Market Data and Bank Supervision: The Transition to Practical Use
Incorporating market data into the examination process.
Published September 1, 2001 | September 2001 issue
Economic research conducted over the last several years has shown that market prices contain information on the riskiness of banking organizations. (As a point of clarification, we use the terms bank and banking organization to refer to both banks and bank holding companies.) This finding effectively addresses claims that market data offer no insight into bank risk taking.
But these findings taken on their own do not make a conclusive case that bank supervisors, who must assess the condition of banks, should begin relying on market data. As economists might say: The findings are necessary but not sufficient. Bank supervisors still must determine if and how they can use market signals, given both their need for real-time, easy-to-interpret information and the already established processes and techniques they use in practice.
In fact, obtaining high-quality market data that provide clear signals to supervisors can be challenging. Moreover, the advantages of augmenting the existing supervisory process with market data do not appear overwhelming, at least as of today. That said, we believe a comparison of benefits and costs justifies increased incorporation of market data into supervisors' day-to-day work. Specifically, supervisors could use such signals in three major ways. Market information could be used to help assess the overall condition of institutions, as well as the quality of loans and capital. Market data also could facilitate supervisory responses to institutional risk taking by reducing the uncertainty associated with those responses and ensuring that actions are credible and fair. Finally, supervisors could use market data to more efficiently and effectively allocate scarce supervisory resources. However, we believe that supervisors will need to look beyond the usual data suspects and examine many sources, including signals from equity markets.
We clearly believe that making more routine use of market data in the supervisory process will have short-term benefits. But increased knowledge of and experience with the data's strengths and weaknesses offer the biggest potential return. Some of the most significant concerns about the use of market information, such as uncertainty of interpretation, may simply reflect a lack of experience with the data. As a result, we make several recommendations to encourage increased supervisory use of market data. Other concerns require more applied research on the use of market data, and we make suggestions for a future research agenda.
Arriving at the transition point from basic research to practical use of market data
In the United States, federal and state authorities manage bank risk taking through prudential supervision and regulation. Under this regime, supervisors first assess the riskiness of banks. Part of this assessment generally involves analysis of data from mandated financial reports as well as sources specific to any given bank. This off-site data analysis supplements the on-site supervisory reviews of loan quality, risk management and other factors that take place on the physical premises of the bank. Based on their assessment, supervisors may take necessary steps to ensure that banks operate prudently.
This supervisory approach to banking stands out in an economy that generally relies on market forces to determine how much risk firms assume. Firms usually communicate their plans of action to investors, who assess the riskiness of the firms' proposals. These assessments influence the prices investors demand for their funds as well as the restrictions they place on the firm. Because most observers view this price-centric system as successful, they naturally have called on bank supervisors to make better use of market signals. After all, to some extent market investors and supervisors seek similar outcomes: accurate assessment of, and response to, bank risk taking.
However, before bank supervisors consider using market prices as an input into their assessments, they must have confidence that the prices set by bank creditors will vary in a reasonable way with the riskiness of banking organizations. Market prices, at a minimum, should signal greater risk for higher-risk institutions. Prices set by bank creditors cannot simply be assumed to pass this litmus test. Explicit government programs, such as deposit insurance, and implicit guarantees that may lead some creditors to perceive that they will be bailed out if a bank fails, can distort prices. Indeed, such distortions partly justify prudential supervision. The relatively high cost of evaluating opaque bank assets adds to the challenge for market investors.
Despite the potential hurdles, bank research over the last decade finds that creditors of banks produce market signals that vary in the proper direction with the riskiness of banks.1 Market prices indicate a higher level of risk for banks that are closer to failure, have riskier balance sheets and engage in riskier activities; in addition, higher risk banks tend to make less use of funds that are risk sensitive (that is, funds whose cost will increase along with bank riskiness). Market data thus contain information, not just "noise."
Such research findings are necessary but not sufficient to justify increased use of market data by supervisors. Supervisors already have established processes to assess supervised institutions with which they have extensive experience and comfort. Changing those processes, through increased use of market data, is costly. Part of the cost involves learning how market data can be used. In addition, any switch in supervisory process will entail new expenses such as developing and communicating new guidance. Market data would also impose a cost if it led to inferior supervisory assessments or actions.
Before increasing the use of market data in the supervisory process, supervisory agencies should feel comfortable that the benefits from increased use will outweigh these costs. Ideally, supervisors would carry out this analysis with exact measures of costs and benefits. Due to measurement challenges, supervisors will only have a rough gauge as to costs and benefits. As a result, we believe a case for increased use of market data could rely on a finding of tangible benefits of increased use at very low costs. Based on our analysis of the presentations and discussions at the Federal Reserve conferences highlighted in the introduction to this symposium, we argue that bank supervisors could gain tangible benefits at low cost by using market data in:
- the assessment of bank condition;
- the supervisory response to bank condition; and
- the allocation of supervisory resources.
We discuss uses of market data in each of these three areas, but first we briefly explain why practical use of market data will likely rely on multiple types of signals from the market.
How multiple sources of market data can facilitate practical use of market signals
A bank issues many types of financial claims in addition to insured deposits, including equity, long-term and short-term bonds and notes, and uninsured deposits. Embedded in the prices investors pay for these claimsor equivalently, in the interest rates they demandare assessments of the condition of the bank. Among these many types of claims, most recent policy attention has focused on a single source of market data, spreads on subordinated notes and debentures (SND), as providing the greatest promise for enhancing the assessment of bank risk taking. (See "Market Data Sources in More Detail" for a more detailed discussion of SND and the other sources of market data mentioned below.)
We believe a focus on SND is too narrow, and that reviewing multiple sources will maximize the benefit supervisors can receive from market data. In part, this conclusion reflects current limitations of SND spreads. Only a small number of banks issue SND, although this does include many of the large and systemically important ones. But, even many of the largest banks have a limited amount of SND outstanding at any point in time. The SND issued by one bank might differ in important wayssuch as the time left to maturity on the issuefrom the SND issues of other banks, making it difficult to compare a bank to its peers using SND data. Finally, many SND issues do not trade regularly in venues from which supervisors can dependably and quickly obtain market prices.
Other sources of market data could be valuable additions to the information set:
There are now widely used models for obtaining risk assessments from equity markets. Relative to SND, many banks issue stock and this stock tends to trade in liquid markets. Vendors sell software that generate easy-to-produce, timely equity-based risk assessments.
Senior debt could significantly augment data available from debt markets.
Uninsured deposits may offer the only hope of obtaining market prices for banks without traded SND or equity.
Quantity-based measures, such as changes in the proportion of uninsured deposit funding, can provide signals as well.
Our point is not that these other types of data are better than SND; they clearly are not perfect. For example, the equity-based models are theoretically and computationally complex, and the Federal Reserve does not have ready access to market prices for uninsured deposits. But on balance, these limitations are no more significant than those encountered with SND data. Thus, looking beyond a single source appears likely to enhance supervisors' ability to use market data.
Using market data to assess bank condition
There are three elements of supervisory processes where market data offer bank supervisors a low-cost tool for assessing the riskiness of institutions. First, market data, as an additional measure of overall condition, can augment supervisors' assessment of the condition and riskiness of banking organizations. Second, supervisors can incorporate market data into statistical models used to forecast the future condition of banks. Finally, supervisors could use market data in the analysis of a bank's loan quality and capital adequacy.
Assimilation of market data into summary assessments
Bank supervisors review standard performance ratios on asset quality, earnings and other aspects of bank performance, as well as measures specific to the business of a supervised institution, when assessing condition. Evaluations from on-site exams also contribute significantly to these assessments. Supervisors use all these pieces of information to come to conclusions about components of bank soundness, including funding, capital adequacy and risk management, as well as the overall condition and riskiness of the institution. The judgmental nature of the existing process could allow market data to be incorporated at a fairly low cost, with minimal disruption.
However, some bank supervisors have argued that lack of experience would limit the use of market data in the assessment process. Over the many years of working with traditional financial information, managerial reports and on-site inspections, supervisors have developed an intuition for the relationship between traditional data and the condition of banks. Supervisors have not had the time or experience to develop the same facility with market data. Supervisors argue, for example, that it is difficult to identify atypical market signals, thus limiting the benefits of using market data in the short run. Analysts can address the supervisory need, in part, by describing the distribution of market data signals. For example, distributional analysis can provide the average change in a market signal over recent years, thereby providing supervisors with benchmarks for their assessments (See "Interpreting Market Data Signals" for reports of initial results from distributional analysis.)
Off-site statistical models
While much of the assessment process relies on judgment, supervisors also make use of empirical techniques such as statistical formulas to forecast future supervisory ratings or to identify the probability that a bank will fail. The Federal Reserve's econometric model for forecasting bank failures relies on accounting variables related to the asset quality, capital adequacy, liquidity and earnings of a bank.2 A second model estimates supervisory ratings for a bank and relies on similar financial variables plus the most recent supervisory rating. The output from these models deserves serious consideration; for example, research shows that once an on-site supervisory rating is more than six months old, supervisors would be better off relying on the model than the rating to identify problem banks.3
These models provide an ideal method for determining if market data would provide any additional benefit, given the existence of supervisory ratings and financial data. Specifically, analysts can add market data to these models and determine if this addition improves the statistical accuracy of the model. Based on presentations at the conference, such tests have found that market data, specifically equity market data, could improve the forecasts of such models. 4
The addition of equity-based data improves measures of the predictive capacity of the supervisory rating model by as much as 5 percent to 10 percent (for cases where the bank is downgraded to "unsatisfactory" condition). While these figures might appear quite small, they do not represent the total informational power of equity data but rather the marginal value of adding such data to an existing model. The presentations did not investigate the marginal power of adding traditional supervisory data to a model that already included market data. As such, we do not know whether the traditional supervisory variables in the model would have similar marginal effects on forecast accuracy.
Statistical models would provide the greatest benefits for the supervision of institutions where the supervisory rating has the greatest chance of becoming stale. Many mid-size institutions meet this description because they receive only annual on-site examinations supplemented by off-site monitoring of traditional, often backward-looking data. Market data could prove useful even at larger institutions if they are not subject to continuous supervision.
Market assessments of condition could also prove useful for banks engaged in nontraditional activities. Standard data sources and analyses that were developed for institutions primarily engaged in traditional banking may become less appropriate and more difficult to interpret as the range of activities conducted by banking organizations expands. See more detail on the types of institutions for which market data could be valuable.
Asset quality and capital adequacy
The same market data used in assessing the overall condition of a bank can also help assess the quality of a bank's loans and the adequacy of its capital.
Asset quality. Firms that borrow from banks also may issue equity and debt instruments. Supervisors can use the signals from markets for these financial instruments to evaluate the ability of borrowers to repay the bank. In particular, since supervisors cannot review every loan made by a bank, market assessments can provide a low-cost method for identifying the loans that should receive the greatest level of supervisory scrutiny. Supervisors might, for example, rank-order industries by a market signal of risk. Borrowers in industries that appear to be in the worst condition might receive additional supervisory review. Indeed, over the last several years, a process along these general lines has facilitated the supervisory review of large, syndicated loans (loans of more than $20 million made by one bank but sold to at least three others).
Once supervisors have decided which loans to review, they can use the market assessment as part of their loan analysis. Using market data in this way would fit well with current loan review procedures that require supervisors to assimilate information on the borrower and make subjective judgments. Many examiners now carry software on their laptops that can generate market signals derived from equity-based models, and so can use this information during the loan review process at little or no marginal cost.
Capital adequacy. Market data can assist in assessing the capital adequacy for banks using economic capital models. The Federal Reserve has encouraged banks to make increasing use of economic capital models where appropriate. In the most general terms, these models estimate the amount of capital a bank must hold to withstand likely future losses some percent of the time and over some period (such as one year). These capital models require banks to determine the probability of default on assets and the losses if defaults occur. Lack of institution-specific, historical data has led banks to use market data in their modeling efforts. For example, some banks use equity-based models to determine the probability of default. Although bond data seem to be used less often than equity-related data in capital modeling, some banks also use the prices of defaulted bonds to estimate default-related losses. Supervisors must have an in-depth understanding and appreciation of market signals in order to evaluate and assess the use of these models.
Using market data when devising supervisory action
Once supervisors make an assessment of the condition of a bank, they must determine how they will respond. For institutions that pose few safety and soundness concerns, the supervisory action plans will not require much change for the bank. Alternatively, supervisors may demand that institutions adjust their behavior, with the severity of the supervisory actions corresponding to the assessment of risk. Market assessments can support the supervisory response by reducing the uncertainty associated with supervisory actions, facilitating communication between supervisors and institution management, and providing an independent assessment to reinforce the legitimacy and fairness of supervisory requests. These uses of market data appear consistent with current processes, and thus should pose relatively little disruption or cost.
Reducing uncertainty. As noted by Minneapolis Fed President Gary Stern, market assessments of bank riskiness could speed supervisory actions by reducing uncertainty about the appropriateness of the supervisory response. Consider the case where supervisors and markets appear to draw similar conclusions. For example, supervisors may come to an initial decision to downgrade the supervisory rating of the bank because asset quality, earnings and capital adequacy have fallen; during that same period, debt spreads may have widened relative to peer institutions. Market signals thus reaffirm supervisory judgments, and supervisors could downgrade the bank more promptly and with more confidence.
It might seem that conflicting signals between markets and supervisors would increase uncertainty for supervisors. But at least in principle, the more information a supervisor uses in formulating assessments of bank condition and supervisory response, the tighter the range of possible values their assessment should take. When supervisors develop an initial assessment of a bank's condition, they are in essence making a best estimate, around which they always have some uncertainty. Market signals, whatever they indicate, represent additional raw information. And additional information, properly used, can almost always improve initial estimates.
Flannery (2001) makes the point clearly. "Statistical theory indicates that even an imprecise market assessment (forecast, signal) of bank condition should complement standard supervisory procedures, in the sense that a more accurate forecast can be made using both sources of information than with either one alone." Using both market and supervisory sources of data should prove especially beneficial because the respective assessments rely on different perspectives and calculations. Flannery goes on to note that, "If market information could reduce the uncertainty about a firm's true condition ... the market assessment [could] convince supervisors to act sooner."5
Enhancing communication. Supervisory actions should achieve the greatest effect when boards of directors and managers of the regulated entities view them as well-founded. We might expect institutions to engage in less resistance to supervisory actions when supervisors can support their actions with data the firm takes seriously and can communicate their intent in terms the bank uses. Market data fill the bill in both regards. Banks would have a difficult time rejecting the validity of market data, especially if they incorporate such data in their own business planning and processes, from funding and compensation to business strategy and risk management.
Facilitate "fair" actions. In a democracy, society as a whole must view supervisory actions as fair and legitimate; capricious regulators eventually bear the consequences. Banks can successfully challenge supervisory actions that are inconsistent with regulatory powers or that arise from arbitrary decision-making. Critics sometimes have charged that supervisory actions do not treat all institutions fairly, with some institutions receiving too light a response while others receive a harsh penalty.
Linking supervisory action to market signals offers a method for reducing this impression. Linkage to a neutral measure could help convince legal authorities and others that banking supervisors have acted in a reasoned and fair manner. A formal link might require supervisors to take a specified set of actions when a market signal assumes a specified value (for example, three times higher than peer average). A system of this sort already links bank capital to supervisory action.6
But an additional advantage of linking market data to supervisory action is the third-party nature of market assessments. Supervisors can directly control bank capital levels by forcing a bank to write down assets, but they cannot directly influence market assessments of bank riskiness. Even without formal linkage between market data and supervisory actions, supervisors could use such data to support supervisory activities in legal venues. We do not view market data in this context as a tool to prevent regulatory forbearance, but rather as a backstop to confirm the legitimacy of supervisory action. The market signal to trigger supervisory action would have to be chosen to reflect this backstop role, and trigger levels would be set to catch only banks in very poor condition.
Using market data in allocating supervisory resources
So far we have largely limited our discussion to uses of market data to assess risk and support supervisory action in institution-specific cases. The data could also help regulatory agencies like the Federal Reserve allocate their supervisory resources efficiently. Bank supervisors, aware of the cost of devoting resources to supervision, have sought to economize on those resources; for example, the Federal Reserve tailors its bank examinations to those areas likely to pose the greatest risk (risk-focused exams). Supervisors can use market data to directly advance their risk-focused approach. As noted above, market data can help in asset reviews by freeing up supervisors to spend the most time with the riskiest assets.
In the same vein, supervisors can use market data to rank-order the riskiness of institutions. Those institutions posing the highest level of risk would receive the greatest attention from supervisors. Market signals are better suited to this purpose than existing methods of supervisory assessment. Examination reports are narrative and cannot be used to construct rankings. Supervisory ratings are numerical but also are inadequate for ranking. For example, supervisory ratings of bank holding companies take on only five values, and only three are used with regularity, making it difficult to adequately differentiate risk levels. In addition, the rating numbers themselves have no inherent meaning (they could just as easily be A through E instead of 1 through 5), making it difficult to quantify the relative condition of institutions with different ratings.
Market measures can take on an infinite range of values and thus can help differentiate between institutions. Moreover, some market measures produce values that have meaningful interpretations. For example, many models based on market data can provide an estimate of the probability of failure; a 20 percent chance of failure has a clear meaning relative to a 10 percent chance of failure. Such rankings by market measures could thus supplement current methods for allocating supervisory resources.
A lack of experience has slowed the transition to practical use of market data
We believe there are benefits to adding analysis of market data to the supervisory repertoire. However, this change to the status quo could take some time unless concerted steps are taken. Supervisors have monitored some market data, such as stock prices, for many years. But greater use of market data along the lines discussed in this essay is akin to rolling out a new product with which supervisors do not have significant experience. Product adoption may be slowed by several factors related to this lack of experience.
Supervisors, as noted above, are not familiar with the properties of market data. Although research that provides a better sense of how to interpret market signals (for example, describing more completely how to identify market signals that are sufficiently unusual to generate concern) will help, supervisors may simply need additional hands-on experience to achieve the intuition and comfort level they have with more traditional supervisory data.
Supervisors in the field tend to rely on extensive guidance and templates that assist them in assessing risk and devising a response to that assessment. By and large, the guidelines and templates that currently exist do not address or facilitate the use of market data in the supervisory processes. More generally, there has not been specific direction from senior supervisory staff encouraging explicit use of market data in carrying out supervisors' jobs.
Supervisors normally review financial or managerial data linked to specific components of an institution's condition. For example, they review data on the performance of loans to determine the riskiness of the loan portfolio. Outside of a small subset of fairly well-defined events such as mergers, supervisors cannot currently link changes in market data to specific activities or events. The risk premium on a bank's bonds might increase, but to what aspect of a bank's business is the market reacting? Absent that kind of interpretation, it is difficult for supervisors to translate information into actions.
Supervisors' experience is based on analysis of information not generally available to the public. An emphasis on such private information and in-house analysis leads many supervisors to be at least somewhat skeptical of the benefits of market assessments. After all, from the supervisory perspective, market participants must rely on inferior data to come to their conclusions. This information disadvantage may be offset in other wayssuch as the large number of market participants and the resources they are able to devote to analysisthat make market information valuable, but these offsetting factors tend to be subtle and hard to demonstrate.
Recommendations for transitioning to practical use of market data
Our recommendations in response to the concerns noted above fall into two categories: those that involve supervisory policies and procedures, and those that suggest an applied research agenda to respond to supervisors' questions about the data.
Changes to supervisory processes. Supervisory agencies could directly address some of the key concerns that have slowed the transition to practical use of market data. First, policymakers and senior bank supervision staff should describe their expectations for use of market data in the supervisory process in greater detail. Part of the confusion as to expectations comes from the use of similar terms with dissimilar meanings. In particular, some policymakers have called for "increased use of market assessments" in supervision or "increased market discipline," but appear primarily to be encouraging increased disclosure by banks rather than supervisory uses of data of the type we have been discussing. Clarification could help prevent a disconnect between practice and policy statements.
Second, and related to the first point, the supervisory agencies should move beyond statements of policy to practical guidance for staff. Promulgation of more formal supervisory guidance could provide field-level supervisors with needed direction; the lack of official guidance makes it difficult for supervisors who do not have significant experience with market data to use it effectively. In addition, guidance would emphasize leadership expectations.
Third, supervisors could offer training to staff on use of market data. Training offers another vehicle for gaining experience with market data and could supplement what supervisors learn from making use of market data. Supervisors at Federal Reserve conferences identified training as a high priority.
Additional applied research. Supervisors have raised a number of questions that would require additional research to answer. One of the most repeated queries concerns the relative merits of the various data sources. Does one source provide a signal deserving greater weight in analysis? Carrying out such analysis for the limited number of signals would require additional data collection. In addition, supervisors want confirmation that the more sophisticated market-based signals, such as those derived from equity prices using relatively complex models, outperform simpler market signals such as raw stock prices or price-to-book ratios, which supervisors feel are more intuitive.
A second request from supervisors reflects the need to filter out those aspects of market signals that do not relate to the riskiness of institutions. Stock prices rise and fall for many reasons, not all of which reflect changes in a bank's condition. Debt spreads reflect more than just default risk; for example, changes in liquidity affect the spreads on bank-issued SND. Research that provides techniques to more keenly focus on the relevant supervisory content of market signals would yield big benefits for practical supervisory use.
A third, and similar, strand of remarks from supervisors focuses on the link between market signals and the specific operations of a bank. Ultimately, supervisors with concerns about an institution must ask the bank to do something. Supervisors would prefer to link market signals to specific bank activities so that they can tailor their analysis and response. Additional research could help make these links.
What's the right speed for the transition?
The case for making the transition to greater use of market data seems clear, but how quickly should bank supervisors make that transition? The research evidence that market data have informational content has grown over the last several years, but our understanding of market data remains incomplete. Initial practical experiments by Federal Reserve analysts have given them a much greater appreciation for the challenges involved in making greater use of market data. We have already noted some of these in our discussion above. There have also been bugs in presentations of market data, in calculations of measures of risk based on market data and in interpretations of changes in market data. Anyone involved in the rollout of a new process will not find these imperfections surprising. But pushing a new product out the door before it achieves a certain state of readiness can cause irreparable damage to its reputation. Indeed, some claims made on behalf of market data, such as an ability to routinely identify weak institutions that supervisors incorrectly deem healthy, seem overstated and unsupported. Does prudence dictate that the Federal Reserve should increase its understanding of market data before asking supervisors to make more use of it?
We recommend moving forward quickly in the near term, even with our less-than-perfect understanding because increased use of market data in the short run appears to pass basic cost-benefit tests. The costs of using market data in the ways and for the objectives we have suggested seem very, very low. It is relatively cheap to acquire and process the information with many types of market data available in electronic form from well-known sources. A small, coordinated staff could, and to some degree has, managed data acquisition, analysis and distribution for the entire Federal Reserve System. And although the benefits may be difficult to quantify fully, they appear real and substantial. Our conclusions are consistent with the limited empirical research on this topic, which finds that supervisors have valuable and unique insight into bank condition, but that market data provide additional valuable information.7
Perhaps the more compelling reason to move forward quickly rather than cautiously is to improve bank supervisors' understanding of market data. Only through the limited monitoring of market data done to date have flaws been exposed, additional analyses conducted and remedies devised. In short, the Federal Reserve has gained most of its knowledge about the data from using it. We doubt that a hiatus from practical use would improve our understanding or address the major stumbling block, the limited experience bank supervisors have with market data. Indeed, assessing and responding to bank risk taking involves more art than science, implying a certain futility in waiting for a precise answer as to how supervisors incorporate market and supervisory data into their decisions. Using market data in the low-cost and minimally disruptive ways suggested here is the most straightforward way to address concerns related to the lack of experience.
1 For summaries of this research see Federal Reserve System Study Group on Subordinated Notes and Debentures, Using Subordinated Debt as an Instrument of Market Discipline, Board of Governors Staff Study 172, December 1999; and Mark Flannery, "Using Market Information in Prudential Bank Supervision: A Review of the U.S. Empirical Evidence," Journal of Money, Credit and Banking (August 1998, Part I), pp. 273-305.
2 A discussion of the Federal Reserve's model is found in Rebel Cole, Barbara Cornyn and Jeffrey Gunther, "FIMS: A New Monitoring System for Banking Institutions," Federal Reserve Bulletin January 1995, pp. 2-15.
3 Rebel Cole and Jeffrey Gunther, "Predicting Bank Failures: A Comparison of On- and Off-Site Monitoring Systems," Journal of Financial Services Research,13, 1998, pp. 103-117.
4 Research shows that bond ratings and equity measures better predict bank performance variables than all but the most recently issued supervisory ratings. See Allen Berger, Sally Davies and Mark Flannery, "Comparing Market and Regulatory Assessments of Bank Performance: Who Knows What When?" Journal of Money, Credit and Banking, Vol. 32., No. 3, August 2000, Part 2, pp. 641-670. Subordinated debt spreads predict supervisory rating changes better than some, but not all, measures of bank capital. See Douglas Evanoff and Larry Wall, "Sub-Debt Spreads as Bank Risk Measures," Federal Reserve Bank of Atlanta, Working Paper 2001-11, May 2001.
5 Flannery, M. J. (2001) "The Faces of 'Market Discipline,'" Journal of Financial Services Research, forthcoming.
6 The regulatory regime called "prompt corrective action" groups banks into five categories ranging from "well capitalized" to "critically undercapitalized" based on the amount of capital they hold. Banks face tougher regulatory requirements as they fall into groups defined by lower levels of capital.
7 For examples, see Robert DeYoung, Mark Flannery, William Lang and Sorin Sorescu, "The Information Content of Bank Exam Ratings and Subordinated Debt Prices," Journal of Money, Credit and Banking, forthcoming; and Allen Berger, Sally Davies and Mark Flannery, "Comparing Market and Regulatory Assessments of Bank Performance: Who Knows What When?" Journal of Money, Credit and Banking, Vol. 32, No. 3, August 2000, Part 2, pp. 641-670.