*The crisis was built on an assumption so reasonable that almost nobody examined it: that house prices in different parts of a large country would not all fall at once.*
The financial crisis of 2008 has been examined more thoroughly than almost any event in economic history, and the accounts of it are complicated. Beneath the complexity lies a single assumption, and understanding it is worth more to an ordinary investor than any quantity of detail about the instruments involved.
The assumption was this: that house prices in different regions of the United States were not strongly related to one another. If prices fell in one state, this told you little about what would happen in another. The evidence for this appeared solid, because it was drawn from decades of historical data during which regional housing markets had indeed behaved somewhat independently. There had been local declines, but there had been no nationwide fall in house prices in the period the data covered.
This assumption was load-bearing in a way that is difficult to overstate. It permitted mortgages from many different regions to be pooled together, on the reasoning that the pool would be far safer than any individual loan, since it was implausible that borrowers everywhere would default simultaneously. The pools could then be divided into portions of varying risk, and the safest portions could be assigned the highest credit ratings, because the arithmetic said that they would only suffer losses in a scenario the historical data suggested was extraordinarily improbable.
The arithmetic was correct. The assumption was not. When house prices did fall across the country simultaneously, driven by common causes that affected every region at once, the independence the models had relied upon evaporated. Loans that were supposed to fail independently failed together, the pools that were supposed to be safe were not, and the instruments built on top of them, which had been constructed on the same assumption, failed in the same way at the same time.
This is the general principle the crisis illustrates, and it extends far beyond mortgages. Diversification calculated from historical correlations is diversification that depends on those correlations continuing to hold. In ordinary conditions they generally do. In a crisis they frequently do not, because a crisis is precisely a situation in which a common cause affects everything at once, and a common cause is exactly what makes previously unrelated things move together. The diversification tends to disappear at the moment it is most needed, and this is not an accident but a structural feature.
Leverage converted this analytical failure into a systemic catastrophe. Financial institutions had funded their holdings of these instruments with very substantial borrowing, which meant that a relatively modest fall in the value of the assets was sufficient to eliminate their capital entirely. When the assets fell, the institutions were compelled to sell, and because they were all compelled to sell the same things at the same time, the selling itself drove prices down further, which compelled further selling.
The market for the instruments then ceased to function in a way that is worth understanding precisely. It was not merely that prices fell. It was that no price could be established at all, because no buyer was willing to transact and no institution could therefore determine what its own holdings were worth, or what its counterparties' holdings were worth, or whether those counterparties were solvent. Uncertainty of that kind is more paralysing than bad news, because bad news can at least be acted upon.
For an ordinary investor holding a diversified portfolio of shares, the experience was a decline in the broad American market of roughly fifty-seven percent from its peak in October 2007 to its trough in March 2009. The recovery, in this instance, was comparatively rapid by historical standards, with the market reclaiming its prior peak within about four years, which places the episode much closer to 1987 than to 1929 in its eventual consequences, though nothing about living through it suggested that at the time.
The lesson most frequently drawn, that the crisis was caused by greed and dishonesty, is not wrong but is not especially useful, because it offers an investor nothing they can act upon. The more useful lesson concerns the limits of models built on history. Any calculation of risk drawn from past data is a statement about conditions that have already occurred, and the events that destroy portfolios are frequently the ones that lie outside that record. A model cannot warn about what it has never seen.
The practical implication is a certain humility about apparent safety. An investor whose holdings look well diversified according to some measure should ask what common conditions might affect all of them together, and whether that measure would have detected such conditions in advance. The honest answer is usually that it would not, which argues for owning genuinely different things, avoiding borrowing, and maintaining reserves that do not depend on any market functioning at all.
There is a further element that deserves attention, concerning who was assessing the risk and how they were paid. The credit ratings that permitted these instruments to be held by conservative institutions were issued by agencies that were compensated by the very issuers whose products they were rating. An agency that rated an issuer's instruments too harshly might find that issuer taking its business to a competitor, and the agencies competed with one another for that business. This does not require anyone to have acted in bad faith in order to produce a systematic bias; it requires only that reasonable people, exercising judgement in ambiguous situations, tend to resolve the ambiguity in the direction that preserves their commercial relationships. An investor relying on a rating was relying on an assessment produced within that structure, and the structure, not the individuals within it, is what made the assessments unreliable.
At VESTFY™ the 2008 crisis is presented chiefly as a lesson about correlation rather than about villainy. The instruments were exotic and the behaviour was often disgraceful, but the mechanism that transmitted the failure was a mathematical assumption that seemed entirely reasonable and turned out to be wrong at exactly the wrong moment. That mechanism has not been retired, and it will appear again in some other form.