Since 1933, the Federal Deposit Insurance Corporation has been entrusted with maintaining the stability of and public confidence in the US banking system. It does so mainly by insuring deposits, and its existence has enabled consumers and businesses to benefit from lower interest rates and banks to book higher profits than they might otherwise have done. With reserves of more than $45 billion, the FDIC’s ability to fulfill its mission is not in doubt. But events in 2002 convinced the FDIC’s senior leadership that the government agency was beginning to lag behind best practice in financial risk management.

In that year, a difficult economy caused more banks to fail than had done so in any 12-month period since 1994. From 1998 to 2002, the number of problem institutions on the FDIC reserve list—one of many indicators of financial strain in the banking system—rose significantly, while the total assets of these institutions quadrupled, to $39 billion. Yet over the same period, the FDIC’s Bank Insurance Fund reserve ratio (the ratio of the fund balance to total insured deposits) dropped markedly, to 1.27 percent—just above the legal minimum of 1.25 percent. If it fell any lower, the agency would have to levy insurance premiums on even the safest institutions.

Early in 2003, FDIC officials decided that the prospect of imposing these premiums warranted a determined effort to bolster public confidence in the agency’s reserve methodologies. They reviewed the process of establishing the contingent-loss reserve (the amount, net of recoveries, that the FDIC expects to pay out in the coming year to resolve bank failures) and, taking a cue from the private sector, set out to upgrade the agency’s financial-risk-management practices. Already, the FDIC has changed the way it calculates loss reserves, thus improving the accuracy of its contingent-loss-reserve estimates and increasing the banking industry’s confidence in them. More significant reforms are under way. By no later than the end of 2004, the FDIC hopes to have transformed the way it does business by developing a new credit risk model and by building a risk-management organization to help it mitigate and prevent losses from bank failures.

A measure of confidence

The FDIC was created by Congress to address widespread bank failures during the 1920s and early 1930s. Its mission is to "preserve and promote public confidence in the US financial system . . . by identifying, monitoring, and addressing risks to the deposit insurance funds"¹ and by limiting damage to the economy when institutions fail. Today it insures deposits worth in excess of $3 trillion, in more than 8,000 banks and thrift institutions—up to $100,000 per individual depositor, per institution.

Although founded by Congress, the agency receives no money from taxpayers and is financed entirely by premiums paid by risky and undercapitalized banks and thrifts. In addition to insuring deposits, it is the primary federal supervisor for state banks that are not members of the Federal Reserve System, acts as a receiver for failed banks, and, when necessary, can impose insurance premiums on banks with insured deposits. Congress has also empowered the FDIC to prevent systemic risks. If a large bank deemed essential to the stability of the banking system is in jeopardy, the agency can provide assurances to creditors, make selected classes of creditors whole, or even close the bank before it fails.

When an institution insured by the FDIC does fail, the chartering authority (the state regulator, the Office of the Comptroller of the Currency, or the Office of Thrift Supervision) closes the bank’s doors. The FDIC, hoping to prevent the spread of panic, then finds the cheapest method of making certain that all insured deposits remain available to depositors, either by liquidating the closed bank and paying them off or by paying another bank to honor the deposits.

The FDIC’s Financial Risk Committee (FRC) estimates the contingent-loss reserve and reports this figure both at the end of each quarter and in its annual report at the end of each calendar year. In all but two quarters during the period from 1997 to 2001, the committee’s estimated failure rate—a key component of the contingent-loss-reserve calculation—was off target: the actual failure rate for each quarter was closer to the average failure rate during the previous two years than to the probability of failure forecast by the committee. These discrepancies indicate that its methodology had produced misleading results (exhibit).

Building a new risk-management system

By changing the way the contingent-loss reserve is calculated (see sidebar, "Improving the FDIC’s financial reports"), the FDIC has already produced more accurate estimates. However, these adjustments amount to a short-term fix rather than a long-term solution: they do not, for example, measure the total level of risk to the insurance funds more than one year out.

The goal of the agency is to introduce a new risk-management model that will enable it to monitor and manage financial risks across various time horizons. When completed, the model will, for the first time, give the FDIC a comprehensive view of its level of risk by producing two basic results: the probability that the FDIC will fall below its target reserve ratio and an accurate measure of the minimum reserves it needs to cover the risks it bears. Such a detailed study of credit risk, similar to those undertaken by best-practice private-sector insurance companies, will enable the FDIC to run a broad range of scenarios and analyses quickly and easily. This will facilitate the early detection and management of risk as well as resource planning across the organization.

Modeling credit risk

The core of the new risk-management model will be the functional equivalent of a credit risk model. Used widely by private-sector companies such as insurers, banks, and retailers, it produces a statistically valid distribution of credit losses over different time horizons. The FDIC’s version will be structured like that of any insurer but will incorporate four elements specific to the agency’s needs.

1. The probability of bank failure. The agency recently used CAMELS ratings,² financial metrics (such as capital ratios), and market data (such as bond yield spreads and equity price volatility) to develop a state-of-the-art econometric model of the probability of failure for individual institutions.

2. Estimates of loss rates. The statistical loss-rate model for estimating loss given the failure of any insured institution will incorporate size; the composition of assets (a mix of cash, securities, consumer loans, and commercial loans, for example); the liability structure, or the quantity of insured deposits; and the speed of deterioration.

3. Correlations of relationships among events. Bank failures are rarely truly independent, so the credit risk model will include a correlation structure that can analyze the relationships among failures. A deterioration in the credit quality of consumer loans, for example, will probably lead to an increased chance of failure among subprime lending institutions as a group. If the likelihood of the most severe outcomes is to be estimated accurately, it is crucial to understand how the fortunes of particular groupings of banks are related.

4. A simulation engine. To produce a distribution of credit losses for any given period, a simulation engine will use Monte Carlo techniques to combine the model’s various components.³ This simulation must not only handle correlations of varying complexity in a reasonable amount of time but also have an intuitive user interface and lend itself to integration into the existing IT infrastructure.

•

The FDIC’s Division of Insurance and Research, which is responsible for building and assembling the components of the credit risk model, has already developed a prototype that can generate a basic loss distribution. Although the model currently lacks the necessary event-correlation structure, the prototype incorporates the concept and will serve as a platform for developing a more sophisticated model.

Modeling income and deposits

The risk-management model under construction will integrate credit risk with three other variables—investment income, premium income, and deposit growth—to provide a comprehensive look at the FDIC’s net risk position at any time.

An investment-income model, for instance, will estimate interest and trading proceeds received from placements in Treasury securities. Although the FDIC already estimates its investment income, the new model will provide revenue forecasts over any time horizon. Similar models will estimate current and potential premiums and the overall growth of insured deposits. By combining these models, the FDIC will have a single comprehensive one that focuses its work on risk metrics that are relevant, timely, and actionable.

Strengthening the organization

By the end of 2004, the FDIC expects to be using the final version of this risk model, which will integrate all essential elements, run in near real time, and incorporate market information on publicly traded banks. The agency’s managers will then be able to respond to the growing demands on them to manage the deposit insurance system effectively and to levy fair and accurate premium assessments on banks.

Two groups created to help the FRC overhaul its risk-management practices have already had notable successes. The National Risk Committee (NRC), comprising senior executives in each FDIC division, is charged with identifying and evaluating long-term business risks facing banks and the FDIC insurance funds. It has helped to coordinate the work of the divisions and will improve their performance by linking investment strategies more closely to the agency’s appetite for risk. The Risk Analysis Center (RAC) monitors risk on a daily basis and recommends responses and courses of action. It has increased the agency’s readiness to react to emerging crises and was the base for market-monitoring operations during the August 2003 power outage in the northeastern states. The RAC has also stimulated new research on interest-rate risk, which in view of the boom in consumer credit is a major component of risk to the banking industry.

For the NRC and RAC to reach their full potential, they will need to spur action across divisional lines and to ensure that the FDIC’s operations improve constantly. With no real line authority or power to instruct other parts of the organization to act on any recommendation, the two committees derive their effectiveness from the leaders who serve on them. To convey a message to the agency as a whole, these leaders will have to focus time and attention on documenting and publicizing recommendations internally. The FDIC’s leadership will also want to consider citing the committees’ recommendations more frequently as the basis for its actions.

The challenge of constant improvement is common to many organizations but perhaps more acute for the FDIC, which lacks the driver of external market competition. The agency may therefore find it beneficial to create formal procedures to help its researchers submit new models for possible use in the risk-management infrastructure. It might also establish criteria for judging the models (their accuracy in predicting bank failures, for example), develop a policy to adopt models that prove effective, and create incentives (such as citations, career advancement, and higher bonuses) for the researchers who develop better ones.

The FDIC has many reasons for achieving best practice in risk management. One is particularly important: the better the agency understands all components of the risk it bears—through lagging indicators such as bank examinations and real-time indicators such as events in the marketplace—the more confidence banks will have in its calculations and the better the relationship between the two sides. The FDIC not only would come closer to achieving its founding mission of bringing greater stability to the US banking system but would also create a better understanding of financial-public-policy decisions that affect depositors, consumers of financial services, and taxpayers alike.