Quantifiable Risk Management Data Driven Approaches to Building a Predictive Risk Framework. Andrew Auslander, CFA, FRM

Transcription

1 Quantifiable Risk Management Data Driven Approaches to Building a Predictive Risk Framework Andrew Auslander, CFA, FRM

2 Quantifiable Risk Management Data driven Approaches to Building a Predictive Risk Framework Presentation to GRC Summit 2018 Wednesday, June 6, 2018 Andrew Auslander, CFA, FRM andrew.auslander@gmail.com

3 Overview 1. Risk Management challenges 2. Risk Management is shifting from a compliance-focus to a performance-focus 3. KRIs need to be tailored to business strategy and risk appetite 4. Integrating KRIs, KPIs, and KCIs 5. Data analytics and data mining to derive value from data 6. Example 1: Current Expected Credit Loss (CECL) 7. Example 2: Liquidity Risk Management / FTP 8. Conclusion

4 Risk Management s Challenges The risk management function is tasked with delivering information in almost real-time with the following challenges: Legacy applications, access to the correct data, no centralized, consolidated data sets Complex, time consuming process to finalize results and prepare dashboards for management, and regulatory reports for submissions To address these challenges, Risk Management needs risk dashboards to monitor all risk triggers and limits and an ability to have early warning indicators (KRIs) and performance indicators (KPIs), and ways of monitoring the efficacy of their controls Key Control Indicators (KCIs). In order for this to work, the firm needs good systems and data. 4

5 Risk Management is shifting from a compliancefocus to a performance-focus Leveraging Dodd Frank / CCAR infrastructure Return on investment Performance - Actionable information that allows the business to make better decisions that impact profitability 5

6 KRIs need to be tailored to business strategy and risk appetite Key Risk Indicators (KRIs) should monitor changes in level of risk exposure and act as early warning signs to prevent crises and allow time for mitigation. Attributes of the Indicators: Quantitative, meaningful, and predictive Derived from data produced by operational systems Provide early warning for risks that are increasing A method for aggregation and escalation of material changes Can be the basis for tactical and strategic decisions 6

7 KRIs need to be tailored to business strategy and risk appetite KRIs should be vetted by the subject matter experts at a firm to ensure they are the indicators that will move when root cause events start occurring Automating KRIs allows for easier tracking, escalation, and documenting for management and regulators Organizations should validate trigger levels and thresholds based on their risk appetite Should be tracked regularly Can be used to look for causal and trend relationships 7

8 Integrating KRIs, KPIs, and KCIs Key Performance Indicators (KPIs) measure performance. Firm performance can be negatively impacted by the risks tracked by KRIs. The business can react to changes in KRIs and potentially avoid deterioration of KPIs Key Control Indicators (KCIs) measure the firm s control effectiveness Ideally, technology should allow for a single interface for KRIs, KPIs, and KCIs in order to allow the clearest picture of the firm 8

9 Data analytics and data mining to derive value from data The first key is reliable and robust data. Detailed transaction and market data are required Think long term and gather data for internal analytics Digitize all data as close to the point of origin as possible Ideally, data is sourced from an integrated risk management system to minimize reconciliation Duplication Inconsistencies = Errors 9

10 First Example: Current Expected Credit Loss (CECL) The International Accounting Standards Board (IASB) and Financial Accounting Standards Board (FASB) began to work together on the development of new financial instruments standards, such as IFRS 9 and CECL following the recent financial crisis with the goal of providing more warning for credit events. Deloitte IFRS 9 and CECL represent the biggest change in accounting standards since standards were established. All of an organization s functional areas business, risk, accounting, modeling, and others should be aligned and work collaboratively to develop and implement new loss estimation models. 10

11 CECL: Applicability Loans / Receivables Debt Securities Held-for- Investment (HFI) (CECL) Held-to- Maturity (HTM) (CECL) Available-forsale (AFS) (Allowance for credit losses) 11

12 CECL Definitions Key components of the new standard include the following: Principles based Allowance for credit losses (ACL) Current expected credit loss model (CECL) Available-for-sale debt securities Zero-loss exception Purchased credit-deteriorated assets Disclosures Expected credit loss models 12

13 First Example: Current Expected Credit Loss (CECL) Source: Federal Reserve Board, Form FR Y-9C, Consolidated Financial Statements for Bank Holding Companies. Note: Delinquent loans are defined as loans that are more than 30 days past due or are in nonaccrual status. The blue shading shows periods of recession as defined by the National Bureau of Economic Research. 13

14 ALLL under Incurred and CECL with Perfect Foresight 14

15 CECL: Key Measurement Considerations Current Conditions Lifetime Losses Supportable Forecasts Recognition of credit losses will shift to an earlier point in the credit cycle Current conditions will differ by market (and type) The impact of conditions also differ measurably Lack of data or lack of quantitative support for management estimates Lifetime losses must be recognized at origination, rather than when default is deemed likely Cumulative losses are non-linear how to measure? Prepayments are difficult to measure, especially for commercial portfolios Differences by loan types Need to generate economic scenarios Also, to appropriately link scenarios to credit losses Ability to forecast beyond a year or two into the future is a challenge Unreliable forecasts will mean greater volatility in loss provisions Document everything 15

16 Stress Testing Process Should be Leveraged as the Starting Point for Forecasted Part of CECL Model Validation Scenario Analysis / Stress Testing Models used by ALM and Credit Regulatory Reporting Finance CECL / IFRS 9 Forecasts consistent with strategic plan and changes to initiatives in various scenarios All results need to comply with regional accounting rules 16

17 The Building Blocks of Stress Testing and Expected Loss Modeling are the Same Data Granular with rich dimensionality Model Integration Prepayment Loss Utilization Pricing Analytical Capabilities Forecasting Engine Market Value Estimation Cash Flow Generation Reporting Transaction and Portfolio Level Output 17

18 Expected Loss Requirements: Data Transaction Level Data All data regarding the contract Dimensions for model consumption Dimensions for pooling and aggregation for reporting Historical loss data Macro-Economic Data Historical Current Expected 18

19 CECL: But wait there is more - Methodologies - Models - CECL calculation - Data acquisition and management - Internal education and communications - Alignment with Planning - Financial statements and disclosures - Mergers, acquisitions and trading - Will your Board understand - Will your auditors understand - How will your customers be impacted - How will your shareholders react - Traceability and auditability - Document everything done, not done & why 19

20 Techniques and Expected Impact of CECL Techniques to be used for CECL FASB: CECL does not prescribe any methodologies for arriving at loan-loss estimates Certain examples have been released, which are guiding industry practice: Forecasting techniques using dynamically responding loss models (i.e. regression, migration, etc.) Valuation techniques to estimate the value of credit spreads or the value of credit losses (i.e. regression, credit spread valuation) Basel type expected loss estimation (Lifetime PD * LGD * EAD) Expected Impact of CECL The impact will result in an increase to the ACL, but also investments in improved credit risk measurement capabilities. Should allow a feedback loop to update business s pricing capabilities Allows management to know the sensitivity of the business to stressed economic conditions Can allow ECL metrics to be added to product profitability analysis (KRIs) 20

21 Second Example: Liquidity Risk Management Liquidity Risk Management Better data allows you to keep a smaller buffer and earn more from your core and excess liquidity. Accurate contingent risk pricing Analytical Capabilities Accurate and Current Data Accurate Liquidity pricing Providing Correct Incentives to Business Accurate capital pricing 21

22 The benefits of advanced technology and quality data for liquidity risk management and Funds Transfer Pricing (FTP) Productivity improves when systems are modern and data is reliable Efficient use of resources Requires technology Accurate data Opportunity cost of poor systems and/or data Funds transfer pricing Carrying cost of core liquidity Carrying cost of excess liquidity Accurate funds transfer pricing can help to prevent bad incentives Pricing capital accurately Pricing liquidity accurately Pricing contingent risk accurately 22

23 Conclusion Good systems and data for risk management can be justified for many reasons: Creates feedback loop for better management of business and pricing Planning and implementing KRIs, KPIs, and KCIs are critical to the success of the ERM process. Allows a view on the risks, performance, and controls of the firm Can improve profitability o Pricing of HFI loans and HTM debt securities can take into account current expected credit loss during lifetime of transaction o Addresses the opportunity cost of leaving a large liquidity buffer and investing in cash or very short term instruments Meet regulatory guidelines 23

24 Thank You Continue the conversation on #GRCSummit