Chapter 15: Basel II and Extreme Risk Analysis
- International Monetary Fund. Legal Dept.
- Published Date:
- February 2013
Basel II mandates the maintenance of bank capital to address three broad categories of risk: credit risk, market risk and operation risk. Basel II methodology assumes that credit risk can be reliably specified with respect to particular asset categories. These projections are based on historical experience, reflected in data sets, which are at times general and at times specific to a particular institution. Basel II may have failed, however, to identify the strong shift in the correlation of defaults that accompanied the financial crisis. Market risk assessment displays similar issues of under-anticipated correlation under extreme conditions. Of the three categories, operational risk is the least tractable. It is a catch-all category, reflecting both internal failures and external events. The character of risk shifts markedly between “ordinary times” and extreme events. Consider this matrix:
|Ordinary Times||Extreme Events|
|Credit risk||Defaults may be predicted based on historical data—no strong correlation of defaults||Inadequate data for robust quantification or risk—strong correlation of default|
|Market risk||Loss of value linked to performance of asset||Loss of value linked to type of asset (contagion)|
|Operation risk||More likely to be institution-specific||More likely to be environmental|
This paper addresses the ability of Basel II to respond to extreme events. Extreme events include events thought to be extremely unlikely and events that are unimagined (and hence ex ante unimaginable). Basel II largely examines risk as faced by individual financial institutions. Yet systemic risk (contagion) is a well-recognized feature of the international financial system. The liquidity crisis, however, appears to be a novel (unanticipated) phenomenon. Cross-failure of institutions may be yet another example of unanticipated correlation of outcomes, with a peculiar magnification effect. Basel II does not ask—at least not explicitly—whether there is adequate capital across the entire banking system. The current financial crisis may be an instance where this failed to hold. A revised Basel regime might need to account for additional capital stored outside individual institutions to assure adequate capital under extreme conditions affecting the broader banking sector.
The Basel II international banking regime is undergoing critical examination in the wake of the global financial crisis.1 Among the current critiques is that the Basel II Framework2 failed to provide for extreme events—outcomes which are sufficiently rare so as to not be reflected in risk estimates generated from historical data, but which have sizeable, if not disastrous, consequences.3
This is a distinct critique from the concern that the levels of protection provided were inadequate. Setting a level of protection is always a judgment call.4 Higher levels of protection require increasingly greater costs. The Basel II Framework reflects a series of such compromises between the limits of protection and tolerated risk. For example, a frequent level of protection found in Basel II is the ability to survive a ten-day time period with a ninety-nine percent confidence level. This is not absolute protection; rather, the level of protection admits (and accepts) bank failure in certain cases.
Extreme events can be usefully divided into two categories. The first includes identifiable outcomes (often predicted by history) that occur too infrequently to be effectively managed. Think of the financial equivalents of a 1,000-year flood, an 8.5 point earthquake centered on Los Angeles or a meteor strike. Hyperinflation and systemic bank failures have occurred—we simply feel these events are extremely unlikely to occur at any given moment in time, and their impacts are likely to be so overwhelming that risk mitigation does not appear to be worth the effort.5 At the outer limits, hopelessness sets in.
The second category of extreme events is more frightening. These are the inevitably unaddressed risks associated with unidentified events—events for which there is no historical experience.
Taleb teaches (among other things) that novel events do occur. Our imagination, and hence our ability to anticipate outcomes, is limited by our experiences. Our habits, however, instruct us to assume continuity.
As the financial crisis lingers, critical attention is directed both backward and forward. Looking to the past, one seeks to determine the root causes of the debacle. Looking forward, one seeks to sketch the regulatory and institutional features of a “returned to normal” financial system. The 2004 Basel II framework will be object of both retrospective and prospective scrutiny. Basel II is the harmonized international bank regulatory system developed by the Committee on Bank Supervision under the auspices of the Bank for International Settlements. Basel II principles had recently been implemented, or were in the process of implementation, by the major national supervisors at the time of the onset of the global economic crisis. Most of these same jurisdictions had previously adopted the Basel I framework, which focused almost exclusively on the application of more rigid capital adequacy standards to be problem of credit risk.
Basel II addresses three categories of risk. The original risk category (and the central concern of the 1988 Basel Accord) is credit risk. Credit risk dominates traditional banking—it is the risk that borrowers will not be able to repay lending banks and so will place lending banks into a liquidity or solvency crisis. Credit risk has two important dimensions: the probability of default and the magnitude of loss given default. Credit risk is most relevant to those assets held on a bank’s “banking book” (as opposed to its “trading book”—where market prices anticipate losses accompanying eventual defaults).
The second Basel II category addresses market risk. Increasingly international accounting standards (and consistent national regulation) require banks to mark assets to market prices (as opposed to the historic practice of holding assets at book values). Markets risks can be divided into various subcategories, reflecting the various sources of loss-generating events. Examples include interest-rate risk, exchange-rate risk, basis risk and migration risk.
The third Basel II category covers risks that do not conveniently fall under the prior categories. These are labeled operation risk. Examples include risk of fraud (by inside rogue traders or outsiders), risk of cataclysmic events (hurricanes, September 11) and, to a large extent, counterparty and systemic risk (including old fashioned bank runs and new fashioned liquidity crises). It is within the category of operation risk that many rare and most unforeseen events are likely to develop.6
Within each of these three categories, Basel II prescribes regulatory disciplines. Yet Basel II had always limits to its effective protection in mind. This is most clear in its provisions concerning the use of bank-designed and implemented risk systems, where achievement was defined as maintaining sufficient capital to have a defined level of survival within a defined period, based on the range of historical outcomes observed during a recent period.
A ninety-nine percent survival system openly acknowledges the inherent possibility that (in about one percent of cases) more serious losses, or even catastrophe, occur. And many argue that the current global financial crisis is one of those less than one in a hundred events. Further, the ninety-nine percent survival rate may prove to be overly optimistic. While the occurrence of an event may say little or nothing about its ex ante probability of occurrence, the occurrence of an un-imagined (and hence truly unanticipated) event indicates that ex ante risk assessment was in error. It is not just the frequency of the event that matters; the magnitude of losses associated with an unanticipated event must be considered in accurately assessing risk.
While there are many areas of potential inaccurate assessment of risk within the Basel II Framework, and there are general quarrels with the setting of levels of protection, this paper, following Franklin and Taleb, investigates Basel II’s treatment of “extreme risk.”
Risk assessment as generally practiced is a history-based science, premised on our intuitions as to the cyclical (or at least repetitive) nature of the unfolding of events. Note there are different data that may be missing in different assessments. Some evidence of potential outcomes may be found outside the chosen test period—but are otherwise known and recognized. Data sets drawn from periods of rising house prices might simply miss foreclosures evidenced in darker, but more remote, times. Other outcomes may be missing simply because their possibilities have not been suggested by historical experience. These are the “black swan” events described by Taleb.7 This paper will focus on the inevitable presence of these outlier events and on how a reconstructed international banking system (Basel III) should be designed to take them into account.
Credit Risk in Extreme Circumstances
Both Basel II and the prior Basel I systems address credit risk. The chief technique mandated is the maintenance of minimum levels of capital.8 Regulatory capital acts like a “cushion,” absorbing credit losses without impairing the ability of a financial institution to repay depositors (and otherwise remain solvent). Capital works well where default risks can be reliably forecasted—and when there are not strong default correlations among individual assets (loans). Imagine a bank with 100 loans, each of which is in the amount of $1,000. Assume (based on the bank’s historical experience) that no more than 4 percent of these loans will default and that the average loss on each defaulting loan is 50 percent of face value. In this particular outcome, the bank will experience a loss of $2,000 (4 loans of $1,000 each losing half their value). If a bank maintained more than $2,000 of capital, its solvency would not be impaired. The Basel II system thus attempts to set capital adequacy levels high enough to absorb credit losses under reasonably foreseeable conditions.
Capital adequacy is not a failsafe system. First, the very setting of a level signals a magnitude beyond which capital will be exhausted (analogous to the relationship between the height of a dike and its effectiveness with respect to storm surges). All a particular level of capital can do is to provide a specific level of protection—the magnitude of loss absorption is directly related to a specified level of capital.
Second, it is vulnerable to error due to the potential atypicity of the historical data that form the basis for setting levels. Third, the assumptions concerning the correlation of asset performances may not hold true for extreme ranges. Indeed, the essence of an extreme event is that adverse performance of particular financial assets becomes strongly correlated. Response to an isolated instance of an individual’s disease is distinct from the response to a plague.
If predictions of defaults are based on (relatively) prosperous times, they will greatly understate risk. This appears to be the case with respect to mortgages in the United States during the housing boom. Defaults were low as overall economic conditions were positive, and stressed borrowers were able to manage difficulties by selling the underlying assets during a period of steadily climbing house prices. It was such atypical historical data that served for setting many capital adequacy levels.9 As such, there was a structured underestimation of the potential for a much larger rate of loss.
The loss given default was also underestimated. As we move toward a more extreme event, not only do default rates increase, but so does the magnitude of loss given default. The sharp downturn in collateral value (i.e., house prices) results in bank’s being far less able to recover when mortgages default.
Finally, in an extreme event, defaults become positively correlated. Instead of comprising discrete events, individual defaults loom as a unified event (destroying the effect of any risk mitigation attained through holding a portfolio of assets). The pool of performing assets necessarily shrinks, producing a magnification effect.
Positive loss correlation and magnification also results at the level of the overall banking sector. Many institutions simultaneously experiencing stress are much more dangerous than an isolated bank failure (acknowledging the potential for contagion effects in better times). The ability of any particular bank to withstand contagion effects such as an imitative bank run or payment system breakdown is compromised during periods of extreme financial stress. Banks share weakness in a systemic manner.
These observed features of the current crisis suggest that considerably more capital is needed when an extreme event occurs. The capital need not be fixed within the banks themselves, however. As the initial management of the financial crisis has demonstrated (might demonstrate?) many banks have been able to call on additional capital from national treasuries in their “lender of last resort” role.
Extreme Market Risk
Traditionally banks specialized in the generation and holding of illiquid assets. Many if not most originated loans were held until maturity. Accounting conventions permitted these assets to be held at constant (i.e., book) values. Contemporary finance has changed these practices in profound ways. Banks are less likely to hold assets through maturity; rather they increasingly follow the originate-and-distribute model, profiting from fee income. Other banks (and other financial institutions) purchase assets.
Basel II adopts the distinction between the “banking book” and the “trading book,” The banking book lists assets that are being held in the traditional way, where expected return is the present value of periodic interest and principal payments, as well as any final principal payment extinguishing the obligation. The trading book lists assets that are held for indefinite periods of time, where anticipated return derives from capital appreciation, to be realized when assets are disposed of (prior to maturity).
Holding assets require different amounts of regulatory capital depending on which “book” they are assigned. Assets may be assigned to the “banking book” or to the “trading book,” The assignment of assets to a particular book is reflective of an institution’s intentions with respect to holding an asset. Assets held to maturity are properly assigned to the banking book; assets destined for sale (including traditional assets intended for distribution or securitization) are properly assigned to the “trading book,” Banks effectively have considerable discretion in assigning assets to one book or the other—indeed, the possibility of inter-book arbitrage has emerged as a weakness of the Basel II Framework.
Basel II presumes that assets will be marked-to-market. This affects all assets, irrespective of the book to which they are assigned. As market values increase, additional regulatory capital is “created,” When asset market values decrease, however, capital is consumed. This is true regardless of whether the asset is intended to be held until maturity or not. Marking-to-market has the effect of mandating quasi-instantaneous write-downs. When market prices decline broadly across classes of assets, capital can disappear at an alarming rate, plunging an institution into a capital crisis. This then flows to capital maintenance levels, and the noted counter-cyclicality of Basel II: banks need to raise expensive additional capital precisely when their financial profile is the weakest.
There are many identifiable causes of (or situations associated with) decreases in asset value. Among these are liquidity risk, interest-rate risk, exchange-rate risk and migration risk.
Market volatility is itself a source for concern. Again, banks no longer have the facility to try to “weather the storm” by holding assets until conditions improve. Capital impairment occurs regardless.
The trend towards securitization has amplified the risk of an extreme market shift. One of the appeals of securitization is the (apparent) reduction of risk achieved by pooling assets. Any real reduction of risk depends on low correlations of non-performance by the particular assets forming the pool. In extreme circumstances, where most or all assets in the pool lose value, the pool begins to resemble a single, bad asset.
Securitization vehicles often magnify risk exposure to extreme conditions. First, they may utilize leverage. Depending on the nature of the claim (tranche) a bank might hold, risk exposure may be significantly enhanced. Second, they may be divided into claims where correlation and other magnification effects are underappreciated.
Compounding these concerns is the likely unreliability of shadow “market” prices for claims on securitization pools and other complex products. These are most often not true prices at which claims are exchanged in thick markets, but rather synthetic prices projected from the (already suspect) prices of the underlying assets. But whether or not these prices are accurate, or even real, they do drive capital demands under the Basel II system.
History is simply missing for assets which are novel. Synthetic pricing proved to be a poor substitute for constructing risk assessments. Unanticipated financial performance resembles an extreme event, even if it seems all too probably in hindsight.
Specific Types of Market Risk
The current financial crisis brought home a category of underappreciated risk. Banks engaged in active originate-and-distribute activity anticipated very short holding periods for bank-originated assets. These assets were either sold in secondary markets or packaged into securitization vehicles.
Warehousing and pipeline risk refers to the event where originating banks are unable to off-load assets due to unexpected changes in market conditions. Involuntary holding of these assets expose the bank to losses due to declining values of these assets.
Reputation risk refers both to the prospect of a decline in value of a bank’s goodwill, as well as the possibility that a bank will feel constrained to undertake certain transactions in order to maintain goodwill. This particular risk manifested during the current crisis.
Banks involved in securitization activities frequently transferred assets to formally independent (and bankruptcy remote) corporate entities (often called Special Purpose Entities). Once these assets were transferred, they no longer appear on the bank’s balance sheet. Nor were they treated as off-balance sheet items, unless the bank was contractually obligated to intervene in event of default or other stress.
Many banks re-acquired distressed assets from Special Purpose Entities, notwithstanding the absence of a contractual obligation to do so. Recognition of this moral obligation is explained by concerns by banks for their reputations. Banks realized avoidable financial losses in order to avoid investor wrath. Given this history, it is likely that implicit puts will be recognized when securitization markets eventually re-emerge.
Extreme Operational Risks
Operational risk describes the residual category of identifiable risks beyond credit and markets risks. These include both internal and environmental risks. Internal risks include poor management, losses caused by rogue traders, and fraud. It also includes certain counterparty risks that are not captured under the credit or market risk categories. External risks include those presented by weaknesses in the interlocked financial system, as well as the usual horrors (wars, plagues, insurrections).
An operational risk event may be of an unanticipated magnitude—and not adequately provided for. And it may be perversely correlated with other stresses. Consider the Madoff Ponzi scheme. A Ponzi scheme is subject to collapse even under the most benign conditions, but is much more vulnerable during periods of market pessimism and declining asset values. The recognition of losses from such an event (corresponding to discovery and collapse of the scheme) may well compound simultaneous credit and market losses.
During extreme events salient operational risks are more likely to be external—that is, they cannot be so easily traced to management failures. And so they are more likely be experienced by many institutions—leading in turn to magnification and systemic risk.
Not all risks can be anticipated. Indeed, it is difficult to manage risks that cannot be imagined (those that are outside of experience) although experience teaches that unanticipated and unimagined events do occur. The liquidity crisis had been imagined by some, who—like Cassandra—were ignored. That said, one might credibly say that the “market” failed to imagine or anticipate such an outcome.
Extreme Systemic Risk
Basel II directly controls the activities of individual institutions. That said, like all prudential regulation, Basel II is in some sense more concerned by the spillover effects of an institution’s crisis onto the larger banking system.10
In the current financial crisis, bank capital was not only inadequate as measured on an individual institutional basis; it was inadequate on a system-wide basis. There were no sufficiently well capitalized banks that were able to absorb the capital deficits (and negative equity) of the failed institutions. Government response in the United States, United Kingdom and elsewhere demonstrated the existence of severe de facto capital deficits.
While institutions remain with well-worn, recognizable names, the effective truth is that the pre-crisis banking sector has been utterly destroyed and what exists today results from an ad hoc public recapitalization. Schwarcz has argued that a better approach to the problem of systemic risk should be located outside particular banking institutions in a “liquidity provider of last resort.” In some sense, Schwarcz anticipated the immediate crisis’ responses, which might be described as a bundle of state-provided liquidity and capital infusions to the broad banking sector.
Responding to Low Probability High Impact Events
The Committee’s initial response to the financial crisis has been to call for identification of low probability high impact events through systematic stress testing. Stress testing is to be viewed as a complement to model-based approaches (such as Value-at-Risk), identifying limits and blind spots in those methodologies.
The Committee’s response does not answer—in a quantified way—how much additional capital need be held in order to adequately meet a low probability high impact event. The answer is rather the unquantified (and perhaps unquantifiable) “enough.” In some sense, a stress test eliminates probability as a determinant. Rather, it considers the adverse event to have occurred and asks what the magnitude of loss might be. The resultant quantum of “adequate” capital would be, it seems, equal that magnitude—any lesser amount of capital would prove inadequate to maintain institutional solvency in such a crisis.
It is not clear how management should make operational the results of a stress test for a rare event. One can always push variables more and more adversely, yielding events that call for ever increasingly amounts of regulatory capital. At a certain point, management will cease to engage. Events—like a large meteor strike—may be demonstrable possibilities, but may be of such low probability as to not warrant a response. Or, in the alternative, their impact may be so severe as to overwhelm any prudential measure.
Responding to Unforeseen Events
The Committee’s initial recommendation does not address events that are truly unforeseen (and so escape Value-at-Risk and other modeling reliant on historical data). The Committee urges banks to avoid “failures of imagination” in identifying risks, but beyond this, has little to say.
As such, it must be conceded that there is some set of events which are likely not captured in risk assessment—we operate within our cognitive limits. Recognizing this, however, there do remain alternatives to ignoring these outcomes. Implicitly, regulatory capital may not in fact be adequate to protect banks and the banking system from certain unanticipated outcomes. The default regulatory results are either (1) institutional failure and resultant systemic risk or (2) inevitable state intervention to recapitalize a failed bank or failed banking system.
A final post-crisis assessment of the Basel II Framework requires that we revisit what we expect bank capital to do—and an acknowledgement that there will always be limits.
The Basel II Framework was never understood to be unbreakable. Rather, it was meant to address certain probabilities of certain magnitudes. To observe the failure of Basel II during the current crisis does not imply that Basel II was inadequate to address the risks it anticipated. Another reconstruction of an international harmonized system for bank regulation featuring capital adequacy controls as a central tool will similarly have to define set limits.
That said, there is some value to be derived from observing rare events. We may end up over-engineering, fighting last wars, as inappropriate amounts of capital are drawn to cover losses that now seem all too likely.
Jeffery Atik, “Basel II: A Post-Crisis Post-Mortem,” 19 Transnational Law & Contemporary Problems 731 (2010); Harald Benink and George Kaufman, “Turmoil reveals the inadequacy of Basel II,” Financial Times, February 27, 2008.
Basel Committee on Banking Supervision, Basel II: International convergence of capital measurement and capital standards: a revised framework, comprehensive version, June 2006. The Basel II Framework includes: (1) June 2004 Basel II Framework; (2) Unrevised elements of the 1988 Basel Accord; (3) 1996 Amendment to the Capital Accord to incorporate market risks; (4) July 2005 paper on The application of Basel II to trading activities and the treatment of double default effects.
Nassim Nicholas Taleb, The Black Swan: The Impact of the Highly Improbable (2007) (hereafter, Taleb, The Black Swan).
The 1988 Basel Accord set a general minimum capital requirement of 8 percent of the bank’s risk-weighted assets. Basel II lowers this for certain banks.
It is a bedrock financial principle that the “risk” of default of a U.S. Treasury obligation is zero. Of course, this risk is merely defined as “zero” in order to express other financial risks on a comparative basis; even the U.S. Treasury conceivably could fail. It is not a scenario most enjoy contemplating.
See James Franklin, “Operation Risk Under Basel II: A Model for Extreme Risk Evaluation,” 27 No. 10 Banking & Fin. Services Pol’y Rep. 10 (2008).
See Taleb, The Black Swan, Note 3 above.
Edward S. Prescott, “Regulating Bank Capital Structure to Control Risk,” 87(3) Fed. Reserve Bank of Richmond Econ. Q. 35 (2001).
Under the new methodologies introduced by Basel II, certain banks were permitted to set their capital requirements based on their institution-specific historical data (as opposed to economy-wide data). This magnifies the possible atypicity of the relevant data pool. When unexpected correlation kicks in across institutions, separate institution-specific capital adequacy targets make little sense.
See Steven L. Schwarcz, “Systemic Risk,” 97 Geo. L.J. 193 (2008).