A Breakthrough In Assessment Of Credit Risk
Most bankers now agree that they must incorporate into their loan prices a risk charge which represents a reasonable assessment of both the expected probability of borrower default and the associated costs of that default. But how to determine this risk charge?
The larger banks now routinely array loans on a scale of riskiness based on borrower financial characteristics. These scales generally run from risk 1 (minimal likelihood of default) to risks 8 or 9 (maximum likelihood).
Presumably, the risk charges to be included in prices will reflect the expected loss probabilities in each risk rating, based on historical average default data. There are, however, several problems with this approach. First, the use of historical default averages is itself a second-best approach. Ideally, one would want a forward-looking rather than a backward-looking method of determining likely defaults - that is, a kind of leading indicator, based on current market data, of future creditworthiness.
Second, the loan-risk-rating systems of most banks have only recently been put in place. So there is a dearth of even historical data on expected defaults in each risk-rating category.
And, third, although some banks now feel reasonably confident of the correctness of their risk ratings when loans are originated, they are much less confident of their ability to track the possible degradations in credit quality over time. Thus while loans may have been correctly priced when booked, pricing accuracy tends to slip as some credits inevitably decline in quality, and the system, necessarily rather labor-intensive, fails to detect this deterioration with the requisite speed.
A Model Approach
As a result, many banks have become dissatisfied with their procedures and have been exploring methods of improving them. One approach is to supplement in-house credit evaluations with those provided by external vehicles. In this connection, banks have been looking at various models that purport to measure and monitor credit quality.
Oliver, Wyman & Co. has recently assessed the capabilities of one such model and found it to be superior to that of the other vehicles known to the firm. This model meets the requirements of credit officers for (1) accuracy, (2) speed, (3) cost effectiveness, (4) ease of access, and (5) ex ante rather than ex post information.
The model, called Credit Monitor, was developed by the KMV Corp. of San Francisco, with which Oliver, Wyman has recently formed a joint venture. It enables banks to rate and review the creditworthiness of some 5,000 publicly traded borrowers, based on the recent movement of their stock prices. The model also provides a method for assessing and tracking the credit fortunes of privately held firms, thereby allowing banks to powerfully supplement their analysis of the middle-market customer.
The philosophy underlying the model is that companies default when the market value of their assets falls below the book value of their current liabilities. What concerns the bank lender is the value of the borrower's assets when his bank loan comes due.
Either the value of the assets is greater than the amount of the bank loan or it is less. In the first instance, the firm will repay the bank debt because the market value of its assets is sufficient for it to do so. Even if it lacks cash, the firm can acquire the necessary wherewithal by selling a part of its assets at their market value.
If, however, the market value of its assets is less than the amount of its bank debt, the borrower cannot repay the lender because he has no way of raising the necessary cash.
Calculating Asset Values
Unfortunately, the lender cannot directly observe the market value of his borrower's assets. Even if the lender could put a value on each constituent asset, he could not measure total asset value because of an obvious inability to gauge the firm's going-concern worth.
While not directly observable, the market value of assets can be calculated indirectly. That's because the market value of the assets is conceptually equal to the value of the liabilities - the firm's equity and debt.
The KMV method measures the market value of assets of public firms by tracking the price of their stock and using this information to infer a market value of the debt. Very simply, the value of the assets is obtained by adding the values of the equity and debt, including the current liabilities, which are valued at par.
Suppose, for example, a firm has a mean stock price of $50 per share. Suppose, further, that the volatility of this equity value, as measured by its standard deviation (the square root of the variance) over a relevant time horizon, amounts to 10%. (In practice, such a low volatility would be unusual; the average firm among the KMV 5,000 has a stock volatility of closer to 45%.)
The statistics textbooks tell us that, in a normal or bell-shaped distribution, the mean plus one standard deviation equal about 68% of all possible outcomes. Stated differently, this indicates that 68% of the time this particular company's stock price will range between $45 and $55, with a complementary probability of about one in three that it will fluctuate outside this range.
Since the mean plus two standard deviations equal about 95% of possible outcomes, 95% of the time the company's stock price will range between roughly $40 and $60, with a chance of one in 20 of a greater fluctuation.
Distance from Default
The KMV model calculates the amount of stock volatility, or number of standard deviations, that would be needed to reduce the value of the equity to the point where, together with the imputed reduction in the value of the noncurrent debt, the market value of the company's assets, equal conceptually to the value of its liabilities, is either just equal or below the face value of the current debt. This amount of stock volatility is termed the "distance from default."
This distance from default furnishes us with an expected probability of default, expressed in percentage terms. In the above example, assume that a fall in the company's stock price of just over $10 is enough to drive the value of the liabilities, and therefore that of the assets, below the face amount of the current debt. As noted, stock volatility equal to two standard deviations suggests that the stock price can fluctuate more than $10 to either side of its mean one chance in 20, or 5% of the time.
But the lender is interested only in the downside volatility, which is equal to half this overall range of movement. Therefore, the lender concludes that the probability of this company's equity deteriorating sufficiently to bring it to default amounts to about one chance in 40, or 2.5%
In practice, a 2.5% probability of default is a bit on the high side. The default probabilities in the KMV universe range from 2 basis points to 20%, with a median value of about 2%.
Having determined the probability of default, the lender arrives at his expected loss by adjusting the loss probability downward for the likely recovery rate, which is determined in part by the quality of the collateral and the ease of perfecting it, and upward for the likely cost of carry if the loan becomes nonperforming, and the costs of workout should the loan actually default. The resulting figure forms the basis for the bank's risk charge, which, as noted, should be incorporated directly into the loan charge.
The results churned out by the KMV model have been compared with those of other models and other methods of computing default probabilities. For example, S&P debt ratings provide an implicit probability of default, equal to the percent of defaults in each rating category.
To beat the S&P system, a model should be able to identify a firm's deteriorating credit-worthiness before the agency moves to downgrade its rating. Based on data from 1979 through 1990, the KMV model signaled credit problems on average one to three years before S&P moved to lower a troubled company's rating.
A model that would vie with S&P must also identify (1) a higher percentage of those borrowers who subsequently default as likely defaulters and (2) a lower percentage of those who do not subsequently default as likely defaulters. KMV bests S&P on both counts.
A Useful Surrogate
The KMV model can be adapted for use in analyzing the credit quality of private firms (most middle-market borrowers) through the use of what is dubbed "comparables" analysis. That is, one can find among the KMV coverage of 5,000 publicly traded companies at least one, if not more than one, firm that operates primarily in the same industry and has roughly the same financial characteristics as any given private borrower. The expected loss probability of this publicly traded firm is then viewed as an approximate surrogate for that of the middle-market firm under analysis.
KMV also provides a supplementary method of assigning default probabilities to private firms. It produces what is called an implicit debenture rating. By regression analysis, it is possible to isolate those financial characteristics - values of various key accounting ratios - that explain most of the differences among S&P and Moody's credit ratings.
Thus one can estimate with considerable accuracy the likely ratings that private firms which exhibit these financial characteristics would have obtained had they held publicly traded debt. Knowing this "shadow bond" rating, it is of course possible to infer an expected probability of default that is at least as valid as, if not more valid than, that of the various rating agencies.
When furnished with this armamentarium of tools, banks can determine with much greater accuracy than in the past the credit losses that can be expected from their public and at least their larger private borrowers. They can thus move to price these risks in a rational fashion.
In addition, these tools will facilitate nationwide commercial loan securitization, which has hitherto been impeded by the inability of loan buyers to determine in a cost-effective fashion whether loan sellers were withholding derogatory credit information. And securitization is, of course, the key to solving the problem of underdiversified loan portfolios - the principal cause of bank failure in the United States.
Mr. Rose, formerly senior columnist for this newspaper, is now associated with Oliver, Wyman & Co., a New York-based management consulting firm that specializes in financial institutions.