Session 55 PD, Pricing in a MCEV Environment Moderator: Kendrick D. Lombardo, FSA, MAAA Presenters: Christopher Kirk Brown, FSA, MAAA Seng Siang Goh, FSA, MAAA Kendrick D. Lombardo, FSA, MAAA
PRICING IN AN MCEV ENVIRONMENT: IMPLEMENTATION CHALLENGES May 17, 2016 Christopher Brown, FSA, MAAA PROPRIETARY AND CONFIDENTIAL
AGENDA Modeling and computational challenges Assumption setting challenges Interpretation of results Summary 2 PROPRIETARY AND CONFIDENTIAL
MODELING AND COMPUTATIONAL CHALLENGES Model Creation Models can range from very simple to very complex Items to consider: Line of business (Term, Whole Life, Universal Life) Model cell groupings Risk-neutral scenario generation Tradeoff between precision and runtime Model will only ever be as precise as the underlying assumptions Alignment with models used for reporting 3 PROPRIETARY AND CONFIDENTIAL
MODELING AND COMPUTATIONAL CHALLENGES Model Runtime and Data Processing Model runtime depends heavily on structure and complexity of model Number of economic scenarios For non-interest sensitive products (term), single scenario may be sufficient Interest sensitive products require thousands (or more) of scenarios Indexed UL requires equity scenarios as well Processing of output data Monthly vs. annual results Data size proportional to # scenarios x time intervals x output variables Could be many gigabytes of data Additional runs for sensitivities and CRNHR calculations Pricing is often an iterative process Parallel processing and cloud computing potential solutions 4 PROPRIETARY AND CONFIDENTIAL
MODELING AND COMPUTATIONAL CHALLENGES Model Runtime Example Indexed UL runtime example: 1,000 modeled variables 40-year monthly projection = 480 timesteps 5,000 model cells 12 segments per model cell 28.8 billion calculations 1000 economic scenarios: 28.8 trillion calculations Nested stochastic: quadrillions? Large number of scenarios may be needed to obtain accurate result Important to balance precision with runtime and business needs 5 PROPRIETARY AND CONFIDENTIAL
ASSUMPTION SETTING CHALLENGES Policyholder Behavior Must consider behavior assumptions in extreme economic scenarios Likely very little experience to use as guidance Term Insurance Typically not interest sensitive Key assumptions involve post-level term behavior Post-level term results more important at low discounting rates Universal Life Insurance Very interest sensitive Dynamic lapse modeling in high / low interest rate environments Premium payment patterns Indexed vs. traditional UL could have different behavior 6 PROPRIETARY AND CONFIDENTIAL
ASSUMPTION SETTING CHALLENGES Policyholder Behavior Example Indexed UL dynamic lapse and premium payment behavior Theory 1: customers will lapse if they have more attractive investment options elsewhere Theory 2: customers will stop paying premiums if account values are sufficient to carry the policy to maturity How to implement? Cap rates can t be directly compared to fixed interest investment options (traditional UL) Cash value growth in high return scenarios might lead to billions of dollars of account value for a single policy 7 PROPRIETARY AND CONFIDENTIAL
ASSUMPTION SETTING CHALLENGES Management Behavior Again, must consider behavior assumptions in extreme economic scenarios Term Insurance Typically not many decisions to make Whole Life Insurance Dividend or excess credit crediting strategies Universal Life Insurance Crediting rates for traditional UL Cap and participation rates for indexed UL Potential changes in expense charges? Interaction between management behavior and policyholder behavior Important (but difficult) to set realistic management behavior assumptions 8 PROPRIETARY AND CONFIDENTIAL
ASSUMPTION SETTING CHALLENGES Management Behavior Example How do you model the cap rates on an IUL product? Typical methodology would be based on investment earnings rate Normal scenario: 3% earned rate, 20% volatility, normal cap rate Alternate scenario: 0.1% earned rate, 40% volatility Methodology might generate a cap rate lower than the minimum guarantee Alternate scenario: 75% earned rate, 15% volatility Likely to generate infinite cap rate Scenarios could have prolonged period of extreme rates or jump up and down 9 PROPRIETARY AND CONFIDENTIAL
REAL WORLD VS. MCEV PRICING RESULTS Where do MCEV and real-world results differ? Spread-based products (especially UL) look worse on MCEV basis Example: Traditional UL declared crediting rate of 3% Required spread 50 bp Earned rate 1.5% MCEV vs. 3.5% real world MCEV pricing: 200 bp loss, real world: no gain/loss Products where profits emerge in late durations look better on MCEV basis Term product where much of profits emerge in post-lt period Recognition of actual investment earnings will occur as business matures 10 PROPRIETARY AND CONFIDENTIAL
INTERPRETATION OF RESULTS Pricing metrics New Business Value (NBV) New Business Margin (NBV as percent of premium) Internal Rate of Return (but note differences in interpretation) Return on Capital / Equity Others? No interpretation of the meaning of an individual scenario in a risk-neutral set But individual scenarios can be useful as a sanity check Sensitivity testing can identify weaknesses / strengths in pricing Again, model results are only as precise as the underlying assumptions 11 PROPRIETARY AND CONFIDENTIAL
SUMMARY Pricing models will always have some set of constraints (time, computing power, etc.) Set assumptions as realistically as possible Always be skeptical of your results Pricing is always a mix of art and science This is no different in an MCEV environment 12 PROPRIETARY AND CONFIDENTIAL
Pricing in an MCEV environment 2016 SOA Life & Annuity Symposium Seng Siang Goh, FSA, MAAA May 17, 2016
Agenda Motivation Real World Pricing Market Consistent Pricing Case Study Conclusion May 17, 2016 2
Motivation Framework to explicitly reflect all risks in pricing Non US regulation US subsidiaries Regulation Risk Reflection Economic View Consistency in risk management (FV hedging) Consistency in economic capital There needs to be a bridge (RW vs MC) May 17, 2016 3
MCVNB: Other Motivations Measuring internal performance Capital allocation across lines of business Assessing value of existing business Transparency for users Maybe sounds kind of cool
Real World Pricing Measures IRR Solved rate of return such that PVDE = 0. IRR vs hurdle rate Premium Margin PV of pre tax profits divided by PV of premiums at an assumed rate ROA PV of pre tax profits divided by PV of projected assets (AV) TEV Value of new business based on real world PVDE May 17, 2016 5
Real World Pricing Measures limitations No explicit requirement or framework to allow for all risks in pricing Equity risk premium and corporate spreads. Risk through discount rate is arbitrary and typically not product specific No explicit allowance for cost of options and guarantees, or cost of holding economic capital. 99.5 CI Use sensitivities Results may be counterintuitive and inconsistent, e.g. IRR May 17, 2016 6
MCVNB MCVNB Reflects market s view of the value of new business PV of future cash flows, adjusted for the market risks of these cash flows Cash flows are valued consistently with the prices of similarly traded cash flows in the capital markets Market consistent assumptions (calibrated risk neutral scenarios) Earned rate and discount rate both risk free May 17, 2016 7
MCVNB PVFP TVOG FC CNHR MCVNB MCVNB = PVFP TVOG FC CNHR Usually expressed as a margin (% of PV of premium) Desired MCVNB > 0 (value creation) May 17, 2016 8
MCVNB PVFP TVOG FC PV of future profits (post tax and pre CoC) Reflects the intrinsic value of the options and guarantees Calculated deterministically (Certainty Equivalent) Time value of options and guarantees (GMxB, ULSG) Calculated stochastically (PVFP) and take the arithmetic average TVOG = PVFP (CEQ) PVFP(Stochastic) Consideration given to dynamic p/h behaviour and management action Frictional cost of required capital Reflects taxation and expenses associated with assets backing the required capital CNHR Cost of residual non hedgeable risks (CoC approach) Captures both market risk (unhedged) and non market risk Risks not covered in PVFP and TVOG May 17, 2016 9
Case Study Report on Pricing Using MCEV VA product with GLWB and GMDB GLWB Higher of compound rollup of 5% and annual rachet ROP GMDB Simplification for MCVNB: No use of economic capital; CNHR: Simplified CoC approach under Solvency II. Time zero method instead of nested stochastic. Shock lapse and mortality and use 6% CoC May 17, 2016 10
Case Study Variable Annuity Real World Pricing: Based on a deterministic real world scenario (8% level equity) discount rate equals to the pre tax earned rate: Target 10% IRR Calculate real world TVOG, FC and TEV Risk Neutral Pricing: Determine PVFP using a deterministic certainty equivalent scenario Cash flow discounted using reference rate Project future distributable earnings using the stochastic risk neutral scenarios Calculate TVOG, CNHR and FC and MCVNB One scenario? May 17, 2016 11
Case Study Variable Annuity Real world versus Risk neutral deterministic sources of earnings Account values, and hence asset based cash flows, are lower under risk neutral Higher reserves under risk neutral and thus higher investment income, but in both cases, reserve has no impact since earned rate same as discount rate GMDB and WB claims higher (ITM) under risk neutral May 17, 2016 12
Case Study Variable Annuity Breakdown of stochastic components X No allowance for equity risk premium in MC pricing, adjust for the risk of investments directly in the investment assumption No allowance for CNHR in Real World pricing, may be possible through reflecting appropriate amount of economic capital Thus MC pricing typically lower than Real World pricing May 17, 2016 13
Challenges of MCVNB Fluctuation over time / Changing market conditions Time lag (product priced vs marketed) Product design tweak Senior management buy in Challenge to reconcile MCVNB results to other pricing measures Product can be negative MCVNB but still marketed Challenge to produce RN scenarios, CNHR May 17, 2016 14
Conclusion It s all about reflecting risk sufficiently in pricing Stochastic v. Deterministic Risk reflection Management Market MC pricing reflects market s view of risk; real world pricing can reflect management s view of risk as long as sufficient allowance is given to all risks. Theoretically could be the same.
MCVNB Useful References MCEV Principles Published by the European CFO Forum Solvency II Quantitative Impact Study Published by Committee of European Insurance and Occupational Pension Supervisors Practice Note on MCEV Published by the American Academy of Actuaries May 17, 2016 16