Modeling Originate & Hold Portfolios in. Moody s Analytics - Risk Practitioner Conference

Transcription

1 GE Capital Modeling Originate & Hold Portfolios in RiskFrontier Moody s Analytics - Risk Practitioner Conference Stefano Santilli October 16, 2012 The views expressed are solely the author s and not those of the General Electric Company. None of the materials presented contain any investment advice or guarantees of success and any use of the methods presented is at your own risk

2 Abstract Mark-to-Market/Model ( MtM based on Lattice Valuation) is the most commonly used valuation setting in RiskFrontier Its use is recommended in most cases and it is typically the most appropriate choice for modeling trading assets, securitizations, and all cases where an analysis of the market value of the assets is critical Modeling the Ecap of an Originate & Hold portfolio of loans and leases (banking book) with Lattice Valuation or other market value-based approach, when the portfolio is not actually marked-to-market, could lead to: Results not compatible with the institution s practices Adverse selection in the comparison of alternative origination choices Sub-optimal amount of Ecap in good and bad times The use of Mark-to-Par Spreads alleviates the issue and obtains sensible Ecap measurement; then risk-adjusted return metrics can be calculated outside of RiskFrontier 2

3 GE Capital: A Diverse Portfolio with Global Footprint Latin America 27% Europe Geography ($0.5T assets) 6% Asia Pacific 1% 11% 3% 54% 27% 4% 11% Canada 3% 2% Other 13% U.S. Industry sectors Product Health Care 6% Energy 6% Automotive 3% 3% Others Commercial 29% Consumer Services 4% 4% Consumer JVs Other Sales Finance 2% Auto 5% 4% Personal Loan 4% 9% Commercial Airlines 17% 7% 20% Real Estate 3% Structured inv. 5% Business Services Small and Medium Enterprises Mortgage 30% 39% Cards 3

4 What is Capital for? Economic capital is typically defined as the amount of capital needed for the financial institution to stay solvent. In VaR terms, it is the amount needed to cover potential losses that will be exceeded only with a certain probability (CL) There are at least two very different ways of interpreting losses and therefore capital : Capital as the amount needed to repay lenders in case of principal losses on the risky assets Capital as replacement, held to fill the gap due to market-value losses on the risky assets The first definition of capital is more appropriate for originate and hold portfolios that are not marked-to-market, because it better reflects the institution s practices and considers all losses as credit losses, thus avoiding counterintuitive results 4

5 Issues with Using MtM for an O&H Portfolio: Example Let s consider the theoretical case of three potential loans* that have identical risk characteristics, but different returns: Common Risk Inputs PD 4.55% LGD 35% R2 25% Maturity 3y Principal Amortizing EAD $ 1,000,000 If we could choose which deal to originate, which one would we pick? Is Deal C the obvious choice when looking at the RiskFrontier results? * The three deals were run in RF as part of a portfolio of ~500 loans and ~$10B of exposure 5

6 Issues Cont d MtM (Lattice) at Time of Analysis The MtM value at the analysis-date may accurately reflect the NPV of the different deals The MtM of Deal C correctly reflects that it has a higher value than the others For an O&H portfolio, it is interesting information, possibly useful for pricing purposes, but does not reflect how the deals are managed (and accounting treatment). No corresponding loss or gain is recognized 6

7 Ecap (RC EL ) Issues Cont d EL and Ecap Ecap (TRC EL ) EL and Ecap (in excess of EL, both RC- and TRC-based) are higher for the deals with higher returns Conceptually, the EL and Ecap should be the same, since the risk drivers are the same Actually, it could be argued that the Ecap should be even lower for Deal C than for Deal A, since it receives larger payments until default Is Deal C the obvious choice when looking at the RiskFrontier results? 7

8 Issues Cont d RORAC RORAC (RC EL ) RORAC (TRC EL ) Drawn Spread Drawn Spread The RORAC of the three deals is very similar; even slightly lower for the deal with the higher return Deal C would not be considered better than the others, leading to adverse selection 8

9 Issues Cont d - 2 nd Example Let s consider a different set of three potential loans that are much less risky. Like before, they have identical risk characteristics, but different returns: Common Risk Inputs PD 0.39% LGD 35% R2 10% Maturity 3y Principal Amortizing EAD $ 1,000,000 If we could choose which deal to originate, which one would we pick? Is Deal F the obvious choice when looking at the RiskFrontier results? 9

10 Issues Cont d - 2 nd Example - Results RORAC (RC EL ) Ecap (RC EL ) RORAC (TRC EL ) Ecap (TRC EL ) EL EL Drawn Spread Drawn Spread The higher the return: The higher the EL The higher the Ecap (both RC- and TRC-based) The lower the RORAC This type of results is not an exception, but the norm 10

11 Undesired Effects of MtM for O&H Deals: Summary While important for certain purposes (trading, securitization, capital setting for heldfor-sale assets, etc.) it is not appropriate for setting capital for assets originated to be held-to-maturity and not marked-to-market. Two major issues: Assumption Effect Problem All embedded gains/losses have been already realized at time of analysis For deals with embedded loss, EL and ECap are reduced. For deals with embedded gain, they are increased. Misleading/difficult to interpret Ecap values The definition of loss (and therefore the capital allocated) includes future unearned income If two deals have identical risk characteristics but different return, the one with higher return typically is allocated higher capital. In many cases, the risk-adjusted return is somewhat lower for the higherreturn deal Potential adverse selection What can be done to solve these issues? 11

12 Mark-to-Par Setting in RiskFrontier From the RiskFrontier User Manual: The terms and conditions of a loan or bond are used to calculate spread or fee information for the instrument. If this option is selected, during an analysis, this value replaces any fee or spread information previously provided. Supported instrument types include: Bonds, Term Loan Bullets, Term Loan Amortizing, Revolvers, CRE Loans (Non Mortgage), and CRE with Fixed or Floating Rate. Valuation method during an analysis must be Lattice/Lattice. This option is not available for Retail instruments, CRE Mortgage Loans, or instruments with changing rate types, such as Fixed then Floating. Bond/loan options such as Prepayment, Putable, Callable, Rating Price Grid, or PD Price Grid are not currently supported. Mark-to-Par spreads are the calculated spreads that make the instruments have a Lattice-based MtM t0 of Par All other user-inputted fees are ignored Achieves neutralization effect on returns by assuming that the spread exactly compensates the lender for the risk taken (in expected value terms) 12

13 Mark-to-Par Results: 1 st Example Common Risk Inputs PD 4.55% LGD 35% R2 25% Maturity 3y Principal Amortizing EAD $ 1,000,000 Drawn MTMt0 Spread (Lattice) Deal A 1% 95.6% Deal B 4% 99.9% Deal C 7% 104.2% EL Ecap - RC EL Ecap - TRC EL 1.34% 16.22% 17.42% 1.45% 17.13% 18.50% 1.54% 17.97% 19.49% RORAC RORAC - RC EL - TRC EL 7.33% 6.88% 7.33% 6.84% 7.32% 6.81% MtPar All 4.1% 100.0% 1.45% 16.17% 17.72% 7.73% 7.12% Since all deals have the same risk characteristics, they all get the same values in the risk metrics Since the return is assumed to be the one that compensates for the risk characteristics, the return is the same for all deals (4.1%) The Rorac numbers under MtPar settings are not really useful, because they do not reflect the actual return of the deal. Actual Rorac can be calculated outside of RiskFrontier. Given the same capital, the deals with higher return would have a higher Rorac Mark-to-Par spreads achieve neutralization of undesired MtM effects 13

14 Mark-to-Par Results: 2 nd Example Common Risk Inputs PD 0.39% LGD 35% R2 10% Maturity 3y Principal Amortizing EAD $ 1,000,000 Drawn Spread Deal A 1% Deal B 4% Deal C 7% MTMt0 (Lattice) 100.7% 105.2% 109.7% EL Ecap - RC EL Ecap - TRC EL.117% 1.62% 1.80%.124% 1.67% 1.87%.131% 1.73% 1.93% RORAC RORAC - RC EL - TRC EL 7.18% 6.53% 7.17% 6.50% 7.16% 6.47% MtPar All.54% 100.0%.120% 1.52% 1.69% 7.55% 6.86% Again, all deals show the same risk output, since all deals have the same risk characteristics Note: MtPar Ecap above is not directly comparable to that of the three deals, because entire portfolio was run with MtPar settings and the deal Ecap is allocated from the total portfolio Ecap Return is still crucial for risk-adjusted return analysis, i.e. how to choose the right investment. Capital per se does not need to reflect the return (e.g. Basel II formula). 14

15 Portfolio Ecap: MtM vs MtPar Increase in risk not fully compensated by increasing spreads Reduction in risk precedes reduction in spreads Credit Cycle MtM t0 declines MtM t0 increases Often Portfolio MtM t0 < Par Often Portfolio MtM t0 > Par Ecap MtM < Ecap MtPar Ecap MtM > Ecap MtPar t For an O&H Portfolio that is not marked-to-market, the difference between MtM and Par is not recognized as gain or loss. Therefore MtM may underestimate Ecap in bad times and overestimate it in good times. MtPar corrects this phenomenon. 15

16 Mark-to-Par: Conclusion Pros Corrects unintuitive treatment of return information in O&H portfolios that are not marked-to-market Deals with the same amount of money invested and same risk characteristics attract the same capital The relationship between risk drivers and capital is appropriate for all key risk drivers (e.g. PD, LGD, R 2, Maturity, etc.). This would not be the case if instead we used Defaultonly mode, because Maturity would play no role in risk assessment More correctly assesses the amount of capital for the portfolio, both in good and bad times Additionally, whether or not the portfolio is O&H, Mark-to-Par spreads are useful when fee information is not available Cons Hybrid approach, not conceptually pure. Provides a practical solution to a material issue. Further research is needed. Risk-adjusted Return ratios need to be calculated outside of RiskFrontier MtPar spreads are a useful way to get reasonable Ecap results for non marked-to-market O&H portfolios. More research is needed to refine the MtPar theoretical framework. 16

17 Contact information for Feedback For any question or feedback, please contact Stefano at 17

18 Appendix Mark-to-Par Spread Formula if no Optionality Source: Moody s Analytics document Overcoming Missing Fee Info Using RF 18