Given the lack of early warnings from markets and rating indicators three years ago that big banks were failing, it’s no surprise that finance departments have looked for new ways to gauge the credit quality and counterparty risk of their financial-services partners and customers. One of the crystal balls they now look to is spreads on credit default swaps (CDSs).

But how much can finance rely on the spreads of these credit derivatives to provide clues that a financial institution or large corporate customer is faltering? A new Fitch Ratings study provides some answers.

A contract used to insure the holder of a bond against default by the issuer, a CDS can act as an indicator of default risk. The spread of a CDS indicates the price investors have to pay to insure against the company’s default. If the spread on a Bank of America CDS is 80 basis points, then an investor pays $80,000 a year to buy protection on $10 million worth of the company’s debt. As default risk rises, so does the spread (cost) of the CDS.

The Fitch study shows that CDS spreads were not a reliable predictor of “credit events” —which include defaults on financial instruments, bankruptcies, and debt restructurings — during the financial crisis. This was especially true when the credit event was more than a year away.

Fitch examined the CDS spread movements of 18 corporations, six financial institutions, and three monoline insurers, all of which “defaulted” in some way. Using a standard formula that translates spreads into the probability a company will default, Fitch found that “CDS spreads do not appear to provide a leading signal of default risk for financial institutions.”

Twelve months prior to the credit problems encountered by Fitch’s sample of six financial institutions (including Lehman Brothers Holdings and Anglo Irish), the average CDS spread of the six was 199 basis points. That translates into an average one-year probability of default of 3.3%. When the firms’ credit events were only six months away, the PD had inched up to only 8.3%, meaning fewer than two of the six were expected to default. The CDS spreads of monoline insurers were a little more on track, registering a 63.8% probability of default 12 months out.

Credit default swaps are imperfect barometers, the study indicates, and may not reflect an entity’s fundamental creditworthiness, especially during times of market distress. They can produce “false positives,” says Fitch. In other words, the spread of a CDS can overstate the probability of default by the reference entity (the one whose debt is being insured).

For example, as of December 2008, a group of 29 real estate investment trusts had an average peak CDS spread of 1,154 basis points, implying a default probability of 19.2%, says Fitch. The 18 defaulting companies in Fitch’s sample had a similar spread in the same period. Yet in the ensuing 12 months none of these REITs had a credit event, while the 18 other firms all defaulted, went bankrupt, or restructured their debt.

Since CDSs are financial instruments, their spreads can reflect market liquidity, counterparty risks, and technical factors, such as the high leverage inherent in swaps, say Fitch analysts. A CDS is different from a bond, explains Robert Grossman, chief credit officer of Fitch Ratings. “It reflects the risk of the reference entity, but part of the risk is [also] that the counterparty might not be around to honor the contract,” he says.

In addition, since there are low margin requirements for credit derivatives, a small change in the instrument’s price could produce a large loss, forcing a firm to wind down its position quickly. That could exacerbate price volatility. Says the Fitch report, “Many market participants have a total return orientation based on shorter-term changes in the mark-to-market value of CDS positions; thus spreads do not necessarily reflect a longer-term horizon of fundamental credit risk.”

Despite their flaws, though, the monitoring of CDS spreads can help CFOs and treasurers differentiate relative credit quality across a collection of entities, especially nonfinancial companies. When Fitch compared the average spreads on the 18 defaulting corporates in its sample with the average spreads on an index of 50 high-yield debt issuers over the same time period, it found that the spreads on the defaulting corporates progressively widened relative to the high-yield group, starting 18 months before their credit events.