The 2008 global financial crisis — complexity is not safety
The answer
Between 2007 and 2009, losses on American home loans destroyed several of the largest financial institutions in the world. Lehman Brothers filed for bankruptcy on 15 September 2008 with about $600 billion in assets, the largest bankruptcy in US history. The S&P 500 fell about 57% from its October 2007 peak to its March 2009 low. The Indian Sensex fell from over 20,000 in January 2008 to about 8,160 in March 2009, although no Indian bank failed.
Why this costs you money
Here is the specific error 2008 punished, and it is an error most investors are still making.
You judge a financial company by its profits and its growth. Both are reported clearly and both are easy to compare. Neither tells you what the company owns.
A lender's profits come from the difference between what it pays for money and what it charges for money. That difference gets larger when the lender takes more risk. So a lender showing unusually high profitability is telling you 1 of 2 things: it is unusually efficient, or it is lending to people other lenders will not lend to. Those look identical in the profit statement for several years.
The thing that separates them is the loan book: what the money was lent against, and who owes it. That is disclosed. It is in the annual report. Almost nobody in retail reads it, and it is the only place where the answer lives.
In 2008 this cost people everything, because the institutions that failed had reported record profits for years, and their shares had been recommended by almost everybody, and their bonds had been rated AAA — the highest rating that exists.
How it works
Follow the chain. Every link was sensible on its own.
1. A home loan is made. In the United States between roughly 2003 and 2006, lending standards loosened. Loans were made with little documentation of income, with low starting interest rates that reset higher after 2 or 3 years, and sometimes with no down payment. These are called subprime loans: loans to borrowers with weak credit records.
2. Loans are bundled. Thousands of loans were packaged into a mortgage-backed security. Buyers received payments as the homeowners paid their loans.
3. The bundle is sliced. The security was divided into layers, called tranches. The top layer got paid first and absorbed losses last. The bottom layer absorbed losses first. This is called a collateralised debt obligation, or CDO, when the process is repeated on the pieces.
4. The top slice is rated AAA. Rating agencies assigned their highest rating to senior tranches, on the reasoning that for the top slice to lose money, an implausibly large number of homeowners would have to default at the same time .
5. Institutions bought the AAA slices because their own rules required highly rated assets, and because AAA paper paid a slightly higher yield than government bonds.
Now find the assumption. Step 4 depended entirely on defaults being uncorrelated — on the idea that a homeowner defaulting in Florida told you nothing about a homeowner in Nevada. That is true in ordinary times, when defaults are caused by individual events: illness, job loss, divorce.
It is false when the cause is national. When US house prices fell nationally for the first time in decades, every borrower depending on refinancing or on selling at a higher price was in trouble at once. The defaults arrived together, which is precisely the scenario the AAA rating assumed away.
Two amplifiers made it systemic.
Leverage. Major investment banks operated at very high ratios of assets to equity, in the region of 30 to 1. At 30 to 1, a fall of a little over 3% in asset values erases the equity.
Credit default swaps, contracts that pay out if a bond defaults. They allowed institutions to take positions far larger than the underlying bonds in existence. AIG, an insurance company, had written a large volume of them and was rescued by the US government with an $85 billion facility on 16 September 2008.
The response: the Troubled Asset Relief Program of $700 billion authorised on 3 October 2008, emergency central bank lending, and eventually the Dodd-Frank Act on 21 July 2010.
What it tells you, and what it does not
What it tells you. A rating is somebody's opinion about a model. It is not a measurement. When you buy a rated instrument you are buying the model's assumptions, and those assumptions are usually stated in a document nobody reads.
It also tells you that complexity has a direction. Financial products get complicated for a reason, and the reason is often that complexity moves a risk from a place where it must be reported to a place where it does not. If you cannot understand a structure after a serious attempt, the conclusion is not "I am not clever enough". It is "this was not designed to be understood by me".
What it does not tell you. It does not tell you that all securitisation is bad. Bundling loans and selling them is a legitimate way to fund lending, and India uses it too. The problem in 2008 was that bundling broke the link between the person making the loan and the person bearing the loss. When the lender sells the loan the same week, its incentive to check the borrower disappears.
And it does not tell you that ratings are useless. Rated bonds default less often than unrated ones. The failure was concentrated in structured products, where the rating depended on a correlation assumption, not in ordinary corporate bonds. Knowing which kind you hold is the useful distinction.
The decision rule
For any lender you own, know 3 things: what the loans are secured against, how the lender is funded, and how long the loans run compared with the funding.
The third one is the killer and it has a name: asset-liability mismatch. A lender that borrows short-term money and lends it for 15 years is fine until the short-term money stops arriving. Then it fails, even if every single loan is good.
The conditional version: high growth in a lender is acceptable if the loan book is secured against assets that can be sold, the funding is stable, and provisions for bad loans are rising along with the book. It is a warning if growth is fast, the lending is unsecured or against collateral that cannot be valued independently, and provisions are flat.
Try this now
Fifteen minutes, and it is the most valuable reading exercise in this cluster. Almost every reader owns a lender, either directly or inside an index fund.
- Identify a bank, non-bank lender (NBFC), housing finance company or insurer in your holdings. If you hold an index fund, look up the fund's top holdings — financial companies are usually the largest weight in both Indian and US indices.
- Find that company's most recent annual report. It is on the company's website under Investor Relations. It is free.
- Find the section on the loan book. In India, look for "advances" and the notes giving the breakup by segment. In the United States, look in the 10-K for "loan portfolio composition".
- Write down the answers to 4 questions.
- What are the largest 3 categories of loans? Home loans, vehicle loans, loans against property, unsecured personal loans, gold loans, corporate loans, loans against shares.
- What proportion is unsecured? Unsecured means there is no asset to sell if the borrower stops paying.
- What is the gross non-performing assets (NPA) percentage, and is it rising or falling over the last 3 years?
- Where does the money come from? Deposits are stable. Short-term market borrowing is not. In India look for the mix of deposits, borrowings and commercial paper.
- One more, if you have the patience: find the provision coverage ratio, the proportion of bad loans already provided for.
What you should see. Most readers discover they had no idea. A company they think of as "a bank" turns out to have most of its book in 1 specific kind of lending, sometimes one that depends on a single asset class or industry.
The most common surprise is unsecured lending. A lender growing quickly in unsecured personal loans is earning high margins for a reason. That is not an argument to sell. It is the fact you must know before you can hold sensibly, because when the cycle turns, that part of the book goes first.
If you cannot find the breakup in 15 minutes, that is itself an answer about disclosure quality.
Three real cases
1. Lehman Brothers, 15 September 2008 (United States) — funding, not assets Lehman filed for bankruptcy with roughly $600 billion in assets. It did not fail because every asset was worthless. It failed because it funded long-term positions with very short-term borrowing in the repo market, and that borrowing stopped. The bankruptcy examiner later described accounting transactions known as Repo 105, which moved assets off the balance sheet at reporting dates . What it turned out to be: the clearest demonstration that a financial institution dies of funding, not of losses.
2. IL&FS, India, 2018 — the Indian version, 10 years later Infrastructure Leasing & Financial Services, a large Indian infrastructure lender, defaulted on debt in 2018 despite holding high credit ratings shortly before. The government superseded the board. The default caused a funding freeze across Indian NBFCs, because mutual funds that held NBFC commercial paper faced redemptions. What it turned out to be: an asset-liability mismatch of exactly the 2008 kind. Long-dated infrastructure assets funded by short-dated market borrowing, with ratings that did not reflect it.
3. Silicon Valley Bank, March 2023 (United States) — no bad loans at all SVB failed in March 2023. Its loans were largely fine. It had invested deposits in long-dated government bonds, whose market value fell as interest rates rose, while its deposits were unusually concentrated among technology companies who withdrew quickly. What it turned out to be: proof that the 2008 lesson is not about mortgages. It is about the relationship between how long your assets last and how quickly your funding can leave.
The question that resolves it
A novice looks at a lender and asks: is it growing, and is it profitable?
An expert asks: what is the collateral, and how long can the funding stay?
Growth and profit are outputs. Collateral and funding are the inputs that determine whether the outputs survive a bad year. Only 1 of those pairs is predictive.
What would make this wrong
The falsification condition. The claim is that lender failures are driven by funding structure and collateral quality, not by reported profitability. If a review of major financial failures showed that the failed institutions had been reporting weak profits beforehand, the claim would be wrong. In practice the opposite holds: Lehman, IL&FS, SVB and several Indian NBFCs were all reporting strong results shortly before failing. Reported profit had close to zero warning value.
The honest limit. Reading a loan book will not let you time a crisis. Analysts who correctly identified weak lending in 2005 were early by 3 years, and being early is expensive. This exercise reduces the chance that you are surprised. It does not tell you when to act.
The hindsight trap, and this one is badly abused. The 2008 crisis is usually told as a story with a small number of heroes who saw it coming, which implies it was visible. Look at what a careful investor genuinely faced in 2006.
US house prices had not fallen nationally, on an annual basis, since the 1930s . Every model was calibrated on that history, because that was the history. The Chairman of the Federal Reserve said in 2007 that the problems in subprime were likely to be contained. The rating agencies were long-established institutions with regulatory standing, and funds were legally required to rely on them for some purposes. The instruments were genuinely opaque: a CDO of CDOs could contain thousands of underlying loans, and the loan-level data was not available to ordinary investors.
Being right also required surviving being wrong. Investors who bet against subprime in 2006 lost money for over a year and faced redemptions, and several were forced to close positions before the payoff arrived.
So what could a careful person genuinely have seen?
Three things, all public.
First, the ratio of house prices to household incomes. That data was published by government agencies. It had moved to levels far above its long-run range in several US states. You did not need a mortgage model. You needed a division.
Second, the loan documentation standards. Lenders publicly advertised loans requiring no proof of income. Some issuers disclosed the proportion of low-documentation loans in their pools. A person who read a prospectus for a mortgage-backed security could see what was inside it.
Third, the leverage of the buyers. Investment bank leverage ratios were in public filings. A firm at 30 to 1 is disclosed as being at 30 to 1.
What a careful person could not have seen was the interconnection — who owed what to whom through credit default swaps. That information did not exist in any public place. Regulators did not have it either, which is why the Dodd-Frank Act required swaps to be reported to central repositories. Anybody claiming they could have mapped the contagion in 2007 is describing data that nobody had.
In India
No Indian bank failed in 2008, and understanding why is more useful than the crisis story itself.
Why India was insulated. Indian banks had very limited exposure to US mortgage securities. Capital account convertibility was restricted, so Indian institutions could not freely buy foreign assets. The RBI had raised risk weights and provisioning on commercial real estate and housing lending in 2006 and 2007, before the crisis. And Indian banks were funded mainly by domestic retail deposits, which do not run in response to events abroad.
How India was hit anyway. Through 3 channels. Foreign institutional investors sold Indian shares to raise cash at home. Indian exporters and IT services companies lost demand. And Indian companies that had borrowed abroad found refinancing unavailable. The Sensex fell from over 20,000 in January 2008 to about 8,160 by March 2009 — a larger percentage fall than the S&P 500 experienced, in a country whose banks were fine.
What India learned later. The 2018 IL&FS default showed the risk had moved from banks to non-bank lenders funded by mutual funds. The responses included the Insolvency and Bankruptcy Code of 2016, the RBI's asset quality review of bank loan books in 2015, and tighter liquidity requirements for NBFCs.
Where to look today. For any Indian lender: gross and net NPA percentages, provision coverage ratio, capital adequacy ratio, the proportion of unsecured lending, and the funding mix. All 5 are in the annual report and the quarterly investor presentation.
In the United States
What changed after 2008. The Dodd-Frank Act of 21 July 2010 created the Consumer Financial Protection Bureau, required central clearing for many derivatives, and introduced annual stress tests for large banks. The Volcker Rule limited proprietary trading by banks. Basel III raised capital requirements internationally.
What did not change. Ratings agencies are still paid by the issuers of the securities they rate. Dodd-Frank required regulators to remove references to credit ratings from rules, reducing the legal force of a rating without changing who pays for it.
Where to look today. For a US bank, the 10-K contains loan portfolio composition, allowance for credit losses, and the maturity profile of deposits and borrowings. Since 2023, unrealised losses on securities held to maturity deserve particular attention, because that is what SVB's disclosures showed before it failed.
Where they differ, and what that tells you
The difference is the identity of the depositor.
American bank deposits include very large corporate balances, which move quickly and are often above the insured limit. SVB failed in roughly 2 days because its depositors were concentrated, sophisticated, and connected to each other.
Indian bank deposits are dominated by household savings spread across a very large number of small accounts. Deposit insurance covers up to ₹5 lakh per depositor per bank, which covers the great majority of accounts in full. Those depositors are slow to move, and that slowness is a genuine source of stability.
What that tells you is where to look for fragility in each country. In the United States, examine deposit concentration: how much funding comes from a small number of large accounts, and how much is uninsured. In India, bank funding is comparatively stable, so the fragility has moved to the non-bank lenders, which have no deposits at all and fund themselves in the market.
That is what 2018 demonstrated. So an Indian investor looking for 2008 risk in their portfolio should not spend the time on large deposit-funded banks. They should spend it on any NBFC or housing finance company they own, and on the single question of how that company gets its money.
Carry this
- A rating is an opinion about a model, not a measurement of an asset.
- Lenders die of funding, not of losses.
- For every lender you own: what is it secured against, and how fast can the money leave?