This article began life as a footnote in our discussion paper ‘Banking on Liquidity: Liquidity, Collateral and Derivatives’ⁱ. It identifies the ABX sub-prime mortgage securities indexⁱⁱ as the prime suspect for the spark that lit the crisis bonfire. Numerous studies demonstrate that the mix and volume of banking assets was highly inflammable, but few propose plausible first sparks that might then propagate widely. At the time, assurances were offered by Ben Bernanke and others that the sub-prime problem was containable. Nonetheless, markets broadly did break down and the question that we need to ask, and to answer, is how such a breakdown comes about.

A number of authors have diagnosed this as a question of trust and proposed remedies accordingly; by contrast we see this as the breakdown of the convention that, in financial markets, we ignore much uncertainty and risk in pursuit of the gains from trade. This convention is sometimes known as ‘market confidence’. Some distinction between convention and trust is necessary – a convention exists because it provides mutual benefit while trust is unnecessary unless there is a risk of loss. Driving on the left is one illustration of a convention; it benefits all drivers.

In the early 1970s, with vivid imagery of ‘lemons’ⁱⁱⁱ in used car sales lots, George Akerlof^iv gave us important insights into the effects of adverse selection and uncertainty in markets^v. As we do not know the quality of the car as well as the selling owner, we are exposed to the possibility of exploitation and reflect this in the price we offer to pay. As the potential information asymmetry mounts, distrust prevails and trade declines - then prices reflect only ‘lemons’, poor quality cars. This is a variant on Gresham’s law, that the bad drives out the good. In the absence of the convention, this unfortunate situation, which is a question of trust, will prevail and markets cease to function^vi.

When discussing market valuation in the General Theory, Keynes observed: ‘In practice we have tacitly agreed, as a rule, to fall back on what is, in truth, a convention.... Nevertheless the ... conventional method ... will be compatible with a considerable measure of continuity and stability in our affairs, so long as we can rely on the maintenance of the convention.’ [Italics from the original] The crisis has reminded us of this convention – it is a behavioural regularity that sustains itself because it serves the interests of everyone involved. Chuck Prince’s infamous words on this subject would not have earned him such evident opprobrium in more normal times. This is also why, before the onset of crisis, a risk manager’s cautions are usually disregarded and ignored, like Cassandra; the caution is a challenge to the convention that allows the pursuit of the gains from trade.

The central insights for the analysis of convention and its problems of co-ordination have been known for at least as long as Akerlof’s ‘lemons’; David Lewis’ 1969 ‘Convention: A Philosophical Study’^vii is the seminal work and Robert Aumann’s ‘agreeing to disagree’ gave a rigorous mathematical underpinning to the topic in 1976.

The analytic key is the concept of common knowledge, where the members of a group have similar knowledge and understanding, and also know that all others possess this knowledge. The situation is the old chestnut where you know that I know that you know, and on... and clearly is related to Keynes’ beauty contest analogy for markets, where to win the competition, it is necessary to predict the face most attractive to others, rather than the face most attractive to ourselves. The iterated self referential nature of the concept of common knowledge makes it complicated and best explained by simple example.

The time-honoured illustration is of an island inhabited by people who have either blue or green eyes. There are no mirrors on this island, so no-one can know the colour of their own eyes, but of course, they can observe the colour of all other eyes. The etiquette is that people who know they have blue eyes should leave the island immediately and that discussion of eye colour among the inhabitants is taboo. No-one will leave the island in this situation. Now we will introduce a truth-telling visitor who informs the island population that there is at least one blue-eyed islander present. This is hardly new information; all of the green eyed islanders can see all of the blue-eyed present. Those who are blue-eyed can also see the other blue-eyed islanders present. The information disparity is slight. If there is just one islander with blue eyes, he now knows immediately that he must have blue eyes since all other islanders are visibly green-eyed, and leaves^viii. If there are many (say, m) blue-eyed islanders present then they will leave together on the same day, after that many m days have elapsed^ix. The visitor is merely a catalyst.

This should resonate with market practitioners; how many times have we heard explanations for market movements that we thought were decidedly old news? The market movement only occurred after the news had become common knowledge and was widely understood.

The deteriorating condition of the sub-prime mortgage market was widely known long before the rot set in; researchers from the Federal Reserve of St Louis^x have found that the quality of loans deteriorated for six consecutive years before the crisis and that securitizers were to some extent aware of it. By the end of March 2007, the ABX (2006 – 2) BBB indices were trading (or not) at prices between 70% and 80% of notional value, and the slide down from there was precipitous. Academic studies^xi of the pricing behaviour have concluded that the price declines suffered far exceeded those warranted by fundamental default experience. By summer 2007 continued denial of the problems and uncertainty, maintenance of the convention^xii in pursuit of the gains from trade, was no longer viable for market participants, and market breakdown ultimately ensued. By year end UBS, Citibank and Morgan Stanley had all cited the ABX indices when reporting their write-downs of mortgage assets. To the extent that these indices were reflecting more than default experience, this is a channel for contagion^xiii and exaggerated loss marking, which lowers the banks’ perceived risk bearing capacity.

These indices clearly conditioned the market for breakdown, but also served another purpose. They are contractible - derivatives may be based upon them and cash settled^xiv. Like all indices, they greatly facilitate the writing of OTC derivatives as they lower the ‘lemons’ problems of adverse selection and asymmetric information associated with specific tranches of specific mortgage securities that would usually reduce or eliminate trade. These derivatives were, by design, efficient devices for both hedging and speculative purposes. There is evidence they were used extensively; in the case of Goldman Sachs^xv, mortgage short positions reached 53% of the firm’s total value at risk. Fender & Scheicher^xvi noted that ‘with markets reportedly overwhelmed by large speculative short positions, market liquidity...has been impaired...’

These ABX indices are synthetic^xvii and, like all such products, magnify the consequence of an event in the referenced assets^xviii; to the extent that contracts are written on these indices the losses experienced on the actual mortgages are increased^xix. This rendered incomplete and inadequate analyses, such as Bernanke’s, which considered only the outstanding stock of physical sub-prime mortgages and concluded that their scale was insufficient to result in more general contagion.

The ABX indices certainly seem to have been both spark and dry tinder for the conflagration that followed. Those now seeking retribution, or perhaps merely justice, will be disappointed by the inanimate nature of the cause, an index, which makes this impossible. One consolation, that may mollify this constituency, is that, because of the pricing distortions evident, the TARP policy of buying and holding mortgage securities to maturity should prove ultimately very profitable for the US taxpayer.

Market confidence, the convention, requires that fundamental soundness is commonly understood among market participants. This is a lowest common denominator, which admits a role for research and information discovery in fundamental credit analysis, enhancing a bank’s competitive advantage and performance. Trust though is not the issue.

ⁱ Keating C & Marshall B (2010). ‘Banking on Liquidity: Liquidity, Collateral and Derivatives’. Available online here.

ⁱⁱ The Markit ABX.HE index is a synthetic trade-able index referencing a basket of 20 subprime mortgage-backed securities. It was first created in January 2006 but gained broad press visibility in early 2007. Further details and descriptions of these indices and trade in them are contained in both Fender & Scheicher papers referenced below.

ⁱⁱⁱ The remedies tend to focus upon the central role of experience in the restoration of trust and the more technical have used models of repeated games.

^iv ‘Lemons’ are cars of poor quality.

^v Akerlof G (1970). ‘The Market for Lemons, Qualitative Uncertainty and the Market Mechanism’, Quarterly Journal of Economics. Vol. 84, Issue 3. Available online here.

^vi Democritus took the idea of asymmetric information in markets to the extreme by describing them as ‘places where men meet to deceive.’

^vii Lewis D (1969). ‘Convention: A Philosophical Study’. Reissued 2002 Hoboken, NJ: Wiley-Blackwell Publishers. ISBN 978–0–631–23257-5. Available online here.

^viii The gambler’s adage ‘if you can’t see the fool at the table, it must be you’ is exactly this situation.

^ix This follows as every blue-eyed islander can see M-1 blue-eyed islanders and if no-one leaves on day M-1 and none should, must conclude that he also has blue eyes – all do this on day M and leave together.

^x Demyanyk Y & Van Hemert O (2008). ‘Understanding the Subprime Mortgage Crisis’. Working Paper. Federal Reserve Bank of St Louis. Available online here.

^xi Fender I & Scheicher M (2009). ‘The Pricing of Subprime Mortgage Risk in Good Times and Bad’. Working Paper 1056. European Central Bank. Available online here.

^xii There are many insights that can be gained from this analytic framework and its extensions, for example, the social value of credit ratings and accounting standards . See Morris S. & Shin H., ‘Contagious Adverse Selection’, Working Paper, Princeton University March 2010. The model is simple but congruent with reality, and also renders questionable the ever-increasing demands for transparency. These calls are based on the unrealistic complete efficient markets model of elementary financial theory.

^xiii Arbitrage activity where traders sell other unrelated assets in order to buy the underpriced mortgage securities will provide a cross-market channel for contagion.

^xiv As these indices are trade-able, they can and did develop a life of their own – unlike, say, an equity index their price did not reflect the arbitrage value of underlying reference assets. In other words they can be regarded as reflecting not just mortgage credit risk but rather mortgage credit and market risk.

^xv Memorandum to Members of the Permanent Subcommittee on Investigations entitled: ‘Wall Street and the Financial Crisis: The Role of Investment Banks’ dated April 26, 2010.

^xvi Fender I & Scheicher M (2008). ‘The ABX: How do the markets price subprime mortgage risk?’. BIS Quarterly Review. Available online here.

^xvii This simply means that they do not rely upon the underlying for settlement, though they reference the price of the underlying.

^xviii See Keating C & Marshall B (2010) above for fuller explanation of this point.

^xix They were also prominent instruments in the construction of synthetic CDOs, which accounted for 15-20% of issuance in that market.