General Information & Client Services
  • Americas: +1.212.553.1653
  • Asia: +852.3551.3077
  • China: +86.10.6319.6580
  • EMEA: +44.20.7772.5454
  • Japan: +81.3.5408.4100
Media Relations
  • New York: +1.212.553.0376
  • London: +44.20.7772.5456
  • Hong Kong: +852.3758.1350
  • Tokyo: +813.5408.4110
  • Sydney: +61.2.9270.8141
  • Mexico City: +001.888.779.5833
  • Buenos Aires: +0800.666.3506
  • São Paulo: +0800.891.2518

This article provides a summary of the mid-cycle stress test results, including observations about scenarios, loss estimates and PPNR, disclosures, and areas for improvement.

In September 2013, the 18 Comprehensive Capital Assessment and Review (CCAR) banks released, for the first time, their mid-cycle stress test (MCST) results.1 The MCST differs in significant ways from the annual Federal Reserve (the Fed) CCAR and Capital Planning Review. In particular, firms are responsible for creating their own macroeconomic scenarios and may, in some cases, use models and approaches that are different from the CCAR modeling exercise. The capital actions permitted by the Federal Reserve are certainly different than those for the CCAR exercise, and firms have more discretion with planning and taking management actions throughout the forecast horizon. Importantly, the Fed will not provide an objection, conditional approval, or non-objection to the mid-cycle results, at least not publicly.

The objective for the MCST is to allow firms to better tailor scenarios to their own idiosyncratic risks. The CCAR exercise is a “one size fits all” test that is specified by the supervisory agencies, which does not permit differentiation across business models – differences that may be critical to loss, revenue, provision, and capital calculations.

Scenarios

The recent MCST results have yielded interesting observations. In aggregate, the overall severity of the scenarios used by the banks largely matched the stresses from the Federal Reserve System’s (FRS) CCAR exercise. The banks generally took a conservative approach to the stress scenarios, although there were a few whose scenario assumptions seemed far less severe than anticipated, as indicated by peer comparisons and other stressed economic variables provided by economic research firms. This fact may signal that some banks should better assess their scenario assumptions and compare them to other available scenario sets.

The approximate average of the unemployment rate used in the scenario was 12.12%, whereas one firm’s peak unemployment level was only 10.9%.2 The peak unemployment in the FRS 2013 CCAR exercise was around 12.1%. The impact to GDP was uneven across the MCST disclosures, with a range of peak GDP declines in the forecast from -1.1% to -8%. The peak GDP decline in the FRS scenarios was -6.1%, and the average peak rate across all MCST results was approximately -4.76%, not nearly as severe on average as the Fed scenario. The peak-to-trough House Price Index (HPI) decline from the last FRS scenario was approximately -20%, and the MCST average decline was -22%, with one firm estimating a -43% decline in home prices. Interestingly, the Fed uses the Dow Jones but virtually all of the MCST reporting banks chose to use the broader S&P 500. In the last CCAR round, the Fed scenario posted an approximate -24% decline in the Dow Jones, whereas the average for the banks reporting a severely adverse S&P 500 shock was on average -44%.

The most interesting observation was the use of tailored idiosyncratic scenarios by the banks, an expectation clearly articulated by the FRS. Ally used a massive oil price shock (a peak oil price of $229), as well as a used car after market index shock (the Manheim Index), to tailor the stress test and models to their particular business models. In other cases, like US Bancorp, Fifth Third Bank and KeyCorp, the scenarios were connected with their geographic footprint to account for the regional portfolio concentrations of their credit portfolios. Goldman Sachs included a reputational risk event in their stress test, and one bank – Bank of New York – considered how stresses are transmitted across various risks (i.e., how credit may impact liquidity, market, and operational risks). This is a sound practice given that the interaction across financial risks is a significant contributor to assessing a firm’s overall resiliency.

Loss estimates and pre-provision net revenue

While comparing loss rates from the 2013 CCAR exercise to the MCST disclosures is at best an “apples and oranges” comparison, the overall loss rates disclosed in the MCST are an improvement over the 2013 CCAR results. Figure 1 provides a summary of all the reporting banks.3 All asset classes showed a reduction in overall loss estimates, other than Junior Liens, Credit Cards, and Other Consumer.

Figure 1. Mid-Cycle Loss Rates vs. FYE 2012
Mid-Cycle Loss Rates vs. FYE 2012
Source: Moody's Analytics

Aggregate industry losses were posted at $213 billion, with provisions of $269 billion to cover the losses. The 2012 and 2013 aggregate losses were estimated at $534 and $462 billion, respectively, with provisions of $324 and $317 billion. Across the planning horizon, no bank came close to breaching the FRB’s 5% Tier 1 common threshold, with a minimum of 6% (Ally) and a maximum of 12.3% (State Street). In fact, five of the reporting banks showed a beginning-to-end increase or no change in Tier 1 common capital, with the average decline in total risk-based capital being only -2.71% and Tier 1 common of -4.54%, with an aggregate average total and Tier 1 common of 13.53% and 9.32%, respectively. Eight banks actually increased the total risk-based ratio, reflecting lower balances and a shift to less risky assets. The leverage ratio decline was -3.18%. Credit cards were the largest contributor to consolidated losses, with 35.9% of the total loss contribution across the industry (see Table 1).

Pre-provision net revenue was $315 billion, offset by the aforementioned provision level of $269 billion, trading losses (subject to instantaneous shocks of the global market) of $74 billion, securities losses of $7.7 billion, and other losses of $17.7 billion. These losses were commonly related to goodwill, deferred tax assets (DTA), intangibles, for valuation only (FVO) changes, or related impairments and/or legal reserve builds and expenses.

Table 1. Consolidated losses by asset type
Consolidated losses by asset type
Source: Moody's Analytics

Disclosure components

The various management planning actions were some of the more compelling components of the MCST. For example, some banks thoughtfully adjusted new business volumes in the scenarios to account for lower credit demand and availability. These banks used internal or vendor-supplied loan demand models by asset class. While few banks mentioned the use of sophisticated supply models for credit, some banks did vaguely describe the use of statistical models, combined with historical industry and internal data, to attempt to measure the appropriate rates and spreads associated with the available asset class credit demand.

This is an interesting approach in that as a bank adjusts its rates and spreads, the level of “received” new volumes declines, reflecting a conscious risk appetite decision by the bank’s planners. Higher relative rates and new business credit spreads in the pro-forma plan would naturally imply a lower supply of credit and a tightening of implied credit standards under the scenario, while lower relative rates and spreads might imply a greater supply of available credit. In many cases, this is done qualitatively through interactions between the finance, risk, and lines of business, although quantitative conditional rate and spread models are sometimes used. Given the new Federal Reserve guidance on the incorporation of Basel III capital rules into the CCAR forecast, banks may need to revisit these simple approaches that merely follow Basel I rules, holding risk-weighted asset (RWA) levels constant across the planning horizon.4 This incorporation of Basel III rules should result in more granular and accurate modeling of credit conditioned new business volumes, and increase the communication and interaction between internal functional groups and the risk origination business lines of the bank.

It is clear through the disclosures that many of the banks are also having difficulty measuring, in a quantitatively sound manner, non-interest revenue and non-interest expenses. This is a challenging exercise, as a bottom-up approach relies on measures of business activity, headcount, loan balances, pipeline, service metrics, asset balances and flows, account balances, and other measures that may not be easy to obtain; or if obtained, the underlying quality of the data may be suspect. As a result, many banks are estimating the numbers at the line of business (LOB) level with workbooks and policy guidance provided by the firm’s central stress testing function. The LOBs are working with the firm’s quantitative modeling and finance groups to build appropriate statistical and planning models. Some banks are supplementing this approach with third-party analytical models, using available public Call Report, FR Y-9C, and firm-specific data. Such modeling, consistent with the Federal Reserve’s approaches, allows for a champion/challenger approach to this somewhat opaque area. Using alternative approaches is fully consistent with regulatory guidance5 and may assist banks in the sensitivity analysis or results, a key area of focus of the supervisory agencies, as evidenced by recent studies.6

Lastly, it was interesting to see how critical technical areas were addressed in the stress tests. For example, the modeling of disallowed deferred tax assets was critical in several cases to capital results, as was the need to consider the buildup of legal reserves to cover expected future claims, such as representations and warranty claims with government-sponsored enterprises (GSEs). Capital One’s criticism of the lack of transparency around the FRS’ models and methods, and that “…the Federal Reserve appears to have made a philosophical choice to use industry-wide models without making adjustments for objectively observable business practices and results among banks”7 seems like a well-phrased consensus statement underpinning many of the disclosures and conversations throughout a majority of the MCST filings. While the supervisory agencies are sensitive to the expectation that the banks should not strive to mimic the Fed models, additional transparency and industry dialogue regarding sound and better practices seems reasonable, and should be routine, and be a more transparent element of this ongoing exercise. Many firms believe more transparency and openness would increase the effectiveness of industry sound practice evolution, as well as accelerate the creation of improved standards of practice.

Disclosure scorecard

As a part of our review of the disclosures, we developed a scorecard, based on a qualitative assessment of the MCST disclosures, scoring each firm across several dimensions of reporting detail. While we do not present firm-level scores, we believe Table 2 provides an indicative measure of overall quality.

This qualitative score emphasizes that the overall adequacy of the disclosures can be improved, with immediate attention focused on how the entire MCST, and the CCAR itself for that matter, is governed. It is clear many banks are only accommodating the minimum disclosure requirements. While there seems to be no market interest or reaction to the disclosures, an enriched narrative that expands the depth of the analysis and a more effective discussion around modeling methods, as well as how the systems, models, and methods are used within a firm (not necessarily the stress metrics), will go a long way toward enhancing the process.

Table 2. Disclosure scorecard
Disclosure scorecard
Source: Moody's Analytics

It is also helpful to note that the range of expected disclosures may have been over-specified. It may make more sense to mandate a certain minimum set of disclosures, but provide more guidance and principles around enhanced prudential expectations, and make internal adjustments to the process to encourage banks to meet more than the minimum standards. Expectations might address the depth of analysis, the incorporation of a broader range of metrics (such as operational risk, liquidity, investment portfolio losses, and mark-to-market losses), and key operational challenges. They also might focus on significantly improving management discussions and analysis about governance and how the overall program is expected to evolve to become more practical at enhancing internal risk management, planning, internal capital adequacy assessment process (ICAAP), and risk-based pricing initiatives.

Observed areas for potential improvement

It is important to note that the overall utility of the stress test reports may need to be revisited. One of the gross assumptions of the entire CCAR and MCST exercises is the exclusion of liquidity impacts and measures, as well as the failure to consider systemic transmission and contagion effects across capital, funding, payment systems, and markets. Internally, only one bank considered transmission effects across various risk pools. This seems like a reasonable risk management consideration given that severely adverse loss events will certainly have a sizable impact on liquidity, particularly wholesale funding, depletion of unencumbered liquid assets (due in part to collateral calls and derivative revaluations), and the liquidity effects of other off-balance sheet non-contractual commitments.

While all banks appear to have easily “passed” the 5% minimum Tier 1 common capital target set by the FRS, this largely ignores the potential for massive disruptions across banks, which causes each firm to “lock down” their positions, hoard liquidity, and freeze credit creation, resulting in the evaporation of liquidity buffers. Clearly, the Fed is pursuing enhanced liquidity policy through other exercises, but it may help to consider how these two highly correlated risks could be presented in a more unified analysis.8

While idiosyncratic scenarios are important, it remains unclear what the overall use case is for the stressed measures, other than as an “extreme event” calculation to determine firm-specific capital resiliency. Many firms appear to be struggling to determine the use case for the stressed measures. This is perhaps due to the fact that internal systems and processes across finance, risk, treasury, and trading remain focused on creating the stressed measures – rather than linking and automating the various business processes – while treating stressed measures as a “special case” of enhanced risk, finance, and balance sheet planning. Importantly, with the emerging US liquidity risk reporting requirements, the ability to measure interaction effects between liquidity, credit, and capital will likely become more important.

Aggregate industry losses were posted at $213 billion, with provisions of $269 billion to cover the losses… the largest contributor to consolidated losses was credit cards, with 35.9% of the total loss contribution across the industry.

The highest utility of the exercise, to date, may be in increasing the communication, as well as the functional and technical integration, across various business lines, and ensuring that the same models used for stress testing are also used in day-to-day risk management. The collection and use of the underlying data for static-pool analysis and risk assessment, likewise, potentially generates medium-term positive benefits, as long as the firm has the right technical data foundation, data models, and associated platform tools to create helpful line-of-business and related risk reports. Certainly, such data will be critically important for supervisors, as it will enhance continuous off-site supervision.

The underlying data could also potentially be used to enhance recovery and resolution planning. The increased data standards may also be helpful for various M&A activities. The data collected could essentially become the functional equivalent of a firm-level “data room” available at all times, which seems like a valuable safety and soundness mission and could certainly support the FDIC’s Title II Orderly Liquidation Authority (OLA) mandate. Such data might also assist in better risk-based deposit insurance pricing, and aid the Fed in discount window lending under FRA 13(3), or other crisis-based emergency lending programs.

Finally, it is a mystery that the loan and counterparty-level data standards are not being applied to the $10-50 billion banks. While the supervisory agencies are wise to consider reducing burdens on Main Street banks, more M&A activity will occur in this space than in the large bank space. Having access to common data feeds would clearly help accelerate the due diligence, bidding, and (possibly) multiples received. If the stress testing program persists – it seems like we are only in the first inning of a long ball game – better data, risk management, risk integration, and financial planning will benefit all well-run banking organizations.

Other areas that may be considered as areas for improvement are:

  • Enhanced disclosures around sensitivity tests, using champion and challenger model results to facilitate the analysis
  • Better disclosures around model performance by disclosing out-of-sample test results and similar measures, which could be added as a technical addendum
  • Significant lack of discussion and disclosure around securities portfolios, including other than temporary impairment (OTTI) numbers and clear mark-to-market (MTM) in the base-case and the forecast
  • Operational risk losses are buried in noninterest expense lines throughout the disclosures – this metric should be reported in more detail, with stress measures by operational risk category type
  • Improved methodology disclosures and discussion; current disclosures are too high level

It seems clear that many of the CCAR and MCST banks continue to view the stress testing exercise as a chore and compliance burden, not as an exercise to improve and refresh their internal systems, business processes, and integrated risk calculation capabilities. Perhaps this is to be expected if the overall regime has become over-specified and, as noted previously, if the various requirements handcuff innovation due to rules rather than principles. While standards need to be clear at the data level, similar to the Financial Products Markup Language (FpML), it may be useful to consider alternatives at the implementation and functional application layer.

While we appreciate that liquidity risk is being assessed in a separate regulatory silo, we are collectively attempting to break down silos, not build new ones.

Banks should change their focus to view the stress testing exercise as many had originally hoped – to create a more integrated view of forecasted risks across a more fulsome range of risk types and scenarios, with the tools and technologies that can link front, middle, back-office, risk assessment, and finance. This strategy creates a more agile, transparent, risk-aware, and efficient organization. The Federal Reserve deserves credit for undertaking such a difficult, multi-year exercise. However, developing useful tools, systems, data, and risk assessment methodologies that allow for dynamic and integrated balance sheet, income statement, cash flow, regulatory, and economic capital forecasts – which can be used every day, not simply twice a year under strict, rule-based conditions and limited scenarios – might enhance the overall utility of the large investments being made. The lack of more dynamic risk measures, methodologies, and integrated infrastructure appear to be a persistent challenge, perhaps as a direct result of possible over-specification.

It seems intuitive that the interaction effects between a “credit loss dominated” stress test and the transmission of such a shock, idiosyncratic or systemic, to the funding markets, particularly wholesale, should be directly incorporated into the stress testing exercise. In the MCST, it is unclear how liquidity runs and stresses are incorporated, and there was little-to-no discussion around models for assessing funding flight risk under stress. While we appreciate that liquidity risk is being assessed in a separate regulatory silo, we are collectively attempting to break down silos, not build new ones.

The ability to create unified measures is available, but it is clearly not cheap, nor quick. The need is real, though, and the regulatory agencies recognize this fact. Banks will need to increase their investments. As an immediate area of focus, and due to the required incorporation of the Basel III framework into the projection beginning in 2014, banks will need to consider, and begin building soon, much more profound and accurate measures for planning conditional credit-adjusted new business volumes and measuring the associated risk-weighted assets (RWAs) in a dynamic and granular fashion. Blunt statistical tools, portfolio average risk ratings, and PDs/LGDs/EADs, will not (and should not) suffice. While it may take the supervisors some time to catch up to this fact, banks should begin planning now.

While we are still learning a lot through these exercises, it is clear that continued improvement in data, architecture, risk analytics, reporting, governance, and the establishment of a comprehensive set of “use cases” for the overall framework, will need to evolve. The investment in this exercise is significant. It would be a mistake if the developed data, modeling capabilities, and internal governance structures did not result in a much more effective method for enterprise-wide balance sheet, credit, financial, and liquidity risk management, rather than merely establishing a highly assumptive and error-prone assessment of capital adequacy and planning.

Ultimately, the supervisors and the banks desire to be strong credit intermediaries, while earning a reasonable return on equity. In order to achieve this objective, the investment in CCAR infrastructure should be biased towards useful business goals and objectives. This means creating a platform that allows for dynamic interaction across firm-wide risk pools, an exercise that – rather than being an annual compliance burden – should evolve into a monthly business, financial planning, balance sheet, and performance management tool.

With the issuance of FR SR 12-17, which deals with the new consolidated supervisory framework for more complex institutions, the need to focus on capital and liquidity planning, governance, and recovery planning will continue to drive the stress testing program forward.9 As the overall project evolves, and as the industry, supervisors, and other third parties continue to learn, it is in the best interest of policy that the process moves in a constructive direction. This creates not only assessments of capital adequacy, but also a stronger, safer, and more sound financial system – one that is resilient, growing, profitable, and better informed about current and potential future risks.

Sources

1 Section 165(i)(2) of the Dodd-Frank Wall Street Reform and Consumer Protection Act and related regulations require large bank holding companies with total consolidated assets of $50 billion or more to conduct two stress tests each year. In the mid-cycle Dodd-Frank ActStress Test (DFAST), each firm is required to conduct stress tests under a set of internally developed scenarios (baseline, adverse, severely adverse). In the annual DFAST submitted in January of each year, each firm is required to conduct stress tests under a set ofscenarios (baseline, adverse, and severely adverse) developed by the Board of Governors of the Federal Reserve System (Federal Reserve).
Banks are required to submit the results of the mid-cycle DFAST to the Federal Reserve by July 5 of each year, including projections of preprovision net revenues (PPNR), trading and counterparty losses, provision for loan and lease losses, and capital levels over a nine-quarter planning horizon (Q2 2013 to Q2 2015 for the 2013 mid-cycle DFAST) under these scenarios. Mid-cycle DFAST rules also require each firmto publish an overview and summary of results based on the severely adverse scenario.

2 Note that the scenario averages and other details are approximate given the inconsistent reporting of scenario data across the eighteen MCST banks.

3 Note that banks reporting zero losses for various asset classes were removed from the averages to avoid skewing the numbers. Also, given State Street’s reduction in CRE assets from the CCAR exercise to the MCST, we removed the CCAR CRE loss numbers to better represent loss rates amongst the lenders engaging in the CRE lending business.

4 Board of Governors of the Federal Reserve System, Two interim final rules, September 24, 2013.

5 Board of Governors of the Federal Reserve System, Supervisory Guidance on Stress Testing for Banking Organizations with More Than $10 Billion in Total Consolidated Assets, Principle 2, May 20 www.federalreserve.gov/bankinforeg/srletters/sr1207.htm

6 Board of Governors of the Federal Reserve System, Capital Planning at Large Bank Holding Companies: Supervisory Expectations and Range of Current Practice, August 2013. www.federalreserve.gov/bankinforeg/bcreg20130819a1.pdf

7 Capital One Financial Corporation Dodd-Frank Act Company-Run Stress Test Disclosures, September 16, 2013.

8 See, for example: Federal Register, Proposed new liquidity data collection templates, September 19, 2013.

9 Board of Governors of the Federal Reserve System, Consolidated Supervision Framework for Large Financial Institutions, December 2012. www.federalreserve.gov/bankinforeg/srletters/sr1217.htm

As Published In:
Related Insights

Credit Loss Estimation - Industry Challenges and Solutions for Stress Testing

Thomas Day, Senior Director, Mehna Raissi, Director, and Chris Shayne, Director discuss credit risk management and loss modeling in a stress testing environment.

May 2014 WebPage Thomas Day, Mehna Raissi, Chris Shayne

Data, Analytics, and Reporting Requirements: Challenges and Solutions

In this paper, we first provide a background on stress-testing, discuss infrastructure challenges and issues related to legacy data and remediation requirements, including the costs and benefits of improved data management and the challenges of managing multiple hierarchies and reporting dimensions required by the Supervisory Authorities. Next, we cover data governance issues, the data requirements of meeting U.S. stress-testing mandates, and the basic elements of a sound data management infrastructure.

March 2014 Pdf Thomas Day, John Haley

Enterprise-wide Stress Testing

In this webinar, originally recorded on September 17, 2013, Thomas Day discusses best practices for expected loss and pre-provision net revenue forecasting, integration of stress testing into your business architecture, and transforming stress testing from a regulatory exercise to a strategic management tool.

December 2013 WebPage Thomas Day

CCAR and DFAST: What will 2014 Bring?

Learn how the Comprehensive Capital Analysis and Review and Dodd-Frank Act Stress Tests will impact banks in 2014 and how banks can best prepare for the changes.

November 2013 WebPage Thomas Day

Modeling Credit Losses to Meet Stress Testing Requirements

This article discusses two conceptual approaches for modeling stressed credit losses: top-down and bottom-up. It highlights the benefits and challenges of using each approach and regulatory expectations.

November 2013 WebPage Thomas Day, Anna Krayn

Stress Testing Webinar Series: Macroeconomic Conditional Pre-provision Net Revenue (PPNR) Forecasting

This webinar discusses the primary challenges confronting banks when forecasting macroeconomic conditional pre-provision net revenue (PPNR), best practices for forecasting macroeconomic conditional PPNR, and the tools and techniques used by Moody’s Analytics to address the challenges.

October 2013 WebPage Thomas Day, Dr. Amnon Levy, Robert Wyle

Stress Testing Webinar Series: Macroeconomic Conditional Loss Forecasting Presentation

In this Moody's Analytics webinar, Thomas Day and other Moody's Analytics experts discuss Macroeconomic Conditional Loss Forecasting. Given the criticality of loss estimation, and the need for different models by asset class, we cover loss estimation for Retail Exposures (non-mortgage), Structured Portfolios, Wholesale C&I (non-public), and Wholesale (public).

October 2013 Pdf Thomas Day, Dr. Cristian deRitis, Luis Amador

Emerging Standards: Liquidity Risk Management

Banks' funding activities and liquidity management will be the focus of increased regulator attention in coming years. This presentation helps risk managers meet best practices.

October 2013 Pdf Thomas Day

Summary of 2013 Comprehensive Capital Analysis and Review and Dodd-Frank Act Stress Tests

This article provides a summary of the 2013 CCAR and Dodd-Frank Act Stress Tests, and compares the results with the 2012 stress tests.

September 2013 WebPage Thomas Day

Breaking the Black Box - Managing Stress Testing, Capital Assessment and Risk Appetite Frameworks in a CCAR World

In an effort to improve transparency and increase confidence, global policy leaders have embarked on an ambitious agenda that has as its centerpiece improvements in data architecture and infrastructure, firmwide stress-testing, capital planning, and risk appetite frameworks. In the United States, this is known as the Comprehensive Capital Assessment and Review (CCAR). CCAR, Basel III, and enhanced liquidity risk management expectations aren't easy problems to solve. To

February 2013 Pdf Thomas Day