In this article, we analyze the failures, specifically of stress-testing and other risk management tools, that contributed to the global financial crisis of 2007-2012. The market failure was the root cause of these risk management problems. These market failures fall roughly into three categories:

• Disaster myopia
• Network externalities
• Misaligned incentives

All three have impeccable microeconomic credentials and potentially disastrous macroeconomic consequences.

Disaster Myopia

In a nutshell, disaster myopia refers to agents’ propensity to underestimate the probability of adverse outcomes, in particular small probability events from the distant past. It is well-established in cognitive psychology that economic agents have a tendency to base decision rules around rough heuristics or rules of thumb.

The longer the period since an event occurred, the lower the subjective probability attached to it by agents (the so-called “availability heuristic”).  And below a certain bound, this subjective probability will effectively be set at zero (the “threshold heuristic”).

If the period of stability is sufficiently long – a Golden Decade perhaps? - this subjective approach to evaluating probabilities looks increasingly like a fully-rational, Bayesian approach to updating probabilities.  As time passes, convincing the crowds that you are not naked becomes progressively easier.  It is perhaps no coincidence that the last three truly systemic crises – October 1987, August 1998, and the credit crunch which commenced in 2007 – were roughly separated by a decade.  Perhaps ten years is the threshold heuristic for risk managers.

In the context of financial crises, disaster myopia has been used to explain the LDC debt crisis, the US savings and loans debacle and various commercial property crises. The credit crunch of the past 18 months is but the latest in a long line of myopia-induced disasters.

Network Externalities

Any asset portfolio is, in essence, a financial network.  So the balance sheet of a large financial institution is a network, with nodes defined by the assets and links defined by the correlations among those assets.  The financial system is similarly a network, with nodes defined by the financial institutions and links defined by the financial interconnections between these institutions.

Evaluating risk within these networks is a complex science. When assessing nodal risk, it is not enough to know your counterparty;  you need to know your counterparty’s counterparty too.  In other words, there are network externalities. In financial networks, these externalities are often referred to as contagion or spillovers.  There have been many examples of such spillover during this crisis, with Lehman Brothers’ failure a particularly painful one.  That is why there have been recent calls to calibrate regulatory requirements to these risk externalities.

These network uncertainties make it tremendously difficult for risk managers to identify and price, and hence manage, balance sheet risk.  Consider first evaluating risks across the portfolio of an individual firm.  There is evidence that firms find aggregation of risks across their balance sheet extremely difficult to execute.

To the extent this is done at all, it requires firms to make assumptions about correlations between asset prices.  But at times of stress, asset correlation matrices are unlikely to be stable and correlations invariably head towards one.  So pre-crisis measures of balance sheet risk are likely to be significant under-estimates.

Misaligned Incentives

Finally, and perhaps most contentiously, incentives and governance.  Principal-agent problems crop up in all aspects of economics.  But it is questionable whether there is any event in recent history where these agency problems have been exposed so frequently and extensively as during the current financial crisis.  It is easy to see why.

Financial innovation lengthened the informational chain from ultimate borrower to end-investor.  The resulting game of Chinese whispers meant that, by the time information had reached investors at the end of the chain, it was seriously impaired. In the narrower context of stress-testing, these principal-agent problems appear to have operated at two distinct levels.  First, internally, through the relationship between risk managers and the risk-takers within financial firms;  and second, externally, in the relationship between financial firms and the authorities.  The former principal-agent problem has been rather less discussed, but appears to have been

potent during the credit boom. Decision-making within firms is an arm-wrestle between risk and return, between risk managers and risk-takers.  When returns are high and risks appear low, this arm wrestle can become one-sided.  Power switches from back to front offices and risk managers become the poor relation.

And what is true within individual firms is then amplified by behaviour across the system as a whole, as firms conduct their own armwrestle with competitors for higher returns on equity.  The Bank’s market intelligence suggested this “keeping up with the Jones’s” was a potent force within financial firms during the upswing.

The second principal-agent problem, between firms and the authorities, is different in kind but similar in consequence.  It arises because of a familiar public policy problem – time-consistency.  If the ex-post failure of an institution risks destabilising the system, any ex-ante pre-commitment by the authorities to let it fail will lack credibility.  This is simply a variant of the old adage that if you owe the bank a small amount it is your problem, a large amount it is theirs.  These days, if a bank owes a small amount it is their problem, a large amount it is the authorities.

This time-consistency problem weakens incentives for banks to consider for themselves large-scale risks to their balance sheet which might induce failure.  The safety net becomes a comfort blanket, the backstop a balm.  And the greater the risk these institutions themselves pose in the event of failure, the weaker the incentives to manage risk.  These are topsy-turvy incentives from a public policy perspective, with risk management discipline weakest among those whom society would wish it to be strongest.

Extract from a speech given by Andrew G Haldane – "Why Banks Failed the Stress Test?"