Capital Buffer under Stress Scenarios in Multi-Period Setting 0
Disclaimer The views and materials presented together with omissions and/or errors are solely attributable to the authors / presenters. These may or may not represent the views of the Bank of Montreal (BMO Financial Group) and Moody s Analytics. The graphs in this presentation are not drawn to scale and are provided for illustration purposes only. 1
Capital Buffer under Stress Scenarios in Multi- Period Setting An integrated framework for measuring and managing Economic Capital and CCAR expected loss Managing an organization s expected loss and uncertainty in its capital surplus under CCAR scenarios 2
Capital Adequacy Framework: Measurement + Management Adapt culture to understand the value of the Capital Adequacy Framework & Stress Testing (CAF): seamlessly integrate related activities into normal business processes (e.g. strategy setting; business and capital planning; product and transaction evaluation) Embed CAF through the entire business management cycle including Strategy, Financial Planning, Business Decision Making and Performance Management; support with good governance. Governance Strategy Performance Mgmt. Core ICAAP CAF Financial Planning Business Decision Making Governance 3
Major Elements of the Framework Risk Appetite Risk Profile Business Strategy / Tactics Capital Demand Regulatory / Economic capital over 3+ v year planning horizon Quantitative and qualitative add-ons in prudential buffer to cover other v risks / stress losses Capital add-on v (general risks) Forecast Capital Demand Capital Adequacy Assessment Assess capital requirements v Set capital targets v Maximize Target risk- Capital adjusted Ratios return & Triggers by using each capital measure most appropriately Additional External & Regulatory Expectations Capital Supply Capital capacity from business plan v over three year planning horizon Planned capital v actions Contingency plans to maintain or restore v capital as warranted Total Capital Supply 4
Capital & Buffers Banks create buffers in the form of provisions for losses that might be expected to occur. However, actual losses are often different from expectations, and capital is held to cover unforeseen possibilities (UL). EL 99.95% Unconditional EC UL Conditional EC buffer 5
Capital & Buffers Until recently, the typical framework does not quantitatively estimate the additional EC (capital or EC buffer) required to survive the impact of a downturn in order that the unconditional capital may be effectively all available when needed. As such, these capital buffers are being estimated to take into account downturn scenario revenues, losses, reserves and pro-forma conditional capital levels (demand & supply; planned capital actions) Risk identification (definitions) Materiality (taxonomy, concentrations, risk mitigating factors) Scenario design Forecast impact to capital demand and supply Limits, Governance and Planning 6
Risk Identification Process Risk identification framework Systematic and repeatable approach to risk identification with linkage to businessas-usual risk management and measurement Risk taxonomy Catalog exposure assessment by risk type Provides for a common risk language across the firm Facilitates more complete and transparent materiality assessment Standardized process to facilitate material / non-material assessment of specific risk types for the Risk Register 7
Coverage: description and materiality assessment Risk taxonomy provides a catalogue of financial and non-financial risks Supports assessment of EC and Stress scenarios 8
Risk Register Describes the material risks categories that require Economic Capital to be held based on the information from the Risk Identification process Risk Type summary description of method used to estimate EC (details provided in a separate document: EC Implementation Guidelines) Capital Model Id / Name (by business e.g. Credit has 12 sub-categories) link to model inventory; which includes further details of feeder models, vetting conditions, assumptions, uncertainties, EC levels by Enterprise / Operating Groups / Lines of Business Bridge to model Regulatory Capital (Pillar 1 and 2) CBR EC Limits 9
Risk Register Material risks are defined and evaluated as to whether or not capital can be held as a mitigant 10
Risk Register (contd.) 11
U.S. Key Macroeconomic Variables: Wholesale portfolios 12
Canadian Key Macroeconomic Variables: Wholesale 13
Economic Credit Losses 14
Illustration: Stressed Credit Risk The portfolio composition over the projection horizon is conditional on the stress macroeconomic conditions. EC (unconditional) over a 1-year time horizon is calculated at each time period based on the updated portfolio composition. Portfolio Composition EL EC Portfolio Composition EL EC I I S S W W D % Actual 99.95 D % t2 99.95 Macro economic conditions Portfolio Composition EL EC I S W D % t1 99.95 15
EC stress scenarios Capacity to Bear Risk (CBR) is an estimate of the maximum amount of risk the Bank is able to assume given its capital base under current conditions. It represents the total capital of the Bank, with deductions for non-portfolio assets that may have materially impaired value under stress. EC Adequacy under Stress: Capacity to Bear Risk (CBR) in excess of EC forecast through all periods. Capital Supply (Capacity to Bear Risk) Conditional: [Supply Demand] Capital Demand (Economic Capital) 16
EC Framework to assess Capital Requirements Goal: Book Capital Goal: CBR CBR = Book Capital - Deductions Target: CBR Stress CBR EC Limit EC Limit Stress EC Spot EC Spot EC Demand Pick-up Supply Compression Limit Buffer Deductions: Goodwill, Intangibles, Uncertainties, etc. Main objective Ensuring adherence to risk appetite linking EC to a solvency target (debt rating) incorporating buffers on account of plausible scenarios defining capital requirement: post-stress basis + operating margin 17
EC limits setting linked to risk appetite and reporting CBR base scenario CBR stress scenario EC Limit Total EC stress scenario Total EC base scenario Not drawn to scale t 0 t 1 t 2 Q4/YY Actual Forecast Main objective Managing risk modifying behaviours to ensure risk taking is within limits capturing portfolio correlations and diversification Adequacy under business-as-usual conditions ensure that when starting from the limit, demand will not exceed supply under stress 18
Framework implied risk based capital (& buffer) allocation Operating Ceiling CET1 Capital Target (i.e. Operating Floor) CET1 Capital Goal CET1 (required post-stress) Operating Range Scenario Uncertainty Buffer Stress related compression in Supply vs. Demand Additional Risk Capital Basel Regulatory Capital (e.g. U.S. Basel III Standardized Adequately Capitalized Requirements CET 4.5%) Stress related compression in Supply vs. Demand Economic Capital : risk sensitive capital demand; business-asusual basis Scenario Uncertainty Buffer Stress related compression in Supply vs. Demand Operating Range Additional Risk Capital Basel Regulatory Capital (e.g. U.S. Basel III Standardized Adequately Capitalized Requirements CET 4.5%) Allocate proportionally based on scenario results or hold as a buffer at the enterprise level Allocate proportionally based on scenario results Allocate proportionally based on spot EC: business-as-usual risk sensitive capital demand This allows for business linked optimization a good understanding of the mix of return for risk across different groups / lines of business 19
Risk based optimization / planning EC related risk tolerance thresholds are informed by limits Financial Performance Management uses EC as one of the key inputs into riskadjusted return. EC is used as a relative measure in supporting business decisions and performance management across the Enterprise the annual planning process considers OG and LoB strategies that include forecast estimates of EC and RAROC performance can be assessed at various levels from individual products to sector to LoBs to Operating Groups. 20
Risk based optimization / planning The aim is to identify growth opportunities, good / poor return customers, credit portfolio concentrations and inform risk-sensitive pricing decisions limits, hold levels, etc. set with the objective of avoiding concentration use secondary market instruments and engage in buying, selling, and hedging transactions identify unprofitable customer relationships that can be cleaned out of portfolio address need to satisfy growth targets as well as potential future economic (and regulatory) capital constraints maintain pricing discipline in the front office and re-pricing or de-risking or pruning low RAROC customers opportunity identification through analysis of revenue (size) vs. total RAROC by geography and industry cherry-picking good customers / targeting greater share of total business while avoiding a price-war 21
Quantitative Approach to Modeling Multi-Period Capital under Stress Scenarios 22
Business Objectives Multi-period capital planning framework can be used to understand and compare dynamics of risk of a credit portfolio under various economic scenarios. calculate capital requirements of the portfolio under different economic scenarios. determine appropriate capital buffer level by testing whether capital buffer is able to withstand a particular economic scenario. 23
Business Objectives» understand and compare dynamics of risk of a credit portfolio under various economic scenarios. If economy worsens 24
Business Objectives» calculate capital requirements of the portfolio under different economic scenarios. If economy worsens 25
Business Objectives» determine appropriate capital buffer level If economy worsens 26
Methodology Illustration» For the purpose of the illustration we assume a constant composition portfolio portfolios can reflect the expectation regarding the composition (new volume, etc.) under various stress scenarios.» Portfolios at various time points will reflect the assumed stress conditions PD Term structures for instruments (from EL calculator outputs, stressed up to forward analysis date) Interest rate curves consistent with respect to CCAR scenarios Market risk premiums consistent with respect to CCAR scenarios» Economic Capital Analysis (done in RiskFrontier) : 1 year horizon» Analysis date valuation : Book» Horizon valuation: Lattice Unconditional forward analysis» There is no assumption about economic scenario beyond forward analysis date Conditional forward analysis» There is assumption about economic scenario beyond forward analysis date 27
Sample Portfolio Number of Obligors 4909 Number of Instruments 9818 Exposure Amount Exposure Weighted Average Maturity 4.503 Exposure Weighted Average RSQ 0.3636 $24.77 Billion Exposure Weighted LGD 37.0437» The 4909 obligors are selected as the whole universe of US public companies in the CreditEdge dataset.» Each obligor has two instruments (both term bullet loans), with varying maturities.» All instruments have annual coupon payments. 28
Stressed Forward PDs» The graph shows the exposure weighted average of 1 year stressed forward PDs of all the counterparties.» The forward PDs implied by original PD term structures are in line with the baseline scenario.» PD from the severely adverse scenario becomes smaller than that from the adverse scenario starting the 10th quarter.» PD from severely adverse scenario has its peak at the end of Q4.» The patterns can be explained by the paths of the stressed macroeconomic variables. 29
Stressed Macroeconomic Variables» Levels» Standard normal shocks» For US equity and US VIX, in the last few quarters, the levels are better in the severely adverse scenario than those in the adverse scenario. This helps explain the observation that PD from the severely adverse scenario becomes smaller than that from the adverse scenario starting the 10th quarter.» For the severely adverse scenario, except for unemployment rate, all other three variables have signs of recovery starting the 5th quarter. That is why we observed a peak in PDs at the end of Q4 for the severely adverse scenario. 30
Market Price of Risk 1.60 1.40 Lambda 1.20 1.00 0.80 0.60 0.40 0.20 0.00 Actual Model Baseline Adverse Severe 31
Interest Rate Curves Zero EDF curves, which serve as both discount and reference curves (for floating rate portfolio), are constructed by adjusting treasury yield curves from CCAR scenarios. Changes in Zero-EDF rates are the largest in adverse scenario, and the smallest in severely adverse scenario. 32
US Portfolio Unconditional Analysis» The graphs show 10 bp portfolio capital with respect to total spread for the fixed rate and floating rate portfolios.» The fixed rates and floating rates are obtained by using mark to par feature of RiskFrontier on analysis date. The same rates are used for all scenarios and all forward analysis dates since we assume constant portfolio composition. Furthermore, with the same rates, we can study the effects of different coupon structures. 33
US Portfolio Unconditional Analysis» Difference between capital levels of the fixed rate and floating rate portfolio is the smallest in severely adverse scenario, and the largest for adverse scenario.» This is due to the fact that changes in Zero-EDF rates are the smallest in severely adverse scenario, and the largest in adverse scenario.» Capital is larger for the fixed rate portfolio when projected future Zero-EDF rates are higher than its analysis date level.» Capital is smaller for the fixed rate portfolio when projected future Zero-EDF rates are smaller than its analysis date level. 34
Capital Distribution Dynamics» Capital distribution dynamics can give a more comprehensive picture of how the risk of the portfolio evolves over time under an economic scenario.» Capital distribution dynamics can be used to test the adequacy of capital buffer over time under an economic scenario. Set the capital buffer threshold on analysis date (i.e. 10 bp capital) Calculate the probability of breaching the capital buffer threshold for each future analysis date in each scenario The larger that probability, the less adequate the capital buffer is in that scenario at that time. 35
Capital Distribution Dynamics (Baseline, Floating Rate) unconditional conditional» For unconditional RF capital distributions, probabilities are always around 10bp.» For conditional RF capital distributions, probabilities are much smaller than 10bp (less than 1bp).» Capital requirement remains relatively stable. 36
Capital Distribution Dynamics (Adverse, Floating Rate)» Under unconditional RF approach, probability is always greater than 10bp. It is the largest at the end of Q3 at 1.2%.» For conditional RF analysis, mean of the capital distribution is greater than that from unconditional RF analysis. However, capital requirement is smaller starting end of Q1. This is because, even under stressed scenarios, PDs can decrease for low PD (high quality) firms, thinning the tail of the distribution around the area of 10bp. 37
Capital Distribution Dynamics (Severely Adverse, Floating Rate)» For unconditional RF capital distributions, probability is the largest at the end of Q4 at 35.7%. This is consistent with the paths of stressed variables: clear sign of recovery starting Q5.» For conditional RF capital distribution, probability is the largest on analysis date at almost 12%. Conditional analysis on analysis date (at time 0) takes into account scenarios from Q1 to Q4, the phases of major deterioration of the economy.» Capital requirement is larger than that for a fixed rate portfolio due to a decrease in zero-edf rates compared to analysis date. 38
Capital Distribution Dynamics (Baseline, Fixed Rate)» Probabilities increase for both unconditional and conditional RF capital distributions.» For conditional RF capital distributions, probabilities are smaller than 10bp.» Capital buffer at 10bp level on analysis date appears to be significantly more resilient under forward conditional analysis compared to the unconditional one. 39
Capital Distribution Dynamics (Adverse, Fixed Rate)» For both unconditional and conditional RF capital distributions, probabilities soar in later quarters, reaching almost 36%.» Unlike under baseline scenario, 10bp analysis date capital is not sufficient to cover losses under adverse scenario.» Inadequacy of capital is mainly due to (i) the paths of the stressed variables (ii) much higher zero-edf rates in the future compared to the analysis date. 40
Capital Distribution Dynamics (Severely Adverse, Fixed Rate)» For unconditional RF capital distributions, probability is the largest at the end of Q4 at almost 13%. This is consistent with the paths of stressed variables: clear sign of recovery starting Q5.» For conditional RF capital distributions, probability is the largest on analysis date at almost 12%. Conditional analysis on analysis date (at time 0) takes into account scenarios from Q1 to Q4, the phases of major deterioration of the economy. 41
Compare Capital Distributions (Example) Baseline Adverse Severely Adverse Fixed (End of Q8) Floating (End of Q8) Floating rate provides larger compensation for risk in baseline scenario due to an increase (compared to analysis date) in the zero-edf rates projected. Means of the two capital distributions are aligned. Means of the two capital distributions are not aligned. Generally, larger losses are generated under conditional forward analysis. However, capital distribution from unconditional forward analysis has fatter tail. Floating rate provides larger compensation for risk in adverse scenario due to an increase (compared to analysis date) in the zero-edf rates projected. Means of the two capital distributions are aligned. Means of the two capital distributions are not aligned. Generally, larger losses are generated under conditional forward analysis. However, capital distribution from unconditional forward analysis has fatter tail. Fixed rate provides larger compensation for risk in severely adverse scenario due to a decrease (compared to analysis date) in the zero-edf rates projected. Means of the two capital distributions are not aligned. Larger losses are generated under unconditional forward analysis. Means of the two capital distributions are not aligned. Larger losses are generated under unconditional forward analysis. 42
Stressed PDs» Analysis date average PD for financial firms is smaller.» Financial firms are more exposed to systematic risk (with exposure weighted average RSQ of 0.431) than nonfinancial firms (with exposure weight average RSQ of 0.335). This results in higher sensitivity of stressed PDs of financial firms to scenarios. 43
Sector Effects of Stress Scenarios» Financial firms expected losses are more responsive to the scenarios due to higher sensitivity of their stressed PDs to shocks of stressed macroeconomic variables.» 23% instruments are from financial firms, which account for 29% value of the portfolio.» Proportion of capital allocated to financial firms is always greater than their value proportion.» Capital proportion for financial firms increases for most of the periods under adverse and severely adverse scenarios. 44
Summary Goal:» Develop a Multi-period capital planning framework that can be used to understand dynamics of risk of a credit portfolio under economic scenarios. calculate capital requirements of the portfolio under different economic scenarios. determine appropriate capital buffer level by testing whether capital buffer is able to withstand a particular economic scenario. Methodology» The framework can be implemented by combining stressed EL calculator and RiskFrontier. stressed EL calculator produces stressed PDs RiskFrontier uses stressed PDs as inputs to calculate economic capital» Two types of RiskFrontier analysis unconditional : There is no assumption about economic scenario beyond forward analysis date conditional : There is assumption about economic scenario beyond forward analysis date Results» The analysis allows more comprehensive and granular understanding of the implications CCAR scenario assumptions have on the risk profile of a credit portfolio. Capital buffer behaviors vary significantly under different scenarios. Severely adverse scenario implies much stronger economic recovery compared to adverse scenario, resulting in counterintuitive observation that capital requirement is less stringent for severely adverse scenario. 45
Thank You 46