Subprime: the fuel
The crisis was built on the American mortgage, and on a single load-bearing assumption that turned out not to bear any load at all: that house prices would keep rising. If that held, a borrower who could not actually afford the payments could refinance or sell and walk away whole. If it failed, everything downstream failed with it. Throughout the early 2000s, lenders stretched their standards past recognition. Loans with no income documentation, no down payment, adjustable rates scheduled to reset into numbers the borrower had no plausible way of meeting — these were written by the million and marketed, often quite aggressively, to people who were being set up to lose their homes. I watched some of this from the inside of a Hampton High classroom; the kids whose families took the second mortgage were not the kids whose parents read the prospectus.
The deeper problem was that the bank originating the loan did not hold the loan. Through securitization, thousands of individual mortgages were bundled into Mortgage-Backed Securities and CDOs and sold off to investors around the world. The bank kept the origination fee; somebody in Dusseldorf or Reykjavik kept the risk. The incentive structure this produced was exactly the one a saboteur would design if the goal were to maximize the volume of bad loans written. Volume paid. Quality was somebody else’s problem. One could argue the entire arc of the crisis is contained in those two sentences.
Leverage and the accelerant
If subprime was the fuel, leverage was the accelerant. Bear Stearns, Lehman Brothers, Merrill Lynch, and their peers were running balance sheets at 30:1 or higher. At 30:1, a 3.3% decline in asset values wipes out the equity entirely. There is no margin. There is no cushion. There is, for a firm at that level of leverage, a single bad quarter between being an investment bank and being a bankruptcy filing.
Alongside leverage sat the Credit Default Swap, the CDS — a contract that let you collect insurance on a mortgage bond you did not own. AIG, the insurer, had written hundreds of billions of dollars of these. When mortgage bonds began to default in volume, the claims came in faster than AIG could pay them, and because the counterparties to AIG were themselves the largest banks in the world, a single insurance company turned into a systemic fuse. The interconnections were dense enough that the failure of any one large institution threatened to topple the rest. This is not a metaphor. This is how the plumbing was actually built.
2007: the first cracks
The early warnings came in the spring and summer of 2007. Two Bear Stearns hedge funds, heavily exposed to subprime paper, collapsed in June. Investors in those funds lost essentially everything. The broader market shrugged and kept rallying, and the S&P 500 made its all-time high in October 2007 — a peak that would stand for six years. In hindsight the late-2007 rally looks like the last mile of a long-distance race where the runner has already gone anaerobic and just doesn’t know it yet. The legs are still moving. The oxygen debt is catastrophic. The split that registers afterward is unrecognizable from the split the runner believed they were running.
March 2008: Bear Stearns
Bear Stearns went in March. The mechanism was a classic bank run, modernized: clients and counterparties pulled their short-term funding inside of a single week, and an institution that had looked solvent on a Monday was effectively insolvent by the following Monday. The Federal Reserve orchestrated an emergency sale to JPMorgan Chase at $10 a share, down from $170 a year earlier. Shareholders lost 94% of their investment. The institutional memory this created inside Treasury and inside the Fed matters; by the time Lehman’s moment arrived in September, the people in those rooms had already walked through a version of the scenario once. The muscle memory was there. Whether it helped is a separate question, and the honest answer is that the muscle memory did not save Lehman. It saved AIG, three days later, and possibly the entire investment-grade corporate bond market in the weeks after that.
September 2008: the month the rhythm shattered
September 2008 was the most compressed month in American financial history since October 1929. If most of policy life moves at the rhythm of a Miles Davis ballad — deliberate, patient, with space between the notes — September 2008 was Coltrane’s “Giant Steps”: the chord changes coming faster than any reasonable improviser can track them, the band scrambling to keep the form intact while the form itself accelerates. The sequence was brutal. On September 7, the federal government placed Fannie Mae and Freddie Mac into conservatorship, absorbing the guarantors of roughly half of all American mortgages; combined losses eventually exceeded $100 billion. On September 15, Lehman Brothers filed for bankruptcy — $639 billion in assets, the largest bankruptcy filing in American history. The decision not to rescue Lehman has been debated inside policy circles ever since, and the debate is not settled. What is not debated is the effect: overnight, every institution in the world began asking whether its counterparties would be alive in the morning.
On September 16, AIG received an emergency $85 billion loan from the Federal Reserve after its CDS book threatened the chain-reaction failure the firm’s counterparties had quietly understood was possible. The loan grew to $182 billion. On September 19, Treasury Secretary Henry Paulson and Fed Chairman Ben Bernanke went to Congress and used the phrase “complete meltdown” — a phrase chosen, I think, to be heard. And on September 29, after Congress initially rejected the $700 billion TARP package, the Dow fell 778 points, the largest single-day point decline to that date. The vote was reversed four days later.
Bernanke, writing about this stretch in The Courage to Act, is as measured as a central banker gets about it. The prose is calm. The subject is not.
The bottom: October 2008 through March 2009
The decline ran longer than most traders expected. The Dow peaked in October 2007 at 14,164 and bottomed at 6,547 on March 9, 2009 — a decline of 54% across seventeen months. The S&P 500 fell from 1,565 to 676, a 57% drawdown. Trillions of dollars in retirement accounts, pension funds, and personal savings were wiped out. In Hampton and in thousands of towns like it, the number that mattered was not the index print; it was the foreclosure notice. My father played poker on Fridays with a neighbor named Monroe. By late 2009 the Monroes were gone. I don’t know where. The dry cleaner the Patel family had run on Mercury Boulevard stayed open another three years on momentum and a relative’s small infusion before closing for good.
What diversification did, and did not, do
One of the durable lessons from the 2008 drawdown is that diversification across individual stocks provides almost no protection in a systemic crisis, because in a systemic crisis, all stocks fall together. Correlations converge to one. The conventional wisdom about holding thirty names to reduce idiosyncratic risk continues to be correct in normal markets and continues to be insufficient when the market itself is the risk. What did work, for traders who survived the period in usable shape, was position sizing. A book in which no individual position exceeds roughly 6% to 7% of equity can absorb a 60% to 70% drawdown on its worst name and still take less than 5% off the total. The discipline matters more than the stock picks. That sentence is not new; what 2008 did was move it from the “reasonable advice” column to the “the only thing that mattered” column for a generation of traders who lived through it.
What I think a trader should take from this
Leverage is the common thread across every financial crisis I have studied. The 1929 crash with its 10% margin accounts; the dot-com bust with equity options functioning as corporate currency; 2008 with 30:1 investment-bank balance sheets. The instrument changes; the mechanism does not. For the individual trader the lesson translates cleanly: margin converts a manageable drawdown into an account-ending event, and it does so on a timeline measured in days, not quarters.
Complexity hides risk. The CDOs and synthetic CDOs that detonated in 2008 were understood in full by almost nobody, including the people selling them. Baldwin wrote, in a different context, that the great force of history comes from the fact that we carry it within us. The force of a bad CDO came from the fact that nobody carried its structure in their head clearly enough to price it. If you cannot assess the risk of an instrument, you are not investing. You are guessing with extra steps.
Cash is what buys the recovery. Traders who reached October 2008 with real cash reserves had the kind of opportunity that arrives perhaps twice in a career: blue-chip franchises at 50% to 60% discounts to their pre-crisis prices. Warren Buffett deployed roughly $26 billion across the crisis window and harvested billions in subsequent profit. Cash is not a defensive crouch. It is dry powder.
Markets recover. The S&P bottomed at 676 in March 2009 and made new all-time highs by 2013. Investors who sold at the bottom locked in their losses. Investors who held — or better, bought — were eventually made whole, and then more than whole. This is the hardest thing to internalize in real time, because in March 2009 selling did not feel like panic. It felt like prudence. Honestly, I am not sure I would have held in March 2009 had I been managing money instead of finishing my freshman year at Howard. The kids in the dorm whose parents had been wiped out were not in any mood to hear about long-run mean reversion, and I am not certain the long-run argument would have survived a phone call home.
See also: The Crash of 1929 · The Dot-Com Bubble · Risk Management · George Soros
What the government did
The scale of the 2008-2009 intervention was without American precedent. TARP, the $700 billion Troubled Asset Relief Program, was originally designed to buy distressed mortgage assets and was rapidly repurposed to inject capital directly into banks. On October 13, 2008, the nine largest banks received $125 billion at a single meeting — Paulson essentially required them all to accept the funds so that no single institution would be stigmatized by needing a bailout. This detail matters more than it usually gets credit for; the design choice was about stigma, not about economics. Stigma is a price on a balance sheet that has no GAAP line, and the Treasury team understood this in a way the academic literature did not catch up to for several years.
The Federal Reserve built a series of emergency facilities that extended trillions of dollars in short-term loans to banks, money market funds, and even foreign central banks via swap lines. The Fed’s balance sheet expanded from roughly $900 billion to over $2 trillion — an increase that would have been considered unimaginable in 2006 and that became merely the opening act for the expansions that followed COVID. The federal funds rate was cut from 5.25% to effectively zero between September 2007 and December 2008, the fastest cutting cycle on record. The FDIC raised the deposit insurance ceiling from $100,000 to $250,000 and temporarily guaranteed new senior bank debt. The American Recovery and Reinvestment Act of February 2009 added $787 billion in spending and tax cuts.
The full cost of the response is debated. Estimates run from $1 trillion to $12 trillion depending on what one counts. Most of the direct bank bailout funds were repaid with interest. The lost jobs, the foreclosed homes on my block and ten thousand others, the reduced lifetime earnings for people who entered the labor market between 2009 and 2012 — those are harder to tally and do not appear on the Treasury’s balance sheet. There is a real cost-benefit accounting that includes those numbers, and there is the cost-benefit accounting that the published policy literature uses, and the gap between them annoys me more than I usually admit. The economists who wrote the postmortems were not the ones who lost their houses, and the ones who lost their houses did not write the postmortems.
What the reforms did, and did not, do
The crisis produced the largest body of financial regulation since the 1930s. Dodd-Frank, passed in 2010, overhauled the architecture: new capital requirements, derivatives clearing rules, the creation of the Consumer Financial Protection Bureau, and a resolution framework intended to wind down failing institutions without taxpayer money. Basel III, the international standard, roughly tripled required bank capital against assets. The Volcker Rule barred banks from proprietary trading, a partial return to the Glass-Steagall division that had been repealed in 1999.
Whether any of this is enough is genuinely unresolved. I sat in policy conversations at Treasury during another crisis, a different one, and the honest reading of the evidence is that leverage, complexity, misaligned incentives, and the human oscillation between euphoria and panic cannot be regulated out of the system. They can be managed. The quality of that management changes with political turnover and with the fading of memory. The 2023 regional-bank episode — SVB, Signature, First Republic — showed how quickly the old mechanisms can reassert themselves in new packaging, and how much of the post-2008 apparatus is there to contain the echoes rather than prevent the originals.
What remains, fifteen years out, is uneven. The banks are larger and better capitalized. The neighborhoods that lost houses in 2009 did not all recover; some of the streets in Hampton I walked as a teenager have a different demographic profile today, and the change is not because anyone in those houses got rich. One reading of the 2008 crisis is a story about the architecture of the financial system. Another reading is about which blocks of which towns got the notices taped to their doors. Both are true, and a trader who takes only the first one home has not read the whole book.
