A Practical View of Risk Modelling: Pitfalls And Promise of Basel II
Since Basel II first moved off the drawing board into the boardroom, banks have collectively invested hundreds of millions in changing and upgrading their credit processes. Particular focus has been given to the process by which they evaluate the credit risks arising from their lending portfolios and the sheer growth of the credit risk software business in terms of numbers of vendors, specialist integrators and consultants bears witness to this. Starting salaries for graduate and post-graduate mathematicians as well as newly qualified PhDs have risen sharply as the need for these specialist risk-related disciplines grows, and specialist headhunters and recruitment consultants have also benefited from this skills shortage. New types of leveraged investment vehicles have compounded the impact as the derivatives market and the hedge fund industry have grown simultaneously.
Credit default swaps (CDS) in particular and other forms of risk diversification have multiplied, so that the original del credere risk posed by a borrower can be diluted and spread around a larger market of underwriters. The size of the CDS market is now estimated by the International Monetary Fund to be far greater than the world’s corporate bond markets, so that the pricing of these derivatives in fact sets the pricing for the underlying debt markets themselves. However, while there has clearly been an increase in appetite for the underlying risks, it is arguable whether a reduction in risk has in fact taken place overall, as the credit standing of many of the intermediaries involved has been eroded by their willingness to assume risk in the form of proprietary positions. (As a general rule, more of these firms’ own money is now directly invested in these instruments than would have been the case some years back.) This change in the markets is evidenced by the high prices of assets in the markets, and consequently low risk premia, giving rise to doubt as to whether risk is adequately compensated.
Is there evidence to suggest that profitability of lending operations has increased as a result of this investment? The experience of the last 10 years has been to demonstrate the link between deterioration in the credit quality of borrowers, the conditions of the local economy and banking failures, and therefore to highlight the magnitude of the contribution that could be made by sound credit risk management to financial stability.
In July 2006, the Bank of England highlighted the increasing problem posed by household indebtedness. In the 1990s, this amounted to an average of 100% of household disposable income. It now runs at 150%, and is noted as a key financial sector vulnerability. Although loan write-offs are rising, they are rising from a historically low level. Studies from elsewhere in Europe show that different economies are at different stages of maturity, but indebtedness is rising as a general trend which must affect resilience to external shocks eventually.
The Basel II regulations attempt to connect the amount of capital available to support lending more closely to the actual credit risk as banks estimate it from their experience. Unfortunately, once the capital constraints actually start to bite, the regulations are likely to have the effect of imposing credit rationing, probably at the stage of the economic cycle where borrowers are most sensitive. This has been pointed out in academic studies of the economics of the risk disciplines of Basel II – see ‘Credit Rationing Effects of Credit Value at Risk’ by Slijkerman, Smant and de Vries, 2004.
One way to avoid this happening is to introduce the idea of modelling correlations between credit risk and market risk and eventually also operational risk. However, two problems arise here. Correlations are easy to overestimate, and this could have a worse effect on the rationing of capital in times of scarcity. The second problem is that getting correlations right requires even greater amounts of data to be analysed and taken into account in the mathematical models used. The higher data requirements demand more sophisticated computing power, and more highly organised databases in order to be able to carry out the calculations over a sensible time-frame. Thus, more money must be ploughed into the business, making it harder for smaller institutions to compete effectively.
Unless smaller organisations devise ways to fight back, by outsourcing analytical and processing functions effectively, the higher barriers to entry into each new phase of sophistication will start to tell. Of course, if this does not start to happen, it might be argued that the greater investment is not perhaps justified by practical benefits on the ground. A number of well-informed commentators still regard much credit risk assessment as ‘art’ rather than science, despite increasing evidence that short-term outcomes are becoming easier to predict accurately. This is perhaps telling. Environmental factors are still something of an unknown quantity in such modelling exercises, which might account for accuracy being somewhat limited in the medium-term models.
The FSA in the UK has identified the area of correlations between credit, operational and market risks as a key objective for development among financial institutions active in the UK. They have correctly identified the possibility of severe crystallisation of risks from these differing areas at one time, with a potentially huge impact. While such an event is unlikely, its severity makes it important that banks start to plan against it. This concern was echoed in a recent speech by Timothy Geithner of the Federal Reserve, who pointed out that the very same innovations in credit instruments that have been so successful in making the credit markets less vulnerable to smaller shocks might have the undesirable effect of increasing the potential size of larger shocks. His views are supported by academic studies on the mechanics of contagion, not mere supposition.
Larger institutions, with larger market shares, can prepare more easily and will generally benefit more from detailed analysis of all types of statistical outliers – events that are rare but of significant size. For larger banks the logistical costs of such exercises can be more easily absorbed across a larger portfolio, and they have easier access to larger amounts of data. It has been argued that smaller organisations can benefit more from savings in redesigning business processes as they tend to be more agile. However, interviews with programme managers working in smaller institutions generally show that such changes are in fact rare. Often these programmes are in the hands of career project managers, with little support in terms of subject matter expertise. Change management disciplines often intervene to prevent improvement of the underlying business process, unless such changes were foreseen at the start of the programme.
Surveys of the state of business processes in the insurance market show that, at any one time, more than 60% of the core business processes are exposed to significant levels of unplanned, unmanaged operational risk. Functionality that may have been carefully planned at the outset may be lost during, for instance, cost control drives, a poorly managed succession of staff or just overtaken by events during long implementation phases. The list of causes is long and makes depressing reading. Are banks so very different? We suspect not, although we have not undertaken any quantitative analysis of similar problems in the banking sector. The root cause of much of this chaos is poor alignment of the business process with supporting infrastructure and implementation of point solutions, probably resulting from poor management of projects that set up or adapt the processes in the first place.
So does detailed mathematical analysis of risk correlations make sense given such a backdrop? While modelling techniques may not yet have evolved to an optimum, experienced operations professionals will recognise the wisdom of attempting to tease out the more complex relationships between risks. Product managers will benefit from any understanding of the different dynamics of different market sectors, and determine their segment pricing strategies accordingly, to generate the most profitable mix of customers.
Any evolving discipline places organisations in a quandary as to the split between long-term strategic investment and the need to demonstrate short-term effectiveness. In a sense, investment in new credit disciplines is a type of arms race, with the banks that develop ‘perfect’ models hoping to gain measurable competitive advantage. However, the sheer volume of competition makes it unlikely that success in these areas will be worth the investment in the long run, as returns in credit markets dwindle and move out of line with actual risks.
Like many other innovations in the early stages of development, it is important for banks to keep up with their peer group, without over-spending on the blind alleys of innovation. It seems likely that more and more banks will be attracted towards the mid-range integrated risk management software products, rather than continuing the trend towards buying larger, more hardware-hungry versions. And the smart money in this sector will be looking at smaller vendors with more innovative products, rather than investing in major change management programmes at this stage. It seems unlikely, given the competition in this sector, that any bank will be able to create a favourable business case for much more and, in the long run, overspending in such a low margin environment can only lead to trouble.