The first step - know your credit risks
The first step in the credit risk management process is for an institution to identify the risky credits in its portfolio of loans, deals or transactions. Banks, for example, achieve this by assigning a credit score or rating to each individual borrower, so the bank can begin to understand its credit risks and assess how best to manage them.
Some assets are easier to assign a credit score to than others. The most developed credit scoring techniques are for consumer loans, such as credit cards and auto loans. As early as the 1980s, credit bureaus focused on information such as credit delinquency and debt burden to assess credit quality.
Some aspects of consumer credit scores have become commoditised. A bureau score that captures almost all the measurable risk inherent in a consumer relationship can be purchased readily and cheaply from a credit bureau such as Equifax or Experian.
Credit providers also use information supplied in credit applications, such as income and whether the individual owns or rents his home. Automated scoring techniques provide a three-digit number derived by computer algorithm from the individual's credit report, which is compared with patterns in thousands of past credit reports.
Bureau scores are also applicable to some small businesses - possibly those up to $1 million in asset size - as many start-up companies are considered to share the credit characteristics of their founders. But there are also many proprietary and industry standard approaches to credit scoring in the small business sector.
Increasingly, consumer and small business credit scores are augmented by data mining technology that helps institutions analyse the "reward" part of the risk/reward equation - for example, a customer segment analysis might rank customers in terms of their propensity to carry a revolving balance. This is important because a customer's profitability is not only decided by the chance that they will default, but also by the likliehood that they will take up bank offerings. Marginally "risky" customers are not always loss makers.
Our next Expert Witness explains the complexity -
and importance - of this "value volatility" to the risk management of
consumer portfolios.
In each lending sector, risk rating systems typically use a different combination of experience-based judgement and quantitative modelling. Most consumer loan rating systems rely entirely or almost entirely on models, while commercial loan rating systems presently rely more on judgement.
One reason for this is the problem of data. Information exists on hundreds of thousands of bad credit card debts and millions of goods ones. In contrast, studies on commercial default models rely on relatively tiny numbers of defaulting companies - partly because the universe of potential defaulters is smaller, but partly because reliable data is difficult to obtain and filter.
Outside the US, the data problem extends even to the largest companies. This is partly because public credit ratings have been applied to major US companies for years - Moody's default database holds information on 1,975 public US and Canadian companies that have defaulted since 1980. By comparison, only 14 holders of long-term local currency ratings have defaulted in Europe since Standard & Poor's began rating corporations there in 1975. All but two of these defaulted in the last two years.
Because of a severe lack of publicly available European data on credit losses, data on recovery rates of US corporate bonds - the amount that is likely to be recovered by a creditor if a company defaults - are used to infer European recovery rates. But there is no standard, or accurate, methodology to do this.
The data problem makes it more difficult to apply statistical approaches, and so more subjective methods have to be called upon. These include business report scores - available for a small fee to suppliers and purchasers of trade credit - based on liens, court actions, creditor petitions and company age and size.
In the US, while the ratings system for larger companies and the scoring of small business/consumer credit is well-established, companies in between these two extremes continue to fall into a statistical no-man's land.
Credit scoring for such "middle market" companies relies on how an individual bank interprets various financial ratios derived from a borrower's financial statements, and other data such as industry sector. While the approaches are increasingly sophisticated, they are not standardised and are difficult to compare and back-test against data.
The Credit Cycle we identified earlier looks set to change this. Banks are recognising that without standard ways to assess the credit risk in their portfolios, it will be difficult to convince both regulators and credit investors that loan and other credit-linked middle market portfolios represent a specific level of risk. Meanwhile, major credit rating companies have begun to promote modelling approaches for middle market credits that weight the various financial ratios in a manner that is standard and which can be more easily tested against the limited historical data that is available.
Public credit ratings are not the only way to assess the credit risk posed by larger companies. The most commonly used quantitative method is based on principles expounded by the well-known academic and researcher Robert Merton. These "Merton models" consider the company's equity as a call option on the value of the firm's assets, in which the strike price of the option is related to the liabilities of the firm. The equity value and its volatility, together with the level of liabilities, provide information that allows the credit modeller to estimate the default probability of the quoted company.
Click the button below for a diagram that illustrates how each of credit scoring and measurement models we have discussed is used to assess a spectrum of firm sizes, from small to large.
Traditionally, once credits have been measured or scored, a bank would decide to accept or reject the implied credit risk of the transaction. But the new credit risk modelling, pricing and transfer tools mean that banks can now actively manage their loan portfolios to ensure an efficient risk/reward ratio and sufficient diversification of loans - much as they would an investment portfolio.
The button below shows this evolution in stated bank practice in terms of seven stages, related to the key modelling and industry developments. Out in the world of credit, it's possible to find institutions that occupy each of these stages, from 1-7, though most occupy Stages 2 to 6.
The role of the rating agency
Credit ratings are vital to the credit industry because they offer consistent and publicly available credit scores, produced by independent agencies, for either the creditworthiness of a major entity or for a particular debt security or other financial obligation.
Ratings help to determine how much companies and governments must pay for credit - but they are also about to take on a new role in the banking system under the Basle Committee's revised Capital Accord. Within a year or two, they are likely to become one part of the process that sets the amount of regulatory capital banks must put aside against credit losses.
However, the banking industry successfully argued against the regulators' plans to make rating agencies the cornerstone of the regulatory capital calculation. Their argument was partly inspired by the fact that there are only two big agencies on the world stage - the banks and the financial system might have become dangerously dependent on these agencies.
These two public credit rating agencies are Moody's Investors Service and Standard & Poor's. Their dominance is challenged only by Fitch, formed by the merger of Fitch IBCA and Duff & Phelps in June 2000. Fitch is presently absorbing Thomson BankWatch, the ratings subsidiary of Thomson Financial.
Although the agencies are independent, they are paid by the companies they rate rather than by the users of ratings information - a conflict of interest that has to be carefully managed. Rating agencies have also been criticised for poor risk estimation, and for being too slow to downgrade their existing ratings when problems appear.
For example, in 1994, both S&P and Moody's gave Orange County their highest short-term rating for a $600 million taxable note issue, just months before it filed for bankruptcy. And in July 1997 the Thai baht's peg to the US dollar broke and the currency plunged in value. But Moody's and S&P did not downgrade Thailand's long-term debt until October 1997. Click on the button to find out how S&P and Moody's structure their rating decision process.
Another "ratings problem" is that outside the US, many companies simply do not have public credit ratings. Around 3000 US companies and corporate bond issues are rated, but in Europe the figure is less than 800. However, this is changing. The number of ratings in Europe has risen by 115% since 1995.
The nature of changes in credit ratings over time, known as ratings migration, is one of the key areas studied by those attempting to model default probability. It's important both to the pricing of credit and to credit risk management - if AAA ratings can be shown to decay to a lower rating only rarely, they are clearly a safer bet than if they turn out to be volatile over time.
Historical data on public credit ratings are analysed statistically to determine the mathematical probability of a particular rating migrating to another rating within a given period. The information is often presented in a transition matrix of the kind seen below. Matrices like these can be used by institutions to determine maturity exposure limits and to measure credit risk in the context of value-at-risk models.
However, there are significant technical issues associated with too great a dependence on these matrices for risk management, as the clickable button below explains.
Average one-year world transition rates
Despite high-profile fears about the risks of electronic banking, or losses from derivative positions, inadequate credit risk management is still the biggest source of serious banking problems according to the Basle Committee - the international banking supervisory body.
The scroll-over Credit Cycle diagram, above, helps make clear how interlinked all these changes are. An unholy alliance of factors set the Cycle in motion in the early 1990s: new credit modelling technology, investor demand, bank marketing, the needs of risk managers and the power of diversification, and attempts to arbitrage regulatory rules on capital charges.
KMV
