1 BANKING Portfolio credit management for the credit crunch This article is an extract from BaselBriefing 14, November 28 November 28 FINANCIAL SERVICES
2 1 BaselBriefing 14 Portfolio credit management for the credit crunch An eye on the past A year after the beginning of the credit crunch financial institution write-downs associated with sub-prime lending and leveraged loans have exceeded half a trillion US dollars and could ultimately exceed more than a trillion 1. These losses have materially damaged the capital bases of the institutions concerned, leading to defaults, takeovers, depressed share prices and the need for significant capital raising activity in an attempt to replace depleted buffers against insolvency (see Table 1). What have we learned so far? Among other things, the need forimproved risk management and asset valuation methodologies during periods of illiquidity, a more objective view from rating agencies when risk assessing structured assets and a redressing of the imbalance between short term business-side profit-taking and longer term prudent risk management. The distinction between market pricing and the cost of holding assets has also been overlooked. Aside from these there has been relatively little discussion of important lessons to be learned at the portfolio level rather than asset level. Such lessons should include: the assessment of risk at the aggregate level, portfolio level correlations (aside from correlations within structured products), the build up of risk concentration and of the hedging (or lack of it) at the portfolio level. After all it is portfolio tail risk driven by these factors that threatens institution survival. The first additional lesson is that deal assessment processes should be in place not only to evaluate vanilla and structured transactions on a standalone basis. They should also measure the incremental risk in the context of the portfolio when assessing deal profitability so that significant concentrations cannot accumulate without a high degree of visibility and oversight. That is to say, portfolio managers should do more than focus on single-name descriptions of risk: each new deal or hedge made may make sense on a stand-alone basis, but may build concentration or be value destroying at the portfolio level; Table 1 Largest write-downs and losses with capital raising activity Firm Writedown/Loss (US$bn) Capital Raised (US$bn) Wachovia Corporation Citigroup Inc Merrill Lynch & Co Washington Mutual Inc UBS AG HSBC Holdings Plc Bank of America Corp JP Morgan Chase & Co Wells Fargo & Company Morgan Stanley Lehman Brothers Holdings IKB Deutsche Industriebank AG Royal Bank of Scotland Group Plc Credit Suisse Group AG Deutsche Bank AG Fortis ING Groep N.V Credit Agricole S.A Barclays Plc Mizuho Financial Group Inc HBOS Plc Source: Bloomberg, 2 October 28 2.
3 2 BaselBriefing 14 Secondly, if financial institutions wish to mitigate large losses that threaten their solvency they should seek to measure and manage tail risks. This usually means focusing on Value-at- Risk or Expected Shortfall (or more broadly speaking, Economic Capital) and understanding how tail losses arise (see inset); Thirdly, risk concentration appears to be still very poorly understood in the market. Many institutions are still largely unable to allocate portfolio risk to the obligor level (or do so appropriately for their business) and thus identify the biggest contributors to capital or the most likely drivers of tail losses; and Fourthly, a common risk assessment methodology across the organisation is required so that the risk taken at a sub-portfolio level is consistent with the objectives at the institutional level. An eye on the future Painful though the sub-prime and leveraged loan related losses have been, financial institutions should also focus on the credit risk associated with the rest of their portfolio and consider what additional capital and portfolio management activity is required to support and mitigate the risk therein. Institutions fortunate enough to have avoided the worst of the write-downs from the sub-prime sector should not be too complacent. Thus far there have been relatively few corporate defaults, but the economic outlook is negative. The premium on corporate bond issuance is at its highest level since the early 199s 3 and the five-year investment grade expected corporate default rate has risen five-fold since July Thus the market view of credit risk has significantly increased over the past year. For the banking sector in particular the outlook has not been positive. According to the Federal Deposit Insurance Corporation there were 117 US banks facing near term difficulties remaining solvent as of August 28. Since then the banking sector has undergone significant trauma and many players have needed injections of capital to survive. Nonetheless further capital may be needed to cover future exposure to corporate defaults. In such circumstances of scarce capital supply, increased risk at the asset level and increased hedging costs, institutions should have a clear understanding of the aggregate risk at the portfolio level and understand how much capital is needed to support the portfolio. It is equally important to be able to identify where concentrations lie in the context of the portfolio. Probabilities, value deterioration and tail risk: a basic overview Unfortunately it is not possible to predict precisely which counterparties in our portfolio will downgrade, default or lose value over some time horizon. One can only attach probabilities to these events. A suitable correlation model is then used to tie these model inputs together to determine the likelihood of various portfolio losses (or gains). The resulting graph or histogram is the portfolio loss density, such as that illustrated in Figure 1.The loss density can be used to determine the expected loss, the volatility of losses and the probability that a certain level of loss is not exceeded over the time horizon. The last of these allows us to calculate how much capital is required to support the portfolio to a given level of confidence to maintain solvency. This amount is commonly referred to as Value-at-Risk (VaR). A related risk measure is the Expected Shortfall (ES): the average loss expected in excess of VaR. Both measures of tail risk are commonly used to define Economic Capital (EC). Source: Kevin Thompson, KPMG in the UK, 28. Figure 1 Tail risk. The idealised loss distribution and Value-at-Risk. Loss volatility = 7 Probability.25% Expected Loss VaR(99.75%) ES(99.75%) Loss (US$m) = 8 =1 =134 Souce: KPMG in the UK, 28
4 3 BaselBriefing 14 Self-assessing tail risk: beyond Basel II Unlike the calculation of Pillar 1 capital (under Basel II), there is no hard and fast rule for how one should calculate tail risk and Economic Capital generally because there is no single prescribed credit portfolio model to determine the loss distribution. Financial institutions should therefore continually strive to model the possible portfolio states, their likelihood, correlations and changes in portfolio value as best they can to assess the potential risks for their book. It is a clear advantage for an institution to have its own credit portfolio model and assessment of tail risk so that it has the best view possible of how risks arise, their likelihood and how best to mitigate them. Unfortunately, the Basel II prescription of the capital and aggregate risk calculation is insufficient for these purposes. In addition, the Basel II calculation allocates the same risk and capital to a counterparty regardless of the correlation with the rest of the book and regardless of the tail risk being managed (see below). Furthermore, the Basel II calculation does not consider the time evolution of losses. Nonetheless Basel II has encouraged institutions to improve their data coverage, quality and infrastructure. Many institutions that have implemented Basel II are now in a position to exploit that investment with their own portfolio risk models. Not all risks are equal: an alternative way of viewing risk Which are the biggest risks in the portfolio? Rather than focus on VaR or Expected shortfall there is another way to manage portfolio losses: by directly focusing on a particular loss amount that the financial institution would like to mitigate. This is best illustrated with an example (see Figure 2): The x-axis indicates the possible aggregate loss (L) from the portfolio over some time horizon (ranging from zero to a maximum loss of US$1bn say); and The y-axis indicates the risk contribution to that level of loss: that is the expected loss from a counterparty conditional on incurring the portfolio loss L. In the figure we highlight the risk contribution curves for two counterparties in the portfolio: Counterparty A: a US$1m exposure to an entity with probability of default (PD) of 4 percent; and Counterparty B: a US$8m exposure to an entity with PD of 2 basis points. A portfolio manager whose objective is to mitigate losses of US$2m would view asset A as being the more risky (its risk contribution is US$19m versus US$5m for counterparty B). But if they were managing risk or allocating capital at the US$2bn loss level (a tail risk) it would be asset B that is the more risky by far (consuming US$281m of the US$2bn risk versus US$7m from Counterparty A) (see Table 2). Compare this with the Basel II view, where asset A would be more risky regardless of the loss we are trying to mitigate. The calculation of risk contribution levels is non-trivial and complicated by the fact that the calculation involves the whole portfolio; every piece of counterparty information in the portfolio together with the correlation Table 2 Risk contributions versus portfolio loss Portfolio loss level (US$m) , Risk contribution: Counterparty A (US$m) Risk contribution: Counterparty B (US$m) Source: KPMG in the UK, 28
5 4 BaselBriefing 14 Figure 2 Risk concentration Risk Contribution (US$m) Counterparty A (US$1m, PD=4%) Counterparty B (US$8m, PD=2bp) Portfolio Loss (US$m) 2 Source: KPMG in the UK, 28 model is used in the calculation of a single risk contribution curve. Nonetheless, the curve is extremely valuable as it shows us how the risk from an obligor changes according to the level of loss we are trying to manage. This is critical for efficient use of capital and hedging budget 5. Did the notion of EC fail during the credit crunch? A few institutions that implemented (or claimed to have implemented) an EC framework suffered significant losses during the first year of the credit crunch. Was this a failure of the EC concept? The answer depends on exactly what framework was in place. It is not logical to suggest that just because an institution has a model in place and a means of calculating its aggregate tail risk that it has a good risk management framework and that it is less likely to suffer significant losses. Regarding inputs, a sensible EC framework requires sound assessment of the PD, credit migration and value deterioration. In relation to subprime, many of these parameters may have been insufficiently conservative or responsive, being driven by the prevailing market implied data that was more suitable for hedging than risk assessment. The institution should also have a sensible model for aggregating the risk to the portfolio level as already discussed. This EC calculation itself should also be responsive. As portfolio structure, asset valuation and counterparty creditworthiness change, so will the tail risk. Institutions should be careful not to mask the risk by choosing an unresponsive portfolio model, by dampening down or capping PDs and risk drivers, or by capping overall levels of EC in the calculation process for example. The market will be sceptical if asset level risk increases or portfolio size grows significantly but the reported EC hardly increases. The EC concept should be more than just a calculation of tail risk. Having more realistic inputs and being able to aggregate risk are merely the first step in the EC and portfolio management process. It should be accompanied by the ability to risk-assess new deals and allocate risk to the obligor level in a way that is consistent with that risk aggregation methodology. Furthermore, business functions need to manage the portfolio actively using a consistent portfolio view framework together with prudent single name and single deal analysis at deal inception and in hedging decisions. Most importantly, if management is unwilling to act upon the indicators of the EC framework, the framework is paying nothing more than lip service to risk management. Conclusion What are the desirable features of the credit portfolio model suite? The need for transparent, responsive and flexible in-house models that can be used to measure the aggregate level of portfolio credit risk and capital, assess new deals in a portfolio context, allocate capital and guide hedging decisions has never been greater. What capabilities should a financial institution have when it comes to credit portfolio risk measurement and management? Senior management should ask whether their organisation has the capabilities listed in Table 3 and consider the possible impact on capital and shareholder value if it does not.
6 5 BaselBriefing 14 Table 3 Desirable features of a credit model suite 1) Model transparency: are all components of the model explainable? 2) Is there sufficient counterparty PD, exposure at default and loss given default (LGD) data? It may be necessary to model a distribution around these parameters to reflect the uncertainty or future volatility in these parameters. If exposures are time-varying then the appropriate time varying distribution data should also be sought. 3) Is the correlation model understood? Many institutions use a black-box / 3rd party model that cannot be explained or reproduced in-house. Correlation is such an integral part of the model that black-box correlation models should be avoided wherever possible. 4) Is it possible to correlate PD, LGD and exposure (or value) if desired? 5) Is it possible to assess the aggregate risk on the portfolio and calculate several views of risk: expected loss, loss volatility, Value at Risk etc. simultaneously? It is vital that portfolio managers use multiple risk measures to manage their books. 6) Is it possible to model time-varying exposures (including derivatives, Collateralised Debt Obligations, time varying loan exposures etc) when they exist in the portfolio and understand the evolution of portfolio losses through time? 7) Is it possible to assess new deals in the context of the portfolio? This must be sufficiently fast for practical use in the business at least for standard trades. 8) Is it possible to demonstrate that portfolio level tools avoid the build up of significant concentrations? 9) Is it possible to identify the biggest risk contributors / capital concentrations? The ability to breakdown risk, to allocate any risk measure (including Economic Capital) down to the obligor level and hence determine concentrations at the counterparty, sector and region level is necessary.this includes the ability to determine risk contributions to any level of portfolio loss. 1) Is it possible to perform stress testing and scenario analysis? Not only to carry out sensitivity analysis and stress model parameters (including correlation) but also perform scenario analysis linked to the real world. 11) Can the same risk framework be applied consistently across the portfolio and vertically at different levels in the institution: at the sub- and aggregate portfolio levels? 12) Does it have an active portfolio management activity? Are effective channels of risk distribution available when the price and risk levels are appropriate? That is, does the institution have the infrastructure to transfer risk? 13) Does the institution have the capability to assess portfolio manager performance, effectiveness of hedges and profitability? Can it do this post-deal and through time? 14) Does the institution have a means of measuring (and incentivising) portfolio manager risk reduction activity? Source: KPMG in the UK, 28. Contact References: Kevin E Thompson Director KPMG in the UK Tel: +44 () Thompson, K.E. and A. McLeod (Winter 26/27), Thompson, K. E. and R. Ordovas (23), The Road to partition, Risk, volume 16 (5), Analytic calculation of conditional default statistics and risk contributions using the Ensemble method, Journal of Credit Risk, volume 2 (4), Bloomberg, 2 October Bloomberg, 2 October 28. At the time of writing no updated figures were available. The table does not include the consequences of the default of Lehman Brothers, losses associated with the takeovers of Merrill Lynch and HBOS, or the support of Freddie Mac, Fannie Mae and AIG. 3 Lehman Brother s US Credit index (27 August 28) 4 itraxx Europe and CDX to 1 September 28 5 See Thompson et al. (23, 26) for further details
7 kpmg.com KPMG firms have an international network of regulatory, risk and capital management professionals. To discuss any of the matters raised in this edition of BaselBriefing, or any other regulatory, risk management and data management matters please contact: Kevin E Thompson Director KPMG in the UK Tel: +44 () The information contained herein is of a general nature and is not intended to address the circumstances of any particular individual or entity. Although we endeavor to provide accurate and timely information, there can be no guarantee that such information is accurate as of the date it is received or that it will continue to be accurate in the future. No one should act on such information without appropriate professional advice after a thorough examination of the particular situation. The views and opinions are those of the authors and do not necessarily represent the views and opinions of KPMG International or KPMG member firms. 28 KPMG International. KPMG International is a Swiss cooperative. Member firms of the KPMG network of independent firms are affiliated with KPMG International. KPMG International provides no client services. No member firm has any authority to obligate or bind KPMG International or any other member firm vis-à-vis third parties, nor does KPMG International have any such authority to obligate or bind any member firm. All rights reserved. Printed in the U.K. KPMG and the KPMG logo are registered trademarks of KPMG International, a Swiss cooperative. Produced by KPMG s Global Financial Services practice. Publication name: Portfolio credit management for the credit crunch Publication no: RRD-1447A Publication date: November 28 Printed on recycled material.