Using ALM Models for PPNR and Securities OCI Peter Stoffelen September 27, BOARD OF GOVERNORS of the FEDERAL RESERVE SYSTEM

Transcription

1 Using ALM Models for PPNR and Securities OCI Peter Stoffelen September 27, 2016 BOARD OF GOVERNORS of the FEDERAL RESERVE SYSTEM

2 Disclaimer The opinions expressed in this presentations are those of the author and do not necessarily represent those of the Federal Reserve Board or its staff. 2

3 Stress testing expectations are closely aligned with ALM Capture all key risks and exposures Stress testing framework should include activities and exercises that are tailored to and sufficiently capture exposures, activities, and risks Maintain a thorough and well-managed process Employ multiple conceptually sound stress testing activities and approaches Be forward-looking and flexible Include strong governance and effective internal controls Obtain useful results Stress test results should be clear, actionable, well supported, and inform decision-making 3

4 Adjusting AL models for NII forecasting Chart of accounts management Data aggregation and current position Stratification Business line forecasting New business volume/originations and mix New business pricing (index and spread) Unexpected paydown modeling Prepayment modeling Estimates for nonaccrual balances 4

5 Data aggregation & stratification Lever common aggregation points for ALM and PPNR modeling PPNR categorization is often directed by Y-16, which can be a common point of aggregation Accomplished by matching roll up points in the ALM and PPNR chart of accounts Follow well-defined stratification routines Within the AL Model set up stratification routines are in place to already aggregate like positions (i.e. loans grouped by coupons, products, maturity dates, etc.) Data should be sufficient to capture embedded optionality within the products Data should be sufficiently granular to allow appropriate risk modeling and forecasting 5

6 How balance assumptions are derived: Targeted average balances Most cash flow simulation models will determine the bottom up growth / new volume that is needed to meet an average balance This target average balance is based on the balance sheet forecast from the line of business projections E.g., deposit balance projections Targeted new originations The firm targets an origination volume directly vs. providing an average balance This approach is often used for mortgage, auto and credit card portfolios Discretionary Portfolios Portfolios such as investment holdings and wholesale funding composition are discretionary, determined by Treasury or other based on loan and deposit projections 6

7 Projecting the NII component of PPNR For ALM, the purpose is to isolate risk to earnings from interest rate changes in isolation Key assumptions needed to produce the forecast are generally the same as for ALM Key differences relate to assumptions that are driven off of broader macro variables (e.g., HPI, unemployment, etc.) Impact from macro variables typically are incorporated through the following key assumptions: New business activity (balance and spread) Translation of non-performing loans Prepayment assumptions For PPNR, the purpose is broader, and reflects changes to macro variables in addition to interest rate changes Examples include Decline in economic conditions, which result in reduced new business Change in product demand (i.e., deposit mix) Impact of declining credit conditions (level of non-performing loans) 7

8 Balance Projections for PPNR Stress testing framework should include activities and exercises that are tailored to and sufficiently capture exposures, activities, and risks PPNR projections should reflect a best estimate of how portfolios would migrate under the set of macro variables assumed The process should be forward looking and flexible Key is the level of support provided to generate balance projections Approach some firms take: Develop top of the house regressions using industry experience to project balances Incorporate those regressions into the ALM model using equation writing and what-if capabilities Equations allow flexibility in scenario stress testing Governance and review process follow normal ALM channels How does your firm project balances? 8

9 New business spread projections for PPNR Spread projections for ALM is typically representative of current pricing structure of the firm What spread are we originating new business relative to objectives of the portfolio? Priced relative to firm s competitors (Broad market or subsector of market) For stress scenarios, spreads should reflect changes in market demand to factor in increase in either credit or liquidity risk For assets, stress scenario would assume widening therefore benefit Deposit liabilities vary more relative to scenario assumptions Firms generally review spreads relative to stress to determine impact on reported results 9

10 Prepayments and Non-performing Loans Prepayment Assumptions: Prepayments can have a substantial impact on modeling balances Almost all loan portfolios will have some degree of prepayments; however, mortgage is the most volatile to changes in interest rates Typically prepayment models are tuned for ALM projections, but should be adapted for PPNR use How should tuners be adjusted? Removed? Unscheduled repayments should consist of both voluntary prepayments (behavioral) and un-voluntary(defaults) Need to consider factors other than interest rates, such as HPI (ADCo) Non-accrual loans: For deteriorating credit impacts both credit losses as well as level of non-performing loans (migration of defaults) Projections are provided by credit divisions within organizations, Focus on capturing credit changes associated with scenario conditions 10

11 Other Benefits from using ALM model for PPNR Leverage model for sensitivity testing Streamline model validation and internal audit Leverage ALM governance and controls in place 11

12 Leverage model for Sensitivity Testing ALM and PPNR are best estimates for a given set of assumptions ALM scenario projections are a best estimate for a given rate change PPNR scenarios are best estimate for a rate change plus other macro variables How can those reviewing the scenario results gain comfort? Which macro variables drive the results more? Best solution would be to run various level of sensitivity test to allow reviewers to gauge the criticality of the assumptions Also helps to demonstrate whether variables other than rate changes are impacting projections Source of support to determine materiality to projections Strong foundation for demonstrating effective challenge 12

13 Streamline model validation and internal audit Leverage what you already have! Just need to document the differences Model validation ALM models are validated already- so step 1 is complete Only modeling approaches modified for PPNR need to be then validated (i.e. balance regression Audit- Internal & External The model has already been vetted by audit groups Document the approach differences between PPNR and ALM 13

14 Leverage ALM governance/controls in place Senior management and staff are already familiar with model More time spent on evaluating results vs. infrastructure Leverage common reporting platforms Process and procedures developed for IRR can be easily modified for PPNR stress testing process Audit review Regulators review Value added benefit of improving ALM given expanded use for PPNR On-going process that improves data as well as information provided to senior management and the Board 14

15 OCI valuation for stress testing Drivers of Fair Value Valuing Agency MBS Securities Reinvestment Assumptions 15

16 Securities Valuation Key drivers to determine market value: Changes in cash flows due to prepayment projections given scenario conditions Absolute change in interest rate curves Spreads on holdings required to cover credit and liquidity concerns 16

17 What to do with Agency MBS securities? For more complex holding such as agency MBS securities, stochastic modeling is used for ALM for starting positions (P0) What about the forward looking quarters within the stress scenario (P1 to P9)? Should stochastics be run each quarter within the forecast? Has your regulator outlined specific expectations to you? 17

18 Reinvestment Assumptions Given investment securities balances and mix are discretionarily derived, questions you should be able to address to support decisions are: How did you determine the allocation of reinvestment activity in securities? Are you shifting investment strategies into higher level of lower sensitive securities? If so, why? Understand changes in investment portfolio balances between ALM base case and the PPNR stress scenarios 18

19 BOARD OF GOVERNORS of the FEDERAL RESERVE SYSTEM Questions? 19