When the Machine Breaks

Late on Tuesday 27 September 2022, officials at the Bank of England took calls warning that by the following afternoon, funds standing behind a large slice of Britain's pension system would be insolvent. Not because they had punted on crypto or emerging-market debt, but because they owned gilts, the safest sterling asset in existence, and gilts were in freefall. The next day the Bank, mid-inflation-fight and days from beginning quantitative tightening, announced it would buy long-dated gilts in whatever size it took. The system nearly died on a Tuesday. This module is about that night, and two other moments when the machine from Modules 4 and 5 stopped working.

September 2019, New York. On an ordinary Tuesday, the interest rate on overnight repo, the collateralised lending that Module 11 called the system's heart, spiked from around 2% to 10%. No war, no default; corporate tax payments and a Treasury settlement had drained bank reserves on the same day, and the "ample" reserves left over turned out not to be ample at all. The fed funds rate broke above the Fed's own target band, the one price it exists to control. Within weeks the Fed was injecting hundreds of billions and buying $60 billion of T-bills a month, while insisting, in Powell's words, that this should in no way be confused with QE. The balance sheet had been shrinking for two years. It never got back to where it started.

March 2023, Santa Clara. Silicon Valley Bank did what the QE decade taught banks to do: it parked a flood of deposits in long-dated bonds at generational-low yields. When rates rose, those bonds sank. Roughly $15 billion of unrealised losses, about the size of its equity. Over 90% of its deposits were uninsured, held by startups that all knew each other. When the hole became public, the run happened at Twitter speed: $42 billion attempted withdrawals in a day, over $100 billion queued for the next. The response: a systemic-risk exception guaranteeing uninsured depositors, and the BTFP, a facility lending against underwater bonds at face value. Mark-to-market, suspended by decree, the moment marking to market mattered.

September–October 2022, London: the centrepiece. The mini-budget of 23 September promised £45 billion of unfunded tax cuts, and gilt yields rose faster than at any time on record. Defined-benefit pension funds ran leveraged LDI strategies, using repo and derivatives on gilts to stretch their assets across their liabilities. Falling gilt prices triggered margin calls; the funds sold gilts to raise cash; the sales pushed prices down further; further falls triggered further calls. A doom loop in the safest asset in sterling finance. The Bank of England, which three days earlier had confirmed plans to sell gilts under QT, was forced to buy them instead, announcing up to £65 billion of purchases to break the spiral, while Governor Bailey publicly gave the funds three days to sort themselves out.

Three breaks, one lesson: the machine cannot be allowed to clear. In each case the market found a true price (for reserves, for deposits, for gilts) and in each case the true price was declared unacceptable within days, or hours. The interventions worked; that is the honest part. The Bank of England spent only £19 billion and calmed the gilt market; BTFP stopped the contagion; the repo facility ended the spikes. But every rescue writes a promise into the machine's expectations. Leverage that gets caught learns to lean harder; each backstop widens the one that will be needed next. A system that cannot clear cannot price risk. It can only accumulate it, and forward it, with interest, to the next Tuesday.