The argument this series makes, in 47 steps. It is the same spine every module is built on, and no module states more of it than its own share.
American house prices had never fallen across the whole country in the post-war period, and that assumption was priced into every loan written against them.
Prices peaked in 2006 and then fell by 27.4% before reaching bottom in February 2012, so the assumption in link 1 was wrong.
The defaults came first and fastest where prices had risen fastest, so the loss was concentrated rather than spread thinly.
The lending written against those prices was increasingly to borrowers who would not have qualified before: by 2006 it was between a fifth and a quarter of all new American mortgage lending.
Those loans were written to be refinanced rather than repaid, so they only worked while the price kept rising, which link 2 says it stopped doing.
The money funding all of this came from outside America: countries running persistent surpluses were hunting for safe dollar assets, and America took the money in and sent it back out again.
A pool of mortgages was turned into a bond and sold on, so no lender in the chain was left holding the loan it had made.
Slicing the pool into tranches let most of a pile of risky loans be sold as low-risk, and that arithmetic only held while the loans failed independently of each other, which link 3 says they did not.
The rating agencies stamped these bonds triple-A at a rate of tens per working day and were paid by the firms whose products they graded.
When the loans failed together, the losses marked against these bonds ran to trillions of dollars, which is why a quarter off American house prices was enough to matter to the whole world.
The firms holding these bonds were funding thirty-year assets with borrowing that had to be renewed every morning, outside the banks and outside deposit insurance.
Those firms were running on so little equity that a small fall in asset values wiped them out.
The run was not depositors queueing: lenders demanded far more collateral for the same loan, and against the worst of the bonds they would not lend at any price.
Money market funds and commercial paper were where ordinary people's cash met that run, and guaranteeing them was a separate emergency of its own in September 2008.
No agency was watching the system as a whole: responsibility was split between agencies, and supervisors judged banks one at a time with no measure of what the banks were doing to each other.
British banks had lent far more than their savers had given them and made up the difference by borrowing short-term in the same international wholesale markets link 13 had just shut.
Northern Rock took only about a fifth of its money from savers, so when the market it borrowed the rest in dislocated on 9 August 2007 it could not refinance itself.
British banks were hurt far less by owning the American bonds than by borrowing in the market it poisoned.
The shortage was a dollar shortage happening outside America, because European and British banks had built up dollar assets they could not fund in dollars.
Britain's answer to Northern Rock was slow because the job was split three ways, between the Treasury, the Bank and the FSA.
The authorities rescued Bear Stearns in March 2008 and took the two government-backed mortgage companies into conservatorship on 7 September 2008, so by that autumn the market believed big firms would be caught.
Lehman Brothers filed for bankruptcy on 15 September 2008 with $639bn of assets and no buyer found over the weekend, and what separated Lehman from Bear was not size but that it was let go.
The next day the Federal Reserve lent to an insurance company it did not regulate, and central banks became the lender of last resort for firms outside the banking system.
Because the shortage was a dollar shortage happening abroad, the Federal Reserve lent dollars to other central banks so that they could lend them on.
The Bank of England lent against the mortgage assets the market would not buy, and it took nearly four years to unwind.
What actually stopped the runs was a promise rather than a loan: guarantee the deposits and the bank debt, and the queue goes away.
On 8 October 2008 Britain put public money into the banks as capital rather than lending it, and took ordinary shares - actual ownership - where most other governments bought instruments closer to a loan.
The state took 84% of RBS for £45.5bn and 43% of Lloyds for £20.3bn, and the two rescues were not the same rescue.
At the peak the British state had £1,162 billion of support to the banks outstanding, and all but £133 billion of it was promises rather than cash paid out.
Britain froze an Icelandic bank's assets using anti-terrorism law, because £4.5bn of British retail money was inside it.
America's own version was TARP, which paid out $443.5bn and cost $31.1bn over its life, neither the profit nor the disaster it gets described as.
Alongside the rescues the Bank of England cut Bank Rate nine times in fifteen months, from 5.75% to 0.5%.
When Bank Rate reached 0.5% and could go no lower, the Bank started buying government bonds instead, and had bought £895bn of them by the time it stopped.
Bank lending to British companies swung from rapid growth into outright contraction, and unsecured lending to households shrank for the first time on record.
This was banks lending less, not only firms borrowing less: the Bank of England's own estimate is that credit supply shocks took 8.3 percentage points off lending growth in 2009.
The American recession that followed ran eighteen months, the longest since the second world war, and American unemployment peaked at 10.0%.
British output fell about 6.3% from peak to trough, and it took twenty-one or twenty-two quarters to get back, depending which vintage of the national accounts you read.
Britain's unemployment went to 8.0%, well short of America's 10.0%, because Britain took the hit in pay instead.
The trade was measurable: Britain lost 6% of its output but only 1.9% of its jobs, where the smaller recession of the 1990s lost 2.5% of output and 3.4% of jobs, and the difference came out of pay.
Because firms kept staff on, the loss showed up as shorter hours and a collapse in output per hour, and British productivity never returned to its pre-crisis trend.
Lloyds was out first: the government finished selling in May 2017, having sold at 76.8p against 73.6p paid, a gain on one measure and a loss on another.
RBS took seventeen years: the last NatWest shares were sold on 30 May 2025, returning £34.98bn against the £45.5bn put in.
The Icesave money came back too: the state paid out £4.5bn and recovered £4.6bn from the Landsbanki estate by January 2016.
What the whole rescue cost depends on whether the interest on the borrowing behind it is counted: the cash came back with a surplus, and the borrowing cost more than the surplus.
Banks must now hold more than three times the equity they held before.
The second reform was about failure rather than capital: a way to shut a large bank without rescuing it, so that link 22's choice does not have to be made again.
And supervisors now test banks every year against an imagined crisis, rather than checking last year's books.