FDIC: Deposit Insurance Options Paper

This options paper is part of a comprehensive review of the U.S. deposit insurance system by the Federal Deposit Insurance Corporation (FDIC). We are undertaking this review to assure the ability of the system to meet its responsibilities over the next decade. Industry consolidation, expanded activities, global-ization and the use of technology have advanced the business of banking and the products and services offered to American depositors. The FDIC wants to ensure that the deposit insurance system continues to protect depositors and contributes to its full extent to the stability of the banking system.

The United States has the oldest federal deposit insurance system in the world, established in 1934 to put an end to the devastating bank runs that shut down businesses and contributed to the Great Depression. The system proved to be a success; following its introduction, deposit insurance restored public confidence in the banking system. For the next three generations, the system served its purpose by helping prevent banking problems from becoming banking panics. In the 1980s, when hundreds of banks and thrifts failed, deposit insurance acted as the anchor for public confidence in the banking system.

In good times and bad times, deposit insurance provides a safe and certain place for people to put their money. By eliminating the disruption caused by bank runs, deposit insurance contributes to the foundation necessary for a robust banking system and by extension, a dynamic financial system. In turn the general economy benefits from the stabilizing influence of deposit insurance.

The success of the U.S. system of federal deposit insurance is particularly evident in contrasting the U.S. experience during the 1980s crisis with recent crises in Asian and Latin American countries that lacked explicit deposit insurance systems. During the U.S. crisis, there were no depositor runs on banks, and bank failures were resolved through a well-established, orderly process. This was not the case for countries without explicit deposit insurance, and it is perhaps sufficient to note that more than 30 countries chose to implement new, explicit deposit insurance systems during the 1990s. The benefits of deposit insurance are appreciated worldwide, and the U.S. system has become a model for the rest of the world.

Nevertheless, the 1980s crisis in the U.S. also provides a sobering reminder that a flawed deposit insurance system can be extremely costly. U.S. taxpayers were billed for more than $130 billion to clean up the savings and loan crisis following the demise of the Federal Savings and Loan Insurance Corporation (FSLIC). This demonstrates that deposit insurance raises complicated issues and requires a careful balancing of competing public policy concerns.

Today, the bank and thrift industries have never been healthier. Bank capital levels are at an all time high, profitability has climbed for the ninth year in a row, and the insurance funds have substantial combined reserves of $42 billion. There will never be a better time to address the latent flaws in the system. Reforms now will also help us maintain the proper incentives for risk and reward to insured institutions, as well as fairness among institutions that present different levels of risk to the system.

The FDIC has identified three fundamental areas for review: the processes for pricing risks, funding insurance losses, and setting coverage limits. This options paper describes various ways in which we might make improvements to the deposit insurance system. The options are intended to prompt analysis and comment from individuals and organizations that have an interest in the issue.

1. The Need for Reform

With the Federal Deposit Insurance Corporation Improvement Act of 1991 (FDICIA), Congress passed a number of significant reforms to shore up the deposit insurance system. These included prompt corrective action, least-cost resolutions, scaling back of too-big-to-fail, the introduction of risk-based premiums, and a mandate to maintain adequate insurance funds. With the Deposit Insurance Funds Act of 1996 (DIFA), Congress ensured that members of the Bank Insurance Fund (BIF) and the Savings Association Insurance Fund (SAIF) would not face significant and arbitrary differences in deposit insurance pricing.

Despite these significant improvements, the current deposit insurance system has several features that work against the effective and equitable functioning of the system:

The continued existence of two separate insurance funds based on an anachronistic distinction;

The current pricing system that creates inappropriate incentives and raises fairness issues;

The requirement that banks are required to fund insurance losses when they can least afford it; and

Uncertainty for depositors as to the future real value of FDIC coverage.

Over the past decade the FDIC has stated its view that the two insurance funds the FDIC administers should be merged. The distinction between the funds is increasingly arbitrary; a combined fund would be stronger and more efficient; and the time to merge them is when they are both healthy. These arguments are laid out in detail in Attachment A. This options paper will not address this flaw, other than to state the FDIC’s position that a merger of the funds is good public policy either on a stand-alone basis or as the prerequisite for any other changes to the deposit insurance system.

The second and third of these problems result from the conflicting mandates of the FDICIA: to price deposit insurance premiums according to the risk posed by individual institutions, and to maintain a target level of reserves within the insurance funds. The tension between the dual mandates of risk-based pricing and a fixed fund level became far more explicit in 1996 as DIFA severely limited the FDIC's ability to price according to risk.

Because of current restrictions on pricing deposit insurance, most banks and thrifts pay no insurance premiums when they are doing well, but pay high premiums when the industry is weak and banks are failing. This does not make sense for the banks or for the communities they serve. It is possible that, in difficult times, deposit insurance premiums could reduce the pre-tax net income of insured institutions by almost $9 billion. Based on current average capital and loan-to-assets ratios for all insured institutions, this reduction in income could lead to a contraction in lending of more than $65 billion at the precise time in the business cycle when loans are most needed.

The current process for setting deposit insurance coverage limits has brought the issue before Congress on a somewhat arbitrary and ad hoc basis. This has resulted in significant fluctuation in the real value of insurance for depositors. The current coverage limit of $100,000 has declined in real value by half since it was established in 1980. This raises the question of whether Congress wishes to continue providing the same level of insurance protection for consumers in real terms, or to allow the coverage level to erode in value by maintaining the status quo.

The current deposit insurance arrangements lead to several questions:

How Should the FDIC Price Risk?

Through a combination of legislative changes, regulatory choices and economic events, the pricing and funding of deposit insurance evolved during the 1980s and 1990s into something fundamentally different from what existed during the first 50 years of the FDIC’s history. Banks that are paying for deposit insurance at the end of the 1990s are those that have run afoul of capital regulations or the supervisory process. This is a significant departure from past practice. Pricing of deposit insurance has evolved into a penalty system for the few, rather than a priced service for all.

Thus, a decade that began with a legislative mandate for risk-based insurance premiums ended with the FDIC providing a free guarantee of almost three trillion dollars in bank and thrift liabilities. As a result, the moral hazard problems FDICIA intended to address with risk-based deposit insurance may have become more firmly entrenched than ever. (Moral hazard problems are discussed in more detail in Section IV, "Coverage Limits.")

A striking feature of a zero premium is that not only may the rate paid by vastly disparate banks be identical, but the dollar amount as well: a bank with $100 billion in deposits and a complex risk profile can be billed the same amount for its insurance as the smallest and most conservatively run community bank. Presumably, the rationale behind a statutory zero premium is that, as long as a fund is above its target level, it does not need additional funds. However, aside from raising money for the insurance funds, premiums also serve to align economic incentives. When a valuable product is offered at zero cost, it leads to that product being overused, causing distortions throughout the marketplace and, in the case of deposit insurance, potentially exacerbating moral hazard.

If deposit insurance were priced according to risk, it is likely that every bank in the U.S. with insured deposits would pay something for deposit insurance, for the same reason that every bank pays at least some spread over Treasuries for unsecured debt. However, since shortly after the BIF was recapitalized in May 1995, its members that are in the best-rated, 1A-assessment category have not been required to pay deposit insurance premiums. Members of the SAIF that are rated 1A have paid no premiums since January 1997.¹

Footnote: ¹ More details on the risk categories in the current premium system are presented in Attachment C.

At year-end 1999, only 7 percent of all banks and thrifts paid premiums into the deposit insurance funds. Ninety-three percent, or more than 9,500 institutions, do not pay premiums. This stands in stark contrast to the first 50 years of the federal deposit insurance program, when every insured institution paid an annual rate of 3.3 to 8.3 cents for every $100 of insured deposits.

Despite the uniform assessment ratings given to these 1A institutions, they do not all present uniform risks to the deposit insurance funds. The current premium matrix does not recognize institutions that, by objective measures and historical experience, have a higher risk profile, unless the institution fails to maintain the minimum level of capitalization to be considered "well-capitalized" as defined for prompt corrective action purposes or is subject to heightened supervision.² In a less favorable economic environment, many of these 1A-rated institutions would deteriorate faster than others, yet that higher degree of risk is not built into the current assessment scheme.

Footnote: ² Federal supervisors rate insured institutions on six factors: Capital; Asset Quality; Management; Earnings; Liquidity; and Sensitivity to market risk (CAMELS). Institutions receive an overall rating ranging from 1 to 5, with 1 being the best rating.

The original decision by the FDIC to lump CAMELS 1- and 2-rated institutions into the same risk category for premium purposes was largely codified into law in 1996 by the DIFA. Federal Deposit Insurance Funds Act, Pub. L. No. 104-208, §§ 2708(b) and 2708(c) (1996) (codified at 12 U.S.C. §§ 1817(b)(2)(A)(iii) and (v)). As a result, the FDIC is largely prohibited from distinguishing between CAMELS 1- and 2-rated institutions for determining premiums.

How Should New Deposits be Treated?

Most banks and thrifts established since the recapitalization of the insurance funds have never paid for deposit insurance. Through March 2000 this included 844 new banks and thrifts whose insured deposits totaled more than $44 billion. The responsibility for maintaining the $550 million needed to capitalize these deposits at a 1.25 percent DRR falls on the other members of the deposit insurance system.

Similarly, institutions that are rated 1A can grow their insured deposits without paying assessments. This zero marginal cost of insurance clearly differs from the private insurance industry, in which higher coverage amounts entail higher charges. With the marginal cost of deposit insurance at zero, the same issues of fairness arise that occur under the new bank scenario: all insured institutions eventually are assessed to cover deposit growth at the fastest-growing, 1A-rated institutions. In a deteriorating financial environment, it will be necessary to raise assessment rates earlier or by a greater amount to make up for the dilution of the reserve ratio attributable to unfunded insured-deposit growth.

Under some circumstances, insured-deposit growth could occur rapidly, accelerating the need to raise assessment rates for all insured institutions. This could happen even in a favorable economic environment in which deposit-insurance losses remain low. In early 2000, an investment company announced plans to convert some of its customers’ funds into FDIC-insured accounts. Reports in the media suggested that as much as $100 billion could be converted in this manner in a relatively short period of time. Sudden growth of this magnitude at 1A-rated banks, with no corresponding growth in the fund balance, would dilute the fund’s reserve ratio. In this example, the BIF reserve ratio would fall by 5 basis points. With a reserve ratio of 1.35 percent as of March 31, 2000, such a decline would leave the fund’s reserve ratio above the statutory minimum of 1.25 percent, but the industry would be closer to mandatory rate increases for all insured institutions, depending on insured-deposit growth and insurance losses. From March 31, 2000, through June 30, 2000, insured deposits at the banks affiliated with the investment company grew by $12 billion.

There is also the possibility of a large shift of household assets into insured deposit accounts in the event of financial market volatility. There is currently more than $11 trillion outstanding in U.S. equity holdings (including mutual fund shares) alone. In a protracted bear market, some of these funds could be transferred to insured deposits. And it is still too early to gauge the probable impact of electronic banking on insured deposit growth. Obviously, the likelihood of deposit inflows from these examples, as from a myriad of other possibilities in an era of financial modern-ization, cannot be known. The question is whether the current deposit insurance system is capable of addressing the issues raised by these possibilities.

Conversely, institutions that shrink their deposits are not compensated for the indirect benefit they confer on other members of the system. Most BIF members have paid no premiums since 1995, and most SAIF members have paid none since 1996, but all insured institutions paid very high rates in the earlier 1990s. The issue of deposit growth and shrinkage becomes important in any discussion of rebates (other than the refunding of current assessment income). Any such program would require legislation, but the question of who is entitled to how much is complicated by the existence of institutions whose deposit growth or shrinkage was atypical. For example, aggregate BIF-insured deposits grew by 10.5 percent from year-end 1995 to year-end 1999, during which time one bank grew its insured deposits (without any acquisitions) from $19 million to $1.2 billion (up 6,140 percent), and another bank reduced its insured deposits from $763 million to $423 million (down 45 percent). Of these two banks today, the one with a lower level of insured deposits paid considerably higher total assessments in the 1990s.

How Should Losses be Funded?

In reaching a point where the FDIC does not collect assessment revenue from most institutions during good times, we have clearly departed from any concept of spreading insurance losses over time by collecting revenue on an ex ante or long-run expected loss basis. In contrast, prior to 1989 it could be argued that Congress intended the FDIC to operate under a form of long-term expected loss pricing. During the period 1933-1989, when premiums were set by statute and never departed from a range of between 3 and 8.3 basis points per annum, accumulated premiums and the investment income on those balances enabled the system to roughly pay for itself. The system in place today, in contrast, amounts essentially to charging nothing in times of prosperity, and a lot in times of adversity, thereby potentially magnifying swings in the banking cycle.

The current "cushion" in the BIF, the amount by which the fund exceeds 1.25 percent, is $2.3 billion.³ If insurance losses not covered by the systemic risk exception were to exceed this amount—as they did in each year from 1988 through 1992⁴—and the fund fell below 1.25 percent and was expected to remain there for a year or more, the FDIC would be forced to raise average assessment rates to a minimum of 23 basis points. Therefore, all banks would be forced to pay substantially higher premiums at a time when many banks were under stress. On a strict pay-as-you-go basis, banks would have had to pay approximately 62 basis points in 1991.

Footnote: ³ Despite growth of the fund during the first quarter of 2000, this cushion fell from $2.5 billion at year-end 1999 because of insured-deposit growth in the first quarter.

Footnote: ⁴ Annual losses ranged from $2.7 billion to $6.9 billion during this five-year period. These are actual losses and not loss provisions, which were even higher but were partially recovered when many projected failures did not occur.

If the FDIC had more latitude in setting rates when the reserve ratio falls short of the DRR, the recapitalization period could be extended with rates less than 23 basis points. This would help to avoid a credit crunch and to moderate the negative impact of deposit insurance premiums on real economic activity.

How Should the Coverage Levels be Determined?

The current process for setting deposit insurance coverage limits has brought the issue before Congress on a somewhat arbitrary and ad hoc basis. This has resulted in significant fluctuation in the real value of insurance for depositors. Deposit insurance has a simple, but important purpose: to provide a safe place for depositors to keep their money, as a way to prevent bank runs and maintain the stability of the banking and financial system.

Since 1934, the basic coverage amount has increased five times, from $5,000 to $100,000. Most of the increases more or less reflected cost-of-living adjustments, but the most recent increase is an exception. The 1980 jump from $40,000 to $100,000 had more to do with attracting deposits to insured institutions in a competitive market of very high interest rates. Today, 20 years later, $100,000 of deposit insurance has lost about half its value, based on the Consumer Price Index.

The next several decades will be a time in which the population is aging, retirement costs are increasing, and the supply of federally-backed investment vehicles, such as Treasury notes and bonds, may decline. Thus, a long-term perspective may argue for allowing for the coverage limit to keep up with changes in the price level, household wealth, or other measures relevant to households.

However, there are trade-offs to consider. Higher coverage limits can increase moral hazard. The 1980 increase is widely viewed as contributing to the high cost of the savings and loan crisis. Also, the impact of higher coverage limits on insured deposit growth is difficult to predict, and the likely distribution of benefits is subject to debate.

2. Overview of Options Paper

This remainder of this paper organizes the discussion into three major areas: pricing risk, funding insurance losses, and coverage levels.

Section II of this paper discusses the pricing of deposit insurance for individual banks. If deposit insurance is viewed as a service that banks use, the question is how this service should be priced. One answer is that the price should reflect the risk that the bank presents to the deposit insurance system. This expected loss approach to pricing is consistent with the best practices that have developed in the banking industry in recent years.

The next question is what information should serve as the basis for pricing. Supervisory ratings are appealing because they are based on quality information and reflect the judgment of experienced supervisors; however, too great a reliance on ratings raises concerns about consistency and subjectivity. This suggests the appeal of more objective information, which could include non-public information (such as credit exposures), Call Report information, and market information. Finally, the FDIC could generate pricing information through risk-sharing contracts with market participants.

Section III deals with how deposit insurance losses are funded from an aggregate perspective. The funding of FDIC losses has evolved over the years from a system that featured steady premiums with a fluctuating reserve ratio to a system that targets a specific reserve ratio and results in volatile premiums. The mandate to maintain a particular ratio can lead to steep premiums during bad times and calls for rebates during good times.

One general approach is a user fee system in which banks have no claim on past premiums. Under such an approach, the question is whether premiums will be relatively stable and consistent with expected loss pricing, or whether premiums will be more closely tied to current losses or the reserve ratio in order to guard against premiums that are too high or too low.

A mutual approach would differ from the user fee system in that banks would have some claim on past premiums. This could take the form of rebates when the insurance fund is viewed as too large; this raises the question of how to allocate these rebates. Alternatively, banks could hold claims on the insurance fund, similar to mutual fund shares. This could address concerns about free rider and pricing problems. Under mutual arrangements, the cash flow between a bank and the insurance fund could have two components: one to price risk at the margin and the other to reflect the bank’s claim on the fund.

Section IV discusses the appropriate extent of deposit insurance coverage. The section begins with a review of the history of coverage levels in nominal and real terms. This is followed by preliminary estimates of how an increase in the coverage limit would be expected to increase the amount of insured deposits. This depends on the behavior of households and businesses, and further study would allow more confidence in these estimates.

It is widely recognized that there is a tradeoff between the stability that deposit insurance brings and the potential for distortion of the market process. Coverage levels speak directly to that tradeoff: higher coverage may provide greater stability during difficult times, while lower coverage may enhance market discipline and minimize distortion. The section addresses this tradeoff with a discussion of moral hazard, implicit protection, and industry structure.

The options in the coverage section include continuing the existing system of ad hoc statutory adjustments; indexing for inflationary adjustments; or simplifying the current system to limit a particular level of coverage to one account per person. Other ideas for changes to coverage include extending higher coverage to municipal and other public deposits; this raises issues similar to those posed by brokered deposits. The section ends with excess coverage options including increased use of private coverage, new excess coverage through the FDIC, FDIC-backed private insurance, or coinsurance systems.

3. Review Process and Comments on Options Paper

This paper is one step in the FDIC’s comprehensive review of the deposit insurance system. FDIC Chairman Donna Tanoue publicly announced the review on March 7, 2000, in a speech before the Independent Community Bankers of America. On April 25, 2000, the FDIC held a Deposit Insurance Roundtable with bankers, their trade group representatives, consumer group represent-atives, and industry experts. The Roundtable provided an opportunity for interested parties to raise issues and discuss broad policy options for consideration in the FDIC’s review. A transcript of the proceedings may be viewed at www.fdic.gov. The Roundtable was followed by outreach meetings with bankers during May and June in Minneapolis, Dallas, and Kansas City. The FDIC also held discussions with members of state banking organizations during their annual spring visits to Washington and with several leadership groups and staff of the national trade associations.

In addition, the FDIC has held discussions with academics and other outside experts. To provide a more explicit "market perspective" on deposit insurance pricing and fund exposure, the FDIC retained the risk-management consulting firm of Oliver, Wyman & Company (Oliver, Wyman). Oliver, Wyman employed an analytical framework similar to that used by the largest financial institutions for analyzing their credit exposures. The purpose was to explore ways to incorporate "best practices" from private-sector risk management into the consideration of FDIC pricing and funding issues.

The ideas and perspectives that were communicated through these various efforts have been incorporated into the options paper. The FDIC will carefully review comments and weigh feedback from the options paper in order to narrow the policy choices and guide the additional analytical work necessary to develop a set of policy recommendations that is appropriately balanced, workable, and fair. As studies conducted by FDIC staff and others are completed, additional discussion—perhaps a new set of roundtables or outreach meetings—will be arranged with interested parties.

Because this paper is not intended to advocate specific approaches but to solicit comment, we have presented a wide range of concepts for dealing with the policy issues under discussion. Examples of how these conceptual approaches might work in practice are included, and readers are encouraged to comment both on the specific examples and on broad conceptual approaches.

Comments on the options paper may be registered on the FDIC web site at www.fdic.gov. The Internet version of the options paper will include a survey which will become available on August 31. Readers are invited to respond to the specific questions that will be posted on the FDIC web survey for each topic and to provide any additional comments relating to the survey questions.

Comments may also be addressed to Robert E. Feldman, Executive Secretary, Attention: Comments/OES, Federal Deposit Insurance Corporation, 550 17th Street, N.W., Washington, D.C. 20429. Comments may also be hand-delivered to the guard station at the rear of the 550 17th Street Building (located on F Street) between 7:00 a.m. and 5:00 p.m. on business days. In addition, comments may be faxed to (202) 898-3838, or sent via the Internet to comments@fdic.gov.

FDIC staff will review comments as they are received and summarize them each month through the fall of 2000, beginning with September. Comments will be available for inspection and photocopying at the FDIC Public Information Center, Room 100, 801 17th Street, N.W., between 9:00 a.m. and 4:30 p.m. on business days.

II. PRICING DEPOSIT INSURANCE FOR INDIVIDUAL BANKS

Depositors value the simplicity and certainty that deposit insurance provides, and banks benefit from being able to offer insured deposits to customers. Like other services banks use, a financial guarantee such as deposit insurance is not costless to provide. The starting point for the discussion in this section is that it is equitable and reasonable for a bank to compensate the parties who bear the risk of providing this benefit.

In this section we will consider ways of differentiating among the risk profiles of the more than 9,000 insured institutions currently in the 1A category for deposit insurance. The most straightforward conceptual approach is for a bank to pay an amount equal to the expected loss the deposit insurer faces from providing deposit insurance to that bank. An "expected loss" pricing system would 1) reflect the differences in risk across banks and 2) generate revenue sufficient to pay for the costs of insuring deposits. The expected loss price for a bank would depend on three things:

the probability of default for that bank;
exposure; and
severity, or loss given default.

At least in principle, every bank could be assigned something similar to a credit rating, with an associated range of default probabilities derived from experience. These default probabilities, combined with customized or standardized assumptions about loss given default, would yield the FDIC’s expected loss per dollar of assessable deposits, and the appropriate premium, for that institution. Such an approach also would provide the raw material to construct hypothetical distributions of FDIC loss exposure using standard credit risk measurement tools, as discussed in the following text box.

Oliver, Wyman & Company Approach to Bank Level Pricing

Expected loss pricing has two key benefits: at the systemic level, setting the price for each bank equal to its expected loss ensures that the premium inflows to the fund are ultimately equal to average long-term loss. This ensures that the fund is self-financing over time. Additionally, risk-based pricing helps to relieve moral hazard problems: Banks that try to use the availability of insured deposits to increase risk will be penalized through higher premiums.

Under Oliver, Wyman's proposed pricing structure, expected loss would be calculated "bottom-up" at the level of individual banks. This can be done by breaking expected loss for each bank into its components: expected default frequency; exposure; and severity. Approaches for estimating each of these building blocks are suggested below:

Expected Default Frequency

Expected default frequency (EDF) is the most significant of the expected loss building blocks.⁵ There are two basic approaches the FDIC could use to quantify the default probability of individual banks. The first method is a fundamental analysis of a bank’s condition, encompassing risk profile, financial strength, management, market position, and future prospects. The second method is to leverage the analysis of others by interpreting market price information.

Footnote: ⁵ EDF is "expected default frequency," or probability of default. The EDF is one of the three components that determine expected loss, which is described by the following relation:
Expected loss = (EDF) x (exposure) x (loss given default).

Fundamental analysis

A private sector company that underwrote deposit insurance would need to have a process for evaluating the riskiness of each customer. This process would likely resemble that which large banks use in evaluating the risks of trading counterparties and correspondent banks. Typically, such an analysis incorporates a number of elements, including the risk profile of the institution (credit, trading, asset/liability, event, and business risks), its financial resources (capital, reserves, subordinated debt, unrealized gains, sellable goodwill) and the quality of its management.

The FDIC already has access to many of the factors that go into a fundamental analysis, either through Call Reports or through the examination process. A key difference from existing CAMELS ratings, however, is how this information is processed. In order to be useful for pricing purposes, the fundamental analysis needs to be summarized in a score or grade which, in turn, is directly calibrated to an absolute measure of default risk (EDF). At many leading institutions, this is done by mapping the results of fundamental analysis to the external agency ratings of rated financial institutions. This mapping is then used to calibrate all scores (for both rated and unrated institutions) to their EDF equivalents.

At a minimum, such an approach can be undertaken with the elements of the existing CAMELS ratings. Oliver, Wyman would, however, advocate that the resulting grades should be more granular and provide for a much greater dispersion of exposure than the current risk buckets.

Market analysis

As an alternative (or supplement) to fundamental analysis, the FDIC could also seek to leverage market price information. In so doing, the FDIC would be outsourcing counterparty evaluation to the credit markets, using the premiums they are charging banks as an indicator of risk.

A number of market instruments could be used as barometers of credit risk, including:

(Uninsured) interbank deposits
Senior debentures
Subordinated debt
Interest rate swaps
Credit derivatives
Equity

In addition, the FDIC could encourage the development of new products for the purpose of providing market price information tailored to FDIC’s pricing needs. For example, banks could be required to purchase private deposit insurance in addition to that from the FDIC, either on a separate class of deposits, or as co-insurance to the FDIC. Recent proposals for a special class of subordinated debt could also serve as a market price indicator.

The main drawback to the use of market prices is that they are limited to banks that are large enough to attract an active market in their securities. While regulatory encouragement could increase the number of banks with appropriate instruments, it is likely that for the majority of smaller banks, even if they were to issue market debt, there would be little trading, and thus relevant price information. For practical purposes, a market-pricing regime would have to be limited to larger banks. Nevertheless, market prices could still be useful as external benchmarks of credit risk,particularly to establishing relative risk on the bases of observed premiums.

Exposure

Exposure measurement is the most straightforward of the expected loss building blocks. For each bank, this should be the amount of insured deposits. An alternative would be to base exposure measurement on total deposits (insured plus uninsured) on the theory that this would also capture exposure in "too big to fail" (TBTF) cases. However, since TBTF coverage would presumably affect only an unspecified number of large banks, insured deposits remains the best single measure of exposure.

Severity

Severity is the size of the loss to the FDIC, as a percentage of the total defaulted exposure to all insured deposits. While important, differences in the expected severity among banks are smaller than differences in EDF. We would thus recommend a simpler treatment of severity estimation. The FDIC experience shows that the severity of loss from small banks is usually greater than that for larger banks. This should be priced into risk-based premiums, either directly, or by investigating the underlying drivers of this difference. Factors which might prove to be good indicators of severity differentiation include business mix, loan concentration, and the structure of liabilities.⁶

Footnote: ⁶ Oliver, Wyman would normally recommend that the severity rates used in pricing be based on historical experience. However, the changes in early resolution brought on by FDICIA, if successful, should lower expected severity. Unfortunately, the experience of bank defaults since FDICIA is too limited to determine if this is the case, and if so, how much loss severity has been reduced.

The FDIC faces some practical constraints on its pricing of deposit insurance. There are limits on the extent to which risk distinctions can or should be made using either objective or subjective measures. Within those limits, however, setting deposit insurance premiums based on expected loss is over the long term likely to minimize the distortions and moral hazard problems associated with deposit insurance, and minimize the cross subsidization of the weakest banks by the strongest.

Conceptually, the question of how deposit insurance is priced at a point in time can be viewed in two ways. As discussed in the preceding text box, a "bottom-up" view would set pricing at the individual bank level and let overall revenue result from the sum of payments across banks. A "top-down" view would instead attempt to estimate appropriate aggregate funding needs and then allocate prices across banks based on risk.

The next section of the paper takes an aggregate perspective and discusses how insurance losses are funded over time. This section deals with how to assess risk at the individual bank level; the discussion here is consistent with either a bottom-up approach to pricing or a top-down approach that features a base price with adjustments for individual banks based on differences in risk.

Options for moving toward expected loss pricing or otherwise differentiating among the risk profiles of institutions in the 1A category can be broadly classified into approaches that rely on supervisory judgment, those that rely on other information, and hybrid approaches. A few specific examples of how some of these approaches could be used to develop "expected loss prices" for individual banks are included for illustrative purposes. We have not attempted to derive expected loss prices for every risk-differentiation option, but in principle one could do this. One could derive historical failure rates and loss rates for groups of banks and use such information to form the first and third elements of the FDIC’s expected loss discussed above. The third element, the proportion of loss borne by the FDIC, could either be based on a historical rule of thumb or, alternatively, could be based on a more sophisticated analysis tailored to an individual bank’s liability composition. (See Attachment E-3 for a full discussion.)

Some of the risk differentiation methods we discuss below are not amenable to historical loss analysis because historical information does not exist. Examples would include using large bank supervisory risk matrices more explicitly, using a newly developed supervisory risk rating, or using elements of non-public bank-specific information that have not been collected in the past. In most of these cases, the risk differentiation methods discussed below are best viewed as tools to help allocate assessments on a risk-related basis. In some cases, for example, by differentiating risk based on specific elements of risk-related information, and depending on the information collected, one could conceivably develop estimates of expected loss at the bank level as described in the preceding text box.

Because all of the methods considered have the potential to distinguish more effectively among A-rated banks, the likely result regardless of the method selected would be that the safest banks pay less over the long run.

Chart 1

Options Relying on Supervisory Evaluations

The case for relying on supervisory judgment as the cornerstone of any attempt to enhance risk differentiation is easy to state: onsite examination provides the most in-depth information available. The options for such an approach include:

Composite CAMELS ratings;
Components of the CAMELS ratings;
Risk matrix results; and
Uniform risk management ratings.

Composite examination ratings. It would be quite simple to achieve additional risk differentiation by application of existing supervisory tools. Specifically, the FDIC could differentiate for insurance purposes between banks with examination ratings of 1 and 2. This would be clearly supportable from a risk

perspective given the relative historical failure rates of these two groups of banks. As shown in Chart 1, the 5-year failure rate for CAMELS 2-rated institutions since 1984 was more than two-and-a-half times the failure rate for 1-rated institutions.

Component ratings. Other currently used supervisory tools could readily be put to use for pricing deposit insurance. For example, the component ratings in the CAMELS system could play a greater role. The component ratings are ratings of the individual elements of the CAMELS acronym that are assigned in the examinations: capital, asset quality, manage-ment, earnings, liquidity and sensitivity to market risk. The component ratings range from 1 to 5, and 1 is the best rating. There is at least conceptual appeal to greater use of these ratings for setting premiums. Purely to illustrate the concept, one could imagine, for example, a rating of "2 minus" for 2-rated banks with a sufficient number of components rated 3 or worse.

Table 1

One concern with this approach is that greater reliance on subjective information could compromise consistency in determining premiums, and more so the finer the distinctions that supervisors are asked to make. This is a general concern that applies to all options involving supervisory information and is discussed further below.

Risk matrix results. For the largest banks, results from the Office of the Comptroller of the Currency (OCC) and the Federal Reserve Board risk matrices could be used more explicitly. As a supervisory tool, large complex banks supervised by the OCC and the Federal Reserve Board are assigned ratings for a variety of risk-management factors. Banks with the same composite CAMELS rating can have substantial differences in these risk ratings and these differences do not affect the premium paid by the institution. One could imagine assigning scores based on these matrices. Again, purely to illustrate the concept, a "2 minus" could be a bank with a sufficient number of high or rising key risk elements in these matrices.

An impediment to this approach is that there is a lack of uniformity at the agency level, with the OCC and Federal Reserve using different matrices and the FDIC and Office of Thrift Supervision not using a matrix approach. An example of the OCC’s Risk Matrix is displayed in Table 1.

Uniform risk management rating. If we consider approaches that require more time to implement, the possibilities expand. For example, one could imagine an interagency effort to develop a uniform risk rating explicitly designed to differentiate among the levels of risk implicit in a 1 or 2 rating. Such a rating could consider such factors as the extent of credit concentrations and the quality of underwriting standards and internal controls—irrespective of the current financial performance and condition of the institution—and would essentially gauge the institution’s overall appetite for risk. This approach could be applied to all institutions, or, alternatively, only to the largest institutions. One concern here is that designing a uniform interagency risk-rating could take years.

There are arguments against basing risk differentiation on supervisory judgment. The examination process might require a greater degree of management discussion in order for examiners to work constructively with banks to obtain needed information, and bankers may have discomfort with additional supervisory discretion. Examiner judgments are subject to several layers of review and an elaborate appeals process, and one of the disadvantages of increasing the reliance on supervisory judgments for pricing deposit insurance could be an increase in the resources bankers and examiners might have to devote to such processes. It could also be argued that some of the tools described above—notably the individual component ratings and the large bank risk matrices—were intended as subjective supervisory tools that should not be put to uses for which they were not designed.

For the 99-plus percent of insured institutions that do not have a continuous onsite examiner presence, there also may be significant issues of timeliness in relying on examiners to differentiate the risks taken by the vast majority of well-performing banks. Most banks are examined on an eighteen-month cycle. The issue concerning the timeliness of onsite data is reduced in larger insured institutions that are subject to ongoing targeted examinations or a more frequent examination cycle.

Changes in risk profiles between exams presently are addressed using offsite monitoring tools, with an onsite exam or other supervisory review as needed; in particular, a downgrade of a bank to a 3-rating is generally validated with an onsite examination. If the premium system were structured to differentiate more among the 9,000-plus institutions not currently subject to heightened supervision, and if changes in an institution’s risk category required an onsite supervisory review, resource demands on the supervisory process would increase.

Options Relying on Objective Factors

Another general method to consider would be "factual" or data-driven approaches. Several sources of information could feed such approaches: non-public bank specific information, bank Call Reports, or market information.

Non-Public Bank-Specific Information

The federal banking regulators have access to a significant amount of non-public information about insured institutions, information that may be a useful supplement to Call Report data for purposes of evaluating institution risk profiles. It may be possible to develop a basic package of objective information that would assist in the process of differentiating the risk profiles of the majority of strong-performing institutions. Much as premiums are now assigned based not only on supervisory evaluations of capital adequacy, but on the capital ratios themselves, one could imagine the FDIC differentiating risk profiles directly from non-public information provided by banks.

The Canada Deposit Insurance Corporation (CDIC) created a risk scoring system in 1999 that is based, in part, on regulatory reporting information and partly on information provided directly to the CDIC by its member institutions. The score incorporates information on a variety of risk-related factors including capitalization, significant credit concentrations, and self-certifications of the extent of adherence to defined risk-management standards. The standards adherence information is submitted by the banks annually to the CDIC, with a copy provided to each institution's primary supervisor. CDIC assigns banks to one of four risk categories for insurance purposes based on both the supervisor’s examination rating and the information banks provide. (Banks that do not provide information are put in the worst premium category.) The algorithm for assigning banks to categories based on the information provided is known to the member institutions. Attachment B-1 summarizes the major features of the CDIC risk scoring system.

The CDIC’s experience has been that its scoring system provides a reasonable balance between qualitative and quantitative factors, and that it is relatively transparent and easily understood by member institutions. The doubling of premiums payable resulting from a drop from one category to the next appears to provide incentives to member institutions to either improve their financial results or their adherence to the standards.

The CDIC faces issues with its system that undoubtedly would be faced under any similar scoring system that might be devised. The selection of a particular scoring methodology creates the possibility that institutions focused on maximizing their scores could achieve the best premium category, but might still be considered risky by other measures that are not encompassed by the system. This issue, and the related question of how, and how often, the information the insurer receives is to be verified, are addressed in part through the supervisory examination process. The details of how that occurs are a subject of continuing discussion between the CDIC and the primary supervisor, as would likely also be the case if the FDIC were to implement such a system. There is also the tradeoff between reporting burden and comprehensiveness of results. For example, the quantitative part of the CDIC’s system includes very little off-balance sheet information, as it was felt that such inclusion would require excessive filing requirements.

The CDIC’s approach is one among many that could be devised along similar lines. The design of a risk scorecard could be exposed both to public comment and the expertise of industry practitioners in both small and large banks.

The advantage of this approach potentially would be in using more detailed risk-related information without imposing a regime where supervisors are asked to make subjective distinctions among healthy banks. Moreover, it could avoid the resource and timeliness issues, described above, that could arise if supervisors were asked to monitor inter-examination changes in risk profiles for over 9,000 banks along a more finely graduated scale than is now required.

Such an approach could raise concerns about the burdens of creating another layer of bank reporting. Those concerns might be allayed if the risk scorecard were either simple, or built on information that is readily available to a well-managed bank. In this regard, a comparison with developments in capital regulation may be appropriate. The momentum towards basing large-bank capital requirements on internal credit ratings accommodates differences in banks’ internal ratings scales, provided those ratings can be mapped to a common rating scale. Thus it may allow for both flexibility at the bank level and analytical rigor in the setting of capital requirements. This new direction in capital regulation is an example of a systematic use of non-public bank-specific information for a significant policy goal. Deposit insurance pricing may similarly be able to benefit from such a systematic use of non-public information.

Bank Call Reports

Better differentiation of risks in the premium system using public financial reports would be conceptually straightforward. There is considerable variation in the financial characteristics of the 9,000-plus insured institutions in the highest assessment category. Using figures reported by banks for year-end 1999, Table 2 displays differences between 1A-rated banks in the top 10 percent on several performance factors and those ranked in the bottom 10 percent. (Attachment C explains the nine risk classifications currently used by the FDIC to assign premiums.) The table reveals substantial differences between these groups that are not reflected in their assessment ratings, since all are rated 1A and currently pay nothing for deposit insurance.

Table 2

(See Attachment D-1 for more details on the dispersion of capital ratios for institutions rated 1A.) As described below, the difficult questions relate to choosing from a multitude of available approaches, and evaluating the usefulness of the results.

A number of financial ratios have been shown to be indicators of the potential for future financial distress, examination downgrade or failure.⁷ An FDIC study of the banking crisis of the 1980s found that high loan concentrations, rapid loan growth and high dependence on volatile liabilities were significantly associated with higher probabilities of failure (FDIC, 1997).

Footnote: ⁷ O'Keefe and Reidhill (1997); Demirguc-Kunt (1991); Gajewski (1989); Whalen and Thomson (1988).

Suppose we identify a set of ratios for which higher values often indicate higher risk, other things equal. One example of how to use these ratios for pricing would be to use peer analysis to identify outliers. (See Attachment D-2 for an example.)

As another example, a bank’s financial ratios might be used as inputs to a statistical credit score designed to estimate failure probabilities based on historical experience of banks with similar characteristics. The resulting score could be used as an indicator for the bank’s premium. Both examples are presented for illustrative purposes only, and represent only two among many potential methods for using reported financial information.

Using Call Report data to better differentiate the risks of the majority of strong-performing institutions has a number of attractive features. Such data are uniform in format and regularly available. At least in theory, such information is objective. And there is a vast body of analytical work to draw from that is expressly designed to measure risk of failure using such information.

One of the issues that would need to be addressed in applying Call Report ratios is whether to use peer comparisons or absolute ratio thresholds. Under the peer approach, an institution could be reclassified because it does poorly relative to its peers, even if all of its ratios are strong by historical standards. Moreover, using purely relative comparisons would make it difficult for bank managers to determine exactly where their institutions would be classified for insurance purposes prior to their actual classification. The alternative is to use absolute benchmarks based upon historical averages. This gives bankers explicit targets, should they choose to shoot for them, and ensures that more institutions move into lower premium categories when the industry is stronger. (See Attachment D-3 for an example of this approach.)

On the other hand, to the extent that the goal of making additional risk distinctions is merely to ensure that higher-risk institutions bear a greater share of the costs of deposit insurance than lower-risk institutions, the purely relative comparisons accomplish this. In driving pricing off such peer-based ratio tests, however, attention should always be paid to the absolute levels of the ratios, in order to avoid making pricing distinctions based on negligible or economically insignificant differences in financial ratios.

Reported information at times has been notoriously inaccurate. The FDIC’s most costly bank failures in recent years have occurred rather abruptly among institutions that had consistently reported strong earnings and capital. In these cases, an examination or another event ultimately revealed that reported earnings had been artificial and overstated while asset values had been inflated unrealistically. In some cases, Call Report-based offsite tools were indicating that these institutions should be candidates for upgrading from a CAMELS "2" to a "1" at the same time that examiners were placing the institutions on the problem list.

Another significant limitation of Call Report information is that it is not detailed enough to fairly compare the risk profiles of insured institutions. As a simple example, consider two institutions whose Call Reports indicate identical concentrations of consumer and residential mortgage loans. One of these institutions could be specializing in subprime loans, and the other in conservatively underwritten loans, but the Call Reports would not show the difference (except perhaps by inference based on loan yields or other indicators). One institution could have significant commercial lending concentrations to a few large counterparties, industries, or geographic areas, while another, with the same Call Report numbers, may be prudently diversified. Call Reports are useful in capturing some types of quantitative data, but despite past revisions, fail to capture certain qualitative factors that also merit consideration. Reported financial information does not provide a picture of the risk profile of the reported loans, the quality of internal controls, or, in some cases, the magnitudes of market-sensitive decisions management has made.

Market Information

More than 750 insured institutions or their holding companies, holding over 50 percent of all insured deposits, currently issue debt or equity instruments that are traded in organized financial markets. As indicated in Chart 2, market prices for these instruments appear to reflect a changing aggregate risk profile over time as well as variations in risk across institutions.

Chart 2 shows the mean and distribution of subordinated debt yield spreads over U.S. Treasury securities with comparable maturities for the period from January 1997 to June 2000. The lower and upper bars of the graph represent the 10^th and 90^th percentile cutoffs, respectively, and are included to show the degree to which spreads are dispersed around the mean. Yield spreads increased significantly over the period, and were particularly volatile in 1998. These widening spreads, in part, may reflect growing investor uncertainty regarding credit risk. Moreover, the expanding spread between the yields at the 10^th and 90^th percentiles suggests that market participants have perceived an increasing disparity among individual institutions in recent periods. This is corroborated by the disparity in EDFs that are based upon asset price volatilities, in Chart 3.

Chart 2

Yield spreads and EDFs can be useful because they convey information derived from prices paid in efficient markets. However, it is not always clear how to interpret these market signals. For example, the sharp increase in yield spreads and EDFs following the Russian bond default that occurred in August 1998 has been widely interpreted as a general shift in the level of risk aversion on the part of investors. Similarly, a perceived scarcity of Treasury securities due to decreased issuance in maturities longer than 3 years appears to have driven down long-term Treasury yields during the first half of 2000, even as short term interest rates were rising.

Conceptually, market prices or credit ratings could be used to identify the level or change in the default risk posed by the issuing institutions, thereby differentiating higher- and lower-risk institutions for the purpose of assigning premiums.

Credit ratings. Our earlier discussion of expected loss pricing was couched in terms of a credit rating framework. The natural question then becomes, why not use credit ratings directly to estimate expected loss deposit-insurance premiums for those institutions that have such ratings? (A discussion of this possibility is contained in Attachment E-1.)

If the goal is to differentiate among acceptable levels of risk, credit ratings are explicitly designed to do so. This is a plausible and straightforward approach but one that raises a number of issues. Not all institutions have credit ratings, but as described below, the FDIC has authority to establish separate insurance pricing mechanisms for large banks, which are most likely to be rated by one of the recognized rating agencies. There are other questions. For example, what debt instruments have payoff characteristics most closely resembling the FDIC’s exposure? Does the credit rating reflect a belief in an implicit federal guarantee that may bias the rating upwards, and result in better ratings for the largest banks? The use of credit ratings from third parties may also raise issues about the appropriateness of a government agency relying on information provided by specific private firms for important public policy decisions, especially when those firms may have less information or experience than the government agency on which to base their judgments.

Subordinated debt. The FDIC currently monitors the spreads individual institutions pay over comparable U.S. Treasury obligations for various debt instruments. Referring again to Chart 2, it is clear that the mean yield spreads on subordinated debt vary over time and that the market differentiates observably among institutions in its pricing of subordinated debt. One could envision a system in which institutions with the highest spreads are classified into a higher-risk category for premium purposes. Attachment E-2 contains a discussion of this issue.

Subordinated debt has received considerable attention as an instrument that could enhance market discipline for large institutions and convey early warning signals to regulators. Subordinated debt is regarded as particularly attractive because the incentives of subordinated debt holders line up with those of the FDIC; these debt holders do not share in the upside from any gambles taken by bank management, and subordinated debt pricing tends to discourage imprudent risk-taking.

Chart 3

Owing to these characteristics, there have been several proposals to require large institutions to issue some minimum amount of subordinated debt on a regular basis. The Gramm-Leach-Bliley Act requires the Treasury Department and Federal Reserve Board to conduct a study of subordinated debt requirements, and the report is due to Congress on May 12, 2001. One of the primary arguments against a subordinated debt requirement is that it interferes with management decisions regarding the optimal liability structure for their institution. Subordinated debt is not widely issued at the bank level, and even where it is, it may be in amounts or on terms different from those that would be mandated by a regulatory requirement.

There are at least two broader sets of issues that arise in connection with the use of market data for deposit insurance premiums. One is that the use of market data would effectively create two pricing mechanisms for deposit insurance, one for larger, publicly traded institutions and another for smaller privately held institutions. The second issue is the difficulty of extracting the appropriate information from market prices. Price movements often reflect more than pure changes in the individual default risk of the issuing institution. For example, price changes can reflect developments in the broader economy or the financial markets that influence the supply or demand for many types of instruments. In analysts’ terms, the "signal-to-noise" ratio associated with price changes is sometimes low, and extracting the true market signal regarding the institution may not be straightforward. In addition, even after the true market signal about an individual institution has been extracted, the result may not correspond to supervisory evaluations of that institution.

The differences between large and small institutions have been growing for some time. Large institutions have increasingly complex risk profiles, global operations, and expanding lines of business, and they are subjected to market scrutiny in an increasingly competitive environment. Small institutions remain more community-based, focused on a limited number of core businesses, and privately held. In recognition of these differences, FDICIA explicitly authorized the establishment of two distinct premium systems based upon the size of institutions. The FDIC has thus far not exercised this authority, given that the 1996 statutory constraint effectively precludes any meaningful distinctions. A well-established movement exists within the bank regulatory community toward separate approaches to supervision and capital regulation for large and small institutions, and it is time to reconsider whether a "one-size-fits-all" approach to deposit insurance pricing will remain suitable going forward.

Hybrid Approaches

We have differentiated between subjective and objective approaches. Our current pricing system uses both, namely examination ratings and capital ratios. (Attachment C provides an overview.) Market prices are likely to incorporate both subjective and objective information, as just discussed. Other hybrid approaches could be considered. For example, the subjective factor could continue to be based on the examination rating, with the objective factor being some type of risk score based on Call Report ratios, market risk indicators, or other non-public information.

All three types of information considered in the preceding sections could be combined into a scoring system for determining premium classifications. The approach used by the CDIC is another example of a hybrid approach. Attachment B-2 illustrates the results of applying a CDIC-style approach to U.S. institutions using a subset of variables. The CDIC system is more complicated than the FDIC system in some respects, but it is also simpler in that it contains only four premium categories.

Customized Financial Contracts

One possible way to use market information to differentiate risks without imposing a particular funding structure on insured institutions might be to go beyond simply monitoring capital markets and begin entering into financial contracts that price and share the risk of failure at individual institutions. The FDIC could enter into financial contracts that, in exchange for a premium paid to the holder, expose the holder to a defined risk in the event of the failure of a specific institution or pool of institutions. The premium that holders require on such contracts provides information relevant to expected loss pricing.

In 1991, the FDIC was granted the authority to "obtain private reinsurance covering not more than 10 percent of any loss the Corporation incurs with respect to an insured depository institution."⁸ The FDIC completed a study in June of 1993 that found that the conceptual attractiveness of a reinsurance program was offset by the complexity of the practical and public-policy issues that would first have to be addressed. Of particular importance was that a market for such risk did not exist at that time and, as a result, the terms under which such coverage could be obtained were not favorable to the FDIC.

Footnote: ⁸ 12 U.S.C.A § 1817(b)(1)(B).

In the last few years, however, financial innovation has greatly expanded the range of nontraditional alternatives in which the capital markets can be employed to finance risk. Examples that may be useful for the purposes of pricing deposit insurance include collateralized loan obligations, credit derivatives (default swaps) and other structured securities. The FDIC could work with market practitioners to explore the feasibility of using such instruments for price discovery, as an input to premium setting.

Making use of market information in this manner can address bankers' concerns regarding subjectivity, given that market prices reflect the collective judgment of diverse, informed parties with personal wealth at stake. Market prices also are inherently forward-looking, and they may serve as a check on any inefficiencies in the FDIC pricing process. For example, through market pricing the FDIC may learn that some reporting requirements have low value-added, given other information, and industry burden could be reduced by eliminating such requirements without sacrificing accuracy in price setting.

Similarly, market information could help to reduce the distortions that can arise when government-administered prices are introduced into otherwise competitive markets. Market participants can be expected to use both subjective and objective information from several sources to price risk—any information that contributes to more accurate pricing.

III. FUNDING DEPOSIT INSURANCE LOSSES

Funding arrangements play a critical role in the design of a deposit insurance system. A well-designed system will ensure that adequate funds are readily available to respond to problems as they arise; inadequate funding can lead to delay in resolving failed institutions and significant increases in costs. The design of the funding arrangements will determine whether the industry is asked to pay for the costs of deposit insurance when the industry is healthy or when it is experiencing problems.

From 1934 until 1950, all FDIC-insured banks were assessed at an annual rate of eight-and-one-third basis points per dollar of domestic deposits. This revenue went into an insurance fund where it earned interest and was available to meet operating expenses and pay for the costs of insuring depositors. Under this fixed premium rate approach, the size of the fund depended on the extent to which revenues and interest income exceeded expenses. A key feature of this system was that banks were required to pay annually for deposit insurance. The fact that the premium rate was stable and the fund grew when the economy was healthy allowed the system to smooth the costs of deposit insurance over time.

One concern with a fixed-premium approach is that the premium rate might prove to be too high or too low. At an aggregate level, this could mean that over time the industry would either be over- or under-charged for deposit insurance. As it happened, in 1950 Congress addressed the industry’s concern that the insurance fund had grown too large by requiring the FDIC to return a portion of excess premium revenue each year. Thus, the effective premium rate was tied to the current year expenses of the deposit insurance system, and could range from slightly more than 3 basis points to 8.33 basis points.

This system was in place until 1989, when the earlier concerns about over-charging were replaced by concerns about under-charging as the insurance fund declined. In the aftermath of taxpayers funding FSLIC losses, Congress addressed concerns about the viability of the surviving funds by significantly changing the assessment system. A DRR was established and premiums depended on whether the reserve ratio was above or below this target. Under this system, which for the first time gave the FDIC some discretion over rate-setting, the effective premium rate could range from 0 to 32.5 basis points, with increases in any one year limited to no more than 7.5 basis points. By 1991, the premium rate had reached 23 basis points.

As noted in the introduction of this paper, in 1991 FDICIA brought further changes by providing broad discretion to the FDIC to achieve two mandates: establishing a risk-based premium system and maintaining the funds at the designated reserve ratio.⁹ The DIFA significantly curtailed that discretion when the funds are above their targets.

Footnote: ⁹ FDICIA included additional mandates to help prevent future crises, such as "prompt corrective action" requirements for bank supervisors to ensure early supervisory intervention for deteriorating institutions, and a " least-cost resolution" requirement to control the costs of resolving bank failures. As with other provisions of FDICIA, these have not yet been tested by adverse economic conditions.

As a result of these changes, the original system with a focus on a steady long-term premium rate has been replaced by a system with a focus on a target fund ratio. The current system has an adjustment mechanism whereby as banks' condition deteriorates, they pay more into the fund. Current arrangements reflect the desire to ensure that taxpayers will be protected from deposit insurance losses. At the same time, it results in what might be called a pay-as-you-go system. During good times, banks pay an insignificant amount; during bad times, the cost of bank failures are passed through to banks when they can least afford it. Losses are determined after the fact and survivors are asked to pay.

Another key question that drives the discussion of aggregate funding is whether banks should be able to receive disbursements from past premiums. If one views the government as bearing all the risk of bank failures with banks paying a "user fee" to compensate the government for doing so, then the answer is no. Another view is that if the government bears only extreme or catastrophic losses while the industry bears losses up to that point through a mutual arrangement, then banks should have some explicit claim on past premiums.

Thus, the options discussed below are organized under two broad headings: user fees and mutual arrangements. Under the user fee, there are two general approaches. The first relies on relatively steady average premium rates designed to equate premium revenue with insurance losses over a long-term horizon. The second alternative is to allow for more variation in the average premium rate by adjusting the rate based on current insurance losses or by linking the rate to the reserve ratio of the deposit insurance fund.

The options under the mutual arrangement heading include: 1) rebates tied to the reserve ratio; 2) a system in which banks hold explicit claims on the insurance fund; or 3) a system which more closely resembles private market provision of deposit insurance.

User Fee Model

As mentioned above, a user fee approach would view the government, not the banking industry, as the provider of deposit insurance and therefore the party responsible for bearing the risk of guaranteeing bank deposits. Under this approach, the industry would pay on a regular basis for access to the deposit insurance system. Because the payment would be viewed as in exchange for something of value, the industry would have no claim on previous payments. This is often compared to private insurance; a driver who does not have an accident does not get his money back.

Long-Term Premium Rates Based on Historical Experience

In the simplest case, the industry would pay a stable average premium rate either set in statute or subject to infrequent change at the discretion of the FDIC. If revenue needs (losses) in a given year exceeded the revenue collected, the government would be responsible for the difference. Likewise, excess revenues would accrue to the government.

One benefit of stable premium rates may be a lower cost of capital for the banking industry. If the volatility of deposit insurance losses were passed directly through to the industry, this would add volatility to bank earnings and the market may discount those earnings. Shaffer (1997) estimates that, based on past FDIC loss experience, steady premium rates could lower the banking industry’s capital cost by $1 billion to $4 billion per year. This is equivalent to additional yearly premiums of 3 to 13 basis points for BIF-insured institutions on a pre-tax basis, or approximately $7 to $29 billion in potential lending that might otherwise occur.

Obviously, a key issue under the stable rate approach is the level at which average premiums are set. The correct rate would depend on the risk of loss faced by the government over a long-term horizon. The main concern with an inflexible rate is that the correct rate will vary over time with changes in industry structure, regulatory regimes, and the competitive environment.

The practical implication of this approach is that policymakers must decide how to set the appropriate average stable rate. In the 1930s, the choice of 8.33 basis points was guided by a review of bank failures from 1865 to 1934. If we choose to look to historical experience as a guide in choosing the appropriate price, the key decision is what time period is most relevant. This is a subjective judgment about how the future will compare to past experience. Table 3 shows the rates necessary to cover operating expenses and insurance losses over various time periods for the BIF.¹⁰

Footnote: ¹⁰ The table presents the rate that would need to be charged on total domestic deposits in order to make total insurance revenues equate to total expenses and losses over the specified time period. The calculations assume that there is no fund balance at the beginning of the time period and that the only income to the fund is assessments charged on deposits. Expenses include operating and administrative expenses plus estimated losses. Deposit and expense figures are taken from the FDIC Annual Reports. The experience of the FSLIC (the prior insurance fund for savings-and-loan institutions) is not included in this analysis. Data are incomplete. Moreover, the 1980s savings-and-loan experience is considered less relevant for today's banking and thrift industries, given the unique balance sheet structures of savings and loans in the earlier era, differences in accounting rules, and other factors that no longer apply to insured institutions.

Attachment F-1 compares the results of a fixed rate approach to actual results for the period 1982 to 1999.

Moving Average Approach

Instead of having Congress or the FDIC Board adjust the rate on an ad hoc basis, an alternative would be a mechanism for adjusting the rate incrementally over time to reflect experience or changes in expectations of future insurance losses.

Table 3

One method would be to use a long-term moving average of insurance expenses as the basis for setting the average premium rate. The moving average would allow premiums to reflect actual insurance expenses, while the long-term horizon would result in gradual changes in premiums over time. This approach has been suggested by Konstas (1992) and Shaffer (1997). Chart 4 shows the moving average of insurance expenses over different time horizons.

Table 4

The annual average premiums that would have prevailed over different periods using various moving averages are set out in Table 4.

Chart 4

One of the concerns with such an approach is that it will result in banks paying higher premiums as a result of banking problems that occurred far in the past. Moreover, this approach could result in extended periods over which the deposit insurance system is not self-financing, and it is unclear whether this would be politically acceptable.

Analytical Approaches

The moving-average approach relies on past performance to set the assessment rate. While this has the advantage of simplicity, it ignores relevant information about potential future losses. An alternative would be to incorporate relevant financial, supervisory, and market information. The previous section of this paper discussed how such information could be used to develop expected loss pricing at individual institutions and how such information could also serve as the basis for aggregate pricing. The very preliminary analysis contained in the Oliver, Wyman & Company report resulted in an expected loss figure for the BIF that translated into slightly more than 5 basis points of assessable deposits; again, this is a highly qualified result. (See the following text box.)

Oliver, Wyman & Company Analytical Framework

Oliver, Wyman's suggested approach for evaluating policy options is to apply analytical tools and methodologies that have been developed for analyzing risk and capital management in banks and other financial institutions to the deposit insurance system. The cornerstone of this approach is to model the loss distribution of the FDIC insurance funds. The loss distribution can then be used to evaluate the appropriate level of fund adequacy and reserving in terms of a stated confidence interval or solvency standard. Furthermore, this analytical approach can be applied to identify pricing options that are consistent with market practice.

While the proposed analytical framework may seem like a novel approach for evaluating the risk of the deposit insurance system, this approach is increasingly the best practice among leading bank and non-bank financial institutions in the U.S. and abroad. In fact, efforts to set capital in relation to an explicit model of a financial institution’s risk profile lie at the core of the current proposals for reform of the Bank for International Settlements' (BIS) capital rules. (See, for example, Jones and Mingo, 1998). Oliver, Wyman's approach applies the same types of methodologies that are under consideration by the BIS Models Task Force to the FDIC’s own loss distribution for resolving similar questions of risk and capital management.

Modeling the Loss Distribution

The first step in analyzing the FDIC’s risk profile is to recognize that the deposit insurance funds are portfolios of credit risks. These portfolios consist of individual exposures to insured banks and thrifts, each of which has a small but non-zero chance of causing a loss to the fund. Such a portfolio is similar to a bank loan portfolio, although the nature of the underlying risks in the FDIC funds raises unique issues.

Chart A shows the typical credit loss distribution for an insured bank. The distribution is characterized by the portfolio’s level of average, or expected, loss; by the size or concentration of individual exposures; and by the correlation among loans in the portfolio. Unlike a normal distribution, the bank credit distribution is heavily skewed: It has a long right tail, meaning that most often, losses are relatively small, but there are cases in which large losses may arise. In order to protect against these losses, a bank is required by regulators (and by rating agencies, uninsured depositors, and other creditors) to hold capital to cover the potential for loss at a high degree of confidence. In this case, the bank holds capital up to the 99.9 percent confidence interval.

If the losses exceed this point, then the bank will become insolvent. To the extent that there are insufficient funds available to repay insured deposits, then the excess deposit losses will be borne by the FDIC.¹¹ Put another way, the FDIC assumes the residual "tail" risk of loss to insured deposits. For the individual bank considered above, the FDIC’s risk profile is shown in Chart B. There is a high probability of no loss, and the residual tail probability of some loss to the fund.

Footnote: ¹¹ In the event of default, history tells us that some of the losses are "recovered." As such, the loss to the insurance fund may be a fraction of what the tail suggests. This is commonly referred to as "severity," or loss given default (LGD). The severity should be taken into account when we move from Chart A to Chart B.

The FDIC’s exposure to individual banks can be added together to create a cumulative loss distribution. Just as with a bank’s credit loss distribution, the FDIC’s cumulative loss distribution will reflect the expected loss of the individual insured banks; the size of individual exposures; and the correlation of losses in the portfolio. Chart C shows conceptually what the cumulative loss distribution for the deposit insurance funds should look like. The distribution will be heavily skewed, with a high probability of very small losses to the fund, but a significant probability of large losses. The potential for large losses will result, in part, from the presence of large banks in the portfolio. The "lumpiness" in the distribution reflects the contribution of individual large banks, each of which imposes a discrete, non-zero probability of a sizeable loss to the fund.

Chart A: Bank Risk Profile

Chart B: FDIC Profile for Single Bank

Chart C: FDIC Cumulative Loss Distribution

Adjustments for Current Insurance Expenses

Whether a long-term pricing mechanism is based on past loss rates or forward-looking analytical techniques, the concern remains that the approach might lead to significant buildups of the insurance funds or prolonged periods of losses to the funds. One way to address this is to tie pricing more closely to current performance. As mentioned earlier, the premium system in effect from 1950 until 1989 featured a statutory rate of 8.33 basis points with a refund of a portion (60 percent) of excess revenue in a given year. The statutory rate was expressed as a maximum rate with downward adjustments for current perform-ance.

A variation of this approach would be to allow an upward adjustment to premium rates in the event of a revenue shortfall. Had such a system been in effect, premium rates would have risen during the 1980s as insurance losses mounted. Table 5 (see next page) shows the rates that would have resulted assuming that any revenue shortfall up to 5 basis points could be charged in a given year. In other words, premiums could rise as high as 13.3 basis points to cover shortfalls or fall as low as 3.3 basis points to reflect the overage of collections above insurance expenses. The table also shows the assessment rates that would have resulted from the current assessment scheme had it been in effect as of 1982. A description of the simulation and methodology is contained in Attachment F-2.

Linking Premium Rates to the Insurance Fund

Under the approaches discussed thus far, the premium rate is based on a long-term view of expected losses, perhaps adjusted for current experience. With these approaches, the insurance fund is free to fluctuate in response to insurance losses, and movements in the fund do not trigger changes in the premium rate.

All the pricing methods above are subject to the concern that the rate-setting mechanism could be biased. If this were the case, over time the fund would come to reflect that bias: excessive rates would result in a fund that was "too large"; insufficient rates could result in a negative fund balance. One way to address the concern about a fund that is too large or too small is to link rates to the size of the insurance fund relative to a measure of fund exposure, such as the reserve ratio.

At this point it is helpful to discuss what an insurance fund means under a user fee system in which the government bears the risk of losses in excess of current revenue. Some have argued that it makes little difference whether there is an insurance fund, given that the government is ultimately on the hook for losses. Feldman (1999) proposes a system in which the insurance fund is replaced by a standing appropriation authority.

Table 5

On the other hand, there are practical reasons for the government to maintain an explicit insurance fund. The insurance fund helps to protect taxpayers from deposit insurance losses by creating a buffer paid for by the industry. The insurance fund also can be viewed as a federal budgeting mechanism to sequester deposit insurance resources from the normal appropriations process. This can help to do two things: first, ensure that adequate resources are readily available when problems arise, thus avoiding potentially costly delay; and second, help to smooth the costs of deposit insurance over time. The size of the fund and the loss distribution determine the probability that a special call on industry capital or an appropriation would be needed. These relationships are considered further in the accompanying text box.

Solvency Standard – Oliver, Wyman & Company

The cumulative loss distribution allows the potential for loss to be directly compared with the reserves and other resources available to the insurance funds.¹² This analysis can also be reversed to determine what level of resources—or reserve ratio—is required to reach a chosen solvency standard. A solvency standard is the desired minimum confidence that the fund will be sufficient to make payments on all its obligations. This can also be expressed in terms of a maximum default probability. One common way of describing solvency standards is in terms of the equivalent credit rating from a major agency. Table 6 shows the one-year default probabilities consistent with each of the rating categories from the two leading agencies, Standard & Poor’s (S&P) and Moody’s. For example, maintaining an investment grade status (BBB-/Baa3 or better) is equivalent to holding sufficient reserves to reduce the one-year default probability to below 0.32 percent. That is, an investment grade quality fund must be at least 99.68 percent confident of being able to meet all of its obligations in the coming year.

Footnote: ¹² The first resource is the expected income from premiums and interest. In most cases, this will be more than sufficient to cover the losses and the fund will have a gain. The next resource is the current balance of the fund itself. If the losses in any period exceed the funds available, a backstop resource, such as a loan or grant from the Treasury, is required to ensure the payment of all obligations.

By creating an explicit link between the potential for loss and reserves, the FDIC can consider the appropriate level of fund adequacy in terms of both a stated confidence interval and market equivalents. For example, just as an individual bank chooses to capitalize according to a desired credit rating, so too can the FDIC choose to capitalize the insurance funds to a desired rating. This approach differs fundamentally from the current system, in which the DRR is set as a fixed percentage independent of the fund’s actual loss profile.

Table 6

*The calculated default probabilities reflect the methodology of Oliver, Wyman.

Setting a Hard Target for the Insurance Fund

A hard target for the insurance fund would mean that premium rates would adjust quickly to changes in the reserve ratio. This is similar to the current system, in that the premium rate effectively drops to zero when the fund is above the target, and rises to a high level, 23 basis points, if the reserve ratio is not expected to return to the target within a year.¹³

Footnote: ¹³ See 12 U.S.C.A. § 1817(b)(2) (West Supp. 1999).

The advantage of keeping the reserve ratio at a target is that it helps to ensure that there is always an adequate buffer between deposit insurance losses and the taxpayer. It does this by giving the FDIC a call on this industry's capital to the extent necessary to maintain the fund ratio target.

The disadvantage is that it can result in long periods where deposit insurance is essentially free followed by short periods where premium rates are extremely high. This pattern makes it virtually impossible to implement a risk-based premium system with expected loss pricing.

Extended periods with zero premium rates also allow deposits to enter the system without contributing to the insurance fund. Some have suggested that this could be addressed by introducing a surcharge for rapid growth or large deposit flows. While this would address free-rider concerns, there are several drawbacks to such an approach. First is the practical question of deciding what constitutes rapid growth or large deposit flows. Applying surcharges on deposit growth in excess of some percentage or dollar volume will result in rather arbitrary distinctions and create incentives for manipulating the system.

Second, deposit growth will often simply reflect the healthy innovative behavior necessary to serve customers in a competitive market. This is true whether the growth is from newly chartered institutions, pioneers in electronic banking, or large financial services firms delivering the benefits of financial modernization. It may be difficult to structure surcharges that address concerns about fairness without inappropriately stifling such innovation.

Setting a Soft Target for the Insurance Fund

There could be funding arrangements designed to maintain a target reserve ratio over time without introducing volatile swings in premium rates. A soft target approach would allow the reserve ratio to return more slowly to the target, thus providing for more stable premiums. The premium rate could vary depending on how far the reserve ratio was from the target.

Attachment F-3 reports the results of a Monte Carlo simulation of a system with a 15 basis point cap and a 4 basis point floor. Premium adjustments do not take place unless the fund balance is ±21 basis points from the reserve ratio target. Premiums would then be adjusted by no more than ±11 basis points for a single change. Using data from 1980 to 1999, the fund balance remains positive in all 300 simulations. This suggests that there may be soft target approaches that reduce premium volatility compared to the current system without materially increasing the risk of fund insolvency.

Mutual Model

Under a user fee approach, the notions of rebates or banks holding claims on the insurance fund are ruled out. With an appropriate pricing mechanism, past premiums represent compensation for the government for bearing risk, not capital of the industry that is held in trust by the government. This section will discuss rebates, bank claims on the fund, and additional private-sector features that could be introduced into the deposit insurance system.

Rebates

The argument for rebates arises from the concern that the pricing mechanism could, over time, result in excessive charges to the industry. This may be a reasonable argument, given the uncertainty associated with deposit insurance losses. Rebate authority would allow the FDIC to price insurance at the margin for each bank, while providing a safety valve against an excessively large insurance fund.

The reserve targeting approach can be used as a mechanism for determining when rebates would be appropriate. One approach is to place a cap on the insurance fund and to rebate funds above that amount. Pending legislative proposals adopt this approach but direct the rebates toward payment of insured institutions' obligation to pay interest on Financing Corporation (FICO) bonds. An alternative is to specify a range where some portion of excess funds is rebated to the industry.

If policymakers reach the conclusion that rebates are appropriate, the question still remains as to how to allocate the rebates. While premiums are based on the current assessment base held by banks, this is not likely to be an appropriate basis for distributing rebates. New institutions and those that had grown rapidly would be given a distribution of income from past premiums that they did not pay. A more equitable approach would be to base rebates on past premium payments, which requires a decision about the appropriate look-back period.

An important feature of mutual arrangements is that rebates could be distinct from premiums. Rebates would be appropriate when the probability of insolvency of the insurance funds becomes sufficiently remote, and would be allocated based on past payments. Premiums would be based on the risk a bank posed to the insurance fund; this would never be zero. The cash flow between a bank and the insurance fund would thus have two components, one reflecting past contributions and the other reflecting risk exposure. The net result of these cash flows might be positive or negative. Even if the net result were zero for many banks, this would still represent an improvement over the current zero-premium system. This is because banks would be protected from the possibility that the insurance funds may grow without limit, while still being charged at the margin for the risk they pose to the insurance fund; this is critical if premiums are to provide appropriate incentives.

Banks Hold Explicit Claims on the Mutual Insurance Fund

Rebates are not the only possible feature of a mutual deposit insurance system. Once the notion of tracking past contributions for the purposes of providing rebates is introduced, the system moves to a mutual system that involves explicit claims; this leads to a significantly different framework for funding deposit insurance.

The current federal credit union share draft insurance system is an example of a system in which insured institutions have explicit claims on the fund. Credit unions are required to maintain a one percent deposit in the insurance fund. When the fund is above a level determined (within bounds) by the deposit insurer, rebates are provided. Another feature is that on an individual institution basis, deposit growth must be accompanied by a proportional contribution to the insurance fund. The credit union model has been criticized for its accounting treatment because the deposits that a credit union must place in the insurance fund are counted as an asset on the books of the credit union and as part of the insurance fund. These and other features of the credit union system are reviewed in a 1997 study by the Treasury Department. On the other hand, Hendershott and Kane (1996) have argued that the credit union insurance model provides positive incentives for monitoring by member institutions.

Following the mutual model, banks also could be required to pay into the fund an amount proportional to deposits. If a bank’s deposits grew, the bank would be required to "top up" its contribution to maintain it at the specified proportion. Likewise, a bank with deposit shrinkage would be entitled to a rebate or credit.

This feature could address the concern that banks are able to bring insured deposits into the system without having contributed to the insurance funds. The key decision here is whether the asset will be carried on the books of the insurance fund or of the bank.

If the asset is carried on the books of the insurance fund, the bank’s payment can be thought of as an "initiation" fee to join the deposit insurance system. Presumably, under this approach, the current fund would be viewed as the accumulation of past initiation fees and existing banks would be given credit for past payments.

The problem with this approach is that the initial cost of chartering a bank or gathering deposits would be significantly increased, once again raising concerns about stifling innovation. Allowing the fee to be paid over time, however, could mitigate these concerns.

Alternatively, if the asset is carried on the books of the bank, the concerns about stifling innovation and growth are diminished. Under this approach, the bank’s payment would be in exchange for a claim on the insurance fund. The accounting treatment and valuation of this claim would depend on the features of the claim.

As an example, the claims could be structured similar to sha