees the deposits. In 2000–03 and 2015–16 nothing in the payment pipes stopped working. The loss fell on people who had lent money to ordinary companies — telecoms and dot-coms in the first case, oil and mining companies in the second — and there is no facility that makes a company able to repay. Defaults, restructurings and time did the work instead.

Both of these did eventually end. They ended without an announcement, and “eventually” is doing real work in that sentence: 44 months and 14 months of the stress line sitting above its firing level.

One axis: how long the wait was

Same kind of event, response times 170 times apart. The number beside each row is the number; the bar is only that same number drawn to scale.

5 days — UK 2022

8 days — US 2020

124 days — Japan 1997

125 days — US 2007

203 days — Euro area 2009

850 days — US 1930

How the bars are drawn: width in per cent = days ÷ 850 × 100, rounded to the nearest whole per cent, with a floor of 3% so that the 5-day and 8-day rows stay visible. 5 days = 1% → 3%; 8 days = 1% → 3%; 124 days = 15%; 125 days = 15%; 203 days = 24%; 850 days = 100%.

Why the speeds differ

Three things decide it. First, what is broken. A funding market — the market for overnight loans between banks, or for government bills — can be backstopped in days by a central bank that already has the authority; a bank that is insolvent, or a government that cannot pay, cannot be fixed by lending alone. Second, whether the authority already exists. In 2007, 2008 and 2020 the Fed used powers it already had as lender of last resort. In 1933 the answer needed a law from Congress and a president who had been in office 36 hours; in Japan it needed public money put into banks for the first time in the country’s history, and the first attempt was too small. Third, whether the currency itself is the problem. In the euro area there was no single government standing behind the bonds, so the answer that finally worked had to wait for a political settlement — 1,053 days.

Notice also the second gap, between the first facility and the one that actually ended the episode. In 2007–08 the first facility came in 125 days and the bond buying that worked in 474; in the euro area, 203 days to the first purchase programme and 1,053 to the one that turned it; in Japan, 124 days to the first injection and 331 to the nationalisation, with a second injection four times larger a year after the first. The first announcement is not the finish line, and a reader who takes it as one will be wrong three times out of the six on this page.

The honest limit of historical analogy

These six all have an answer because history keeps the episodes that ended and files the others under “nothing happened”. So the six are not a probability: the correct base rate is the dashboard’s own finding, 6 of 10 armed episodes, not 6 of 6. Analogy also points the wrong way exactly when it feels most useful. The 2008 template said a credit bust means years of deflation and deleveraging; applied to March 2020 it predicted a long slump, and the recession was over in two months — the next problem was inflation. The 2010–12 template said a currency union without a fiscal union must break up; instead borrowing costs turned within days of a speech on 26 July 2012, and the programme that finally stopped the run — announced on 6 September 2012 — came with a condition that had been the obstacle all along: the country had to be inside a bailout programme first. What history is good for here is narrow and real: it shows the range of waiting times, it shows that the first answer is often not the last, and it shows what an answer looks like when it arrives. It does not tell you the date.

A popular story that is wrong

The claim: “The Fed pivoted in January 2019 with nothing broken.”

The data: the US junk-bond spread was 3.16% on 3 October 2018, 5.38% on 27 December 2018 and 5.44% on 3 January 2019 — above the firing level this dashboard uses for that line. It had come back to 4.37% by 31 January 2019. The “patient” language in the FOMC statement is dated 30 January 2019 — after the break, not before it.

Source: ICE BofA US High Yield Index option-adjusted spread (series BAMLH0A0HYM2), daily closes, all four prints read from the series history — fred.stlouisfed.org/series/BAMLH0A0HYM2; FOMC statement, 30 January 2019 — federalreserve.gov/newsevents/pressreleases/monetary20190130a.htm.

Sources for every date on this page

Federal Reserve History, “Banking Panics of 1930–31” — federalreservehistory.org/essays/banking-panics-1930-31
Federal Reserve History, “Bank Holiday of 1933” — federalreservehistory.org/essays/bank-holiday-of-1933
Federal Reserve History, “Emergency Banking Act of 1933” — federalreservehistory.org/essays/emergency-banking-act-of-1933
Federal Reserve History, “The Great Depression” — federalreservehistory.org/essays/great-depression
NBER, US Business Cycle Expansions and Contractions — nber.org/research/data/us-business-cycle-expansions-and-contractions
NBER, Business Cycle Dating chronology — nber.org/research/business-cycle-dating
NBER, announcement of 20 September 2010 (trough June 2009) — nber.org/news/business-cycle-dating-committee-announcement-september-20-2010
Nakaso, The financial crisis in Japan during the 1990s, BIS Papers No 6 — bis.org/publ/bppdf/bispap06.pdf
BBC News, 9 August 2007 (BNP Paribas funds) — news.bbc.co.uk/1/hi/business/6938425.stm
Federal Reserve, TAF, 12 December 2007 — federalreserve.gov/monetarypolicy/20071212a.htm
Federal Reserve, PDCF, 16 March 2008 — federalreserve.gov/newsevents/pressreleases/monetary20080316a.htm
Federal Reserve, bond purchases, 25 November 2008 — federalreserve.gov/newsevents/pressreleases/monetary20081125b.htm
Federal Reserve, CPFF, 17 March 2020 — federalreserve.gov/newsevents/pressreleases/monetary20200317a.htm
Federal Reserve, PDCF, 17 March 2020 — federalreserve.gov/newsevents/pressreleases/monetary20200317b.htm
Federal Reserve, FOMC statement, 23 March 2020 — federalreserve.gov/newsevents/pressreleases/monetary20200323a.htm
Federal Reserve, FOMC statement, 16 December 2015 (rate raised) — federalreserve.gov/newsevents/pressreleases/monetary20151216a.htm
Federal Reserve, FOMC statement, 30 January 2019 (“patient”) — federalreserve.gov/newsevents/pressreleases/monetary20190130a.htm
ECB, Securities Markets Programme, 10 May 2010 — ecb.europa.eu/press/pr/date/2010/html/pr100510.en.html
ECB, three-year LTROs, 8 December 2011 — ecb.europa.eu/press/pr/date/2011/html/pr111208_1.en.html
ECB, Draghi verbatim, 26 July 2012 — ecb.europa.eu/press/key/date/2012/html/sp120726.en.html
ECB, technical features of OMT, 6 September 2012 — ecb.europa.eu/press/pr/date/2012/html/pr120906_1.en.html
European Commission, A new chapter for Greece — timeline of events — economy-finance.ec.europa.eu/system/files/2018-08/05_timeline_of_events_2.pdf
European Commission, Report on Greek government deficit and debt statistics, COM(2010) 1 final — ec.europa.eu/eurostat/…/COM_2010_report_greek
Bank of England, gilt market operation, 28 September 2022 — bankofengland.co.uk/news/2022/september/bank-of-england-announces-gilt-market-operation
Bank of England, Quarterly Bulletin, gilt market case study, 20 November 2023 — bankofengland.co.uk/quarterly-bulletin/2023/2023/financial-stability-buy-sell-tools-a-gilt-market-case-study
Bureau of Labor Statistics, October 2009 release — bls.gov/news.release/archives/empsit_11062009.htm
Bureau of Labor Statistics, April 2020 release — bls.gov/news.release/archives/empsit_05082020.htm
Reuters, 9 March 2020 (7% fall, 15-minute market-wide halt) — reuters.com/graphics/USA-MARKETS/0100B5L144C
ICE BofA US High Yield Index option-adjusted spread, series BAMLH0A0HYM2 — fred.stlouisfed.org/series/BAMLH0A0HYM2 — full daily history read from the archived series download of 3 November 2025 — web.archive.org/web/20251103031826/https://fred.stlouisfed.org/graph/fredgraph.csv?id=BAMLH0A0HYM2

One day-level note, not a month-level one: the Commission’s own timeline gives 19 October 2009 for the Greek deficit revision; the finance minister’s press statement was the next day, 20 October. The month is not in doubt and the 203-day gap moves by one day either way. Everything else on this page is pinned to an exact day by the source named beside it.