Presentation on theme: "FHLB Income-Based IRR Measurement: Alternative Approaches and Issues"— Presentation transcript:
1FHLB Income-Based IRR Measurement: Alternative Approaches and Issues - Potentially Useful Lessons from the Private Sector -
2Classification of IRR Measurement Techniques AgendaBackgroundClassification of IRR Measurement TechniquesOpportunities and Challenges of:Stochastic income measuresEarnings-at-Risk (EaR)Background: Citicorp’s Risk Measurement Challenge ~1987Background: Citicorp’s SolutionApplication to Income Output from an FHLB using QRM or BancWareApplicationsRisk LimitsDecomposition of the IRR Measure: Improved Understanding and ManagementHedging both Income and ValueRegulatory
3BackgroundIRR measurement and management in private banks is largely focused on reported incomeMortgage banks more oriented toward value because accounting is closer to market value accountingBank America implemented an IRR measurement solution to hedge earnings and value simultaneously in 1990sIRR measurement in many FHLBs has focused on controlling value-based risk measuresFHLBs and its regulators are starting to place more emphasis on income-based risk measures and effects on retained earningsThe private sector has implemented methodologies that are potentially useful to FHLBs and their regulators
4BackgroundSome methodological issues that arise in private banks when measuring income-at-risk:Which definition of “income” to modelHow to simulate interest ratesWhether to include new business assumptions and how to vary new business assumptions in different rate scenariosOver what period to model income-based risk (a.k.a., the “time horizon” problem)How to set risk limits for income-based IRRAll of these questions are relevant when designing income-based risk measures
5Three Dimensions of IRR Measurement Methodologies Time HorizonExisting Only vs. Existing + NewDeterministic vs. StochasticBoth value and income based IRR measures can be categorized using this three dimensional framework
6Three Dimensions of IRR Measurement Methodologies DeterministicIncome BasedStochasticExisting Business OnlyExisting + New BusinessDeterministicValue BasedExisting Business OnlyStochastic12 – 18 Months18 Months to ~5 Yrs360 MonthsShort Term Medium Term Long TermTime Horizon
7Income Based IRR Measurement Methodologies Existing Business OnlyDeterministicIncome Based IRRStochasticExisting + New Business12 – 18 Months18 Months to 5 Yrs360 MonthsIncome Based Methodologies
9Classification of Useful Income Based IRR Measurement Opinion: There are only three approaches to measuring income at risk that offer risk managers much value added and one of them is very complicated and beyond the capabilities of vendor based ALM systems.
10Opportunities and Challenges of Stochastic Income Measures When used with the right software it’s the only methodology available to optimize hedges when hedging from both a value and income perspectives using stochastic methodologiesFor portfolios where value and income accounting are aligned then the potential issues are minimizedWhen mortgage bankers use value based stochastic risk measurement tools they are also approximately hedging income
11Opportunities and Challenges of Stochastic Income Measures Difficult to compute:New business equations are difficult to specify and results are sensitive to these assumptionsExcluding new business helps, but in private sector defining new business for core deposits is assumption intenseMost balance sheets requires two yield curves to simulate unless basis risk is ignored or work-arounds appliedResource intensive to get a credible measureNot easy to produce a validated measure in QRM.BW’s stochastic model is inferiorNot available in trading models with superior stochastic engines that focus on value based risk measurementNot aware of any commercial bank that is using stochastic income for hedge design.
12Opportunities and Challenges of EAR Scenarios can be predefined shocks of almost any form or “what if” scenariosALM models are built for this type of analysisVery useful for short term analysesEasy to understand and communicate resultsRisk attributes can be computed
13Opportunities and Challenges of EaR Can be misusedALM models allow targeting balancing procedures, which can mask riskLimited time horizon for analysis allows risks to be pushed “beyond the radar”Hedging transactions often beyond the time horizonNot useful for assessing long term and strategic risksRisk limits are not applicable when new business sensitivities are includedMany users do not know how to decompose risk into characteristics and can generate “non-actionable” results
14Citicorp’s Risk Measurement Challenge ~ 1987 Background:7 retail banks, 3 thrifts, a mortgage bank, and large credit cards businesses with decentralized management structureCorporate management was concerned that smaller thrifts could take a risk position and bankrupt the corporationPerceived need to develop common, understandable, and actionable risk management metrics and language across multiple management unitsRequirement to understand risk of the combined unitsRisk measures were needed to limit risk in a way that could not be “gamed” by new business assumptionsNo vendor solutions were available
15Citicorp’s Solution“SMEAR”: Spot Measure of Earnings-at-RiskDesigned by Gary Lachmund, former President of National Asset Liability Management Association (NALMA) and then head of ALM at CitibankOriginally developed proprietary model in-house; Can (now) be easily generated in vendor ALM modelsComplementary to analyses of risk that do include new business sensitivitiesAddresses several of the issues relevant to measuring income-based IRR in the FHLBs by the regulatorsWas utilized to limit risk of short- and long-term earnings sensitivity
16Application to FHLBs and FHFB Provides method a solution to “time horizon problem”Easily produced in BancWare and QRMMeasures are complementary to income-based risk measures that include new businessPotential to creates common methodology across 12 regulated banks so risk measures can be consolidated and comparedRegulators can measure position of systemRegulators can rank order positions of individual FHLBs
17SMEAR Procedures Start with current balance sheet Shock interest rates instantaneously, by multiple incrementsMay use flat rates or forwards, but forwards are preferredKey: all rates shocked same amount*Run-off balances based on contractual maturity- or model-based prepayment in each scenarioAs balances run-off replace with overnight funding or placements (a.k.a. the balancing item) at scenario- dependent rateFor repricing assets and liabilities reprice according to contractual rulesAllow no new business* i.e parallel shocks; This assumption eliminates repricing effects from analysis
18SMEAR Procedures Treat equity as an indefinite term maturity item Compute “Pretax Rate Sensitive Earnings” (PRSE) in each scenario and as many time periods as relevantThis allows for fee income, direct expenses, and gains-on-saleGenerates a matrix of solutions for each time period and shockCalculate differences in each time period relative to the base caseGraph the calculated differences
19Definition of Income Applicable to an FHLB Risk Measure Income measure = the net revenues in each time period associated with the book of existing business (i.e. “the risks you already own”) or “NII associated with Existing Book of Business” (NII-EBS)Gains-on-sale are not currently a component of income sensitivitySince this measure explicitly excludes net revenues associated with new business, it does not fall into one of the standard income definitionsFHLBs have derivatives that do not qualify for hedge accounting. The NII-EBS incorporates these obligations by calculating net cash flow differences as their contribution to the income-based risk measureIn order to accommodate a GAAP earnings measure, market value sensitivity of derivative instruments not qualified for hedge accounting treatment can be added back into the analysis separatelyUsing a blend of market-value accounting and accrual accounting in an income-based risk measure can lead to non-economic risk management decisions
20Notes on Long Term Earnings at Risk Measure Focus is on “the risks you own” (certainty) vs. risks you only incur over time in an uncertain futureIgnorance of long term earnings effects can lead to risk positions that increase longer-term exposures that are off the radar screenMarket value sensitivity analyses is not a substitute for longer term earnings exposures
21SMEAR Calculation Steps Step 1: Calculate NII-EBS in each period for the base (“expected” or forward curve) scenario and for each “rate shock” (or “stress”) scenarioStep 2: Subtract NII-EBS shock scenario values from those of the FWD caseStep 3: Plot the value changes for each stress scenarioStep 4: Connect the dots
22SMEAR Example: Income-Based Simulation Results Step 1: Calculate NII-EBS in each scenario and period
23Transforming Income Simulations to Risk Measures Step 2: Calculate DNII-EBS relative to FWD case
24Transforming Income Simulations to Risk Measures Step 3: Plot the relative values
25Transforming Income Simulations to Risk Measures Step 4: Connect the dots
26We’ve transformed tables to graphs. SummarySo far:We’ve transformed tables to graphs.We’ve extended the time horizon for income-based risk analyses.Time horizon can be extended as far into the future as needed for controlling longer term earnings sensitivity associated with the existing balance sheet.Number of shocks can be added so that a broader range of rate shocks is applied as the time horizon is extended
27Further Application can Extend the Benefits of the SMEAR Further ApplicationsFurther Application can Extend the Benefits of the SMEARRisk Measurement Technique:Application I: Risk limits in the SMEAR frameworkApplication II: Decomposition of riskApplication III: Ability to assess value-based hedging on income-based IRRApplication IV: Regulatory
28SMEAR RISK LIMIT FRAMEWORK: Application I Width of rate shock band can be linked to observed market volatilitiesSize of rate shock may vary with the direction of shock if view is that rates are approximately log-normally distributed.Note that the width of the limit is no longer tied to the exact rate shock used in the calculations.Risk limit may vary by time period or direction of shock due to expected offsets in new business activity
29SMEAR RISK LIMIT FRAMEWORK: Application I Limit violation marked in “X” occurs when line intersects the bottom of the SMEAR “limit box”Size of shock utilized in limit increases with time, as does size of limitIncome limits in future periods typically become less restrictive because opportunities exist to mitigate the riskCiticorp limits were invoked out to Year 10, requiring a broader range of rate shocks than shown
30Challenge: “Rates Don’t Move in Parallel Shocks” Citicorp limits were invoked out to Year 10, requiring a broader range of rate shocks than shownThe purpose of risk measurement and a risk limit system is to guide risk management actions“Actionable understanding” is critical. Graphical framework translates to a visual picture of risk components and points the way to managing riskThe actual number used to limit risk is a proxy and shouldn’t be equated with “what if” analysesSetting of size of actual risk limit in each case (defined by direction of shock and time) is critical component of system.The limits should take account of evolution of new business but not the evolution of new interest rate risk positions
31Decomposing income-based Risk Measure: Application II Why Decompose income-based IRR?Decomposition of income-based IRR is a:Risk communication tool, because portfolio composition effects are difficult for the non-technical audience to comprehend (e.g., some members of ALCOs)Risk measurement validation tool, because specific risk measures can be ascribed to individual product characteristics and errors can frequently (but not always) be seenRisk education tool, because it reduces the complexity associated with understanding complex risk characteristics and, therefore, builds broader understanding of the complexity risk management among treasury and non-treasury professional staff
32Decomposing income-based Risk Measure: Approach With instantaneous parallel rate shocks income-based risk can be decomposed into:Repricing Risk: caused by mismatches in the repricing characteristics of assets and liabilities already on the balance sheet; andOption Risk: caused by the options embedded in the structures of financial instruments (e.g., prepayment, calls, and puts)Basis Risk can be added to option and repricing risk by shocking the CO curve by a different amount than the LIBOR curve and adding the results to those generated with parallel shocks
33Decomposing income-based Risk Measure: Approach Yield Curve Risk is directly calculated from product-level decompositions of option risk.Whereas, basis risk can be added to the other risk calculations in the SMEAR framework, total calculated yield curve risk is partially duplicative and cannot be addedRepricing risk component of yield curve risk has already been calculated by shocking interest ratesMissing component is options related effects which can be discerned at the product levelIf desired, income limits can be applied to options risks directly
34Decomposing the Income-Based IRR Measure: Example Total IRR=Repricing Risk Options RiskTotal IRR = Repricing Risk + Options Risk
35Decomposing income-based IRR: Repricing Risk Repricing risk is the sum of the “implicit” repricing exposures on each product type. However, it can be calculated at any level of aggregation, including the entire balance sheet.Aggregate measure of repricing risk includes equity.Repricing risk is best viewed at the balance sheet level. Focusing on offsets at the product level can introduce undesired noise at the balance sheet level.When rates are shocked by equal amounts, repricing risk is “linear” in the risk graphsSince fix-pay (or fix-receive) swap risk profiles are also linear, the mitigating transactions that reduce pricing risk can be easily identified and calculated.Swaps can be designed to be almost “perfect” hedges of measured repricing or “Gap” risk
36Decomposing income-based IRR: Option Risk Option risk is the sum of the options exposures associated with each product.It can be calculated at the aggregate level by subtracting repricing risk from total risk. However, graphical representations of options risk can be complicated when more than one type of option is present.Options risks are best hedged with options, although options exposures are frequently partially hedged with swapsMeasurement of options related risks are highly model sensitive because the exact conditions determining when the option is exercised are often based on specific modeling assumptions.Whereas repricing risk is best analyzed at the balance sheet level, options risk is better understood at the product level.
37Decomposing income-based IRR: Option Risk Identified Embedded Options in the Illustrative FHLB Balance SheetTotal Option Risk equals the sum of options risks embedded in all products and derivative instruments
38Decomposing income-based IRR: Option Risk Classification Scheme for Graphs to Follow
39Decomposing Options Risk : Prepayment and Call Risk Callable Agencies have no extension risk, unless they are expected to be called in the Forward Rate shock. Mortgages have extension risk as prepayment speeds slow relative to those modeled in the Forward Rate shock.Note: Graphs are not drawn on same scale.
40Decomposing Options Risk: Cancelable Advances & COs Cancellation features in CO portfolios raise the average coupon in lower rate levels. In turn, this raises income relative to the forward scenario. In the illustrative balance sheet the CO portfolio was far larger than the Advances portfolio.
41Decomposing Options Risk: Derivatives with Options Swaptions include options to purchase fixed receive as well as fixed pay swaps. There is greater prevalence of cancelable swaps than swaptions and caps observed on FHLB balance sheets.Note: Graphs are not drawn on same scale.
43Value vs. income-based IRR Hedging: Application III BackgroundBank of America built a stochastic interest rate model that calculated both income and economic value simultaneously. The model incorporated consistent simulation of two yield curves (Treasury and LIBOR)An optimizer was constructed to find hedges that minimized both value-based and income-based risk measuresA trade-off was calculatedGiven senior management input on preferences for minimizing variances of value and income over time, an optimal hedge solution was calculatedSeveral FHLBs are designing hedges focused exclusively on value-based IRR measures and have asked:How will value-based hedges impact income-based IRR measures?What methodology can be employed to measure the impact of value-based hedges on income-based IRR measures?
44Value vs. income-based IRR Hedging: Application III Considerations and an Approach using SMEARThe income-based risk measure that most coincides conceptually to the value-based risk measure includes long-term earnings and excludes new businessBank America findings from hedging from both perspectives:The size of hedge adjustments varied by productAdjustments could be thought of as duration neutral adjustments to the cash flow timingSignificant improvement to reducing earnings variances that did not sacrifice value based risk measure could be determined by trial and error
45Value vs. income-based IRR Hedging: Application III Simple SMEAR Test on Current FHLB PositionsCalculate the SMEAR risk in two subsequent time periodsUse risk measures in each period to evaluate the effects of value based risk measures on income at riskSubtract the risk measuresThis is called “Delta SMEAR”Use Delta SMEAR to evaluate the stability of the value based hedge in term of income based riskIterate the process and adjust the hedges accordingly
46Regulatory Extensions and Applications: Application IV Regulators Need an Income-Based IRR Methodology that Can Be:Applied consistently across 12 independently managed FHLBsUsed to assess risk at each bank as well as all banksUsed to assess relative risk of 12 banksUsed to limit risk at individual banksRegulatory limits can be set relative to individual FHLB’s “real” capitalTotal risk of 12 FHLBs can be limited and limits can be allocatedProduced with minimum additional effort, utilizing QRM or BancWare modelsUsed in conjunction with FHLBs other risk measures
4715 Fairway Drive, Novato CA 94949 Contact InformationALCO Partners, LLC15 Fairway Drive, Novato CA 94949Mike Arnold, Principal (415)Bruce Campbell, Principal (949)