Policyholder Behavior Southeastern Actuaries Conference

Transcription

1 Policyholder Behavior Southeastern Actuaries Conference Presented by Paul Fedchak - Milliman June 20, 2013 Agenda General Product Dynamics & Policyholder Behavior PB Assumption Considerations Example of PB Impact Resources for PB Data 2 1

2 Why Policyholder Behavior? Traditional insurance has three possible outcomes. 3 Why Policyholder Behavior? More and more products are sold with policyholder optionality in mind VA w/ Living Benefits ULSG IUL Required more and more for regulatory purposes C3 Phase II AG 43 VM-20 Impact on financials is more noticeable than ever 4 2

3 Life Product Dynamics What life product behaviors might be dynamic? Premium funding level Premium persistency Policy surrender Partial withdrawal Policy loans Dividend use Conversion STOLI/IOLI 5 Premium Funding Level Obviously only applies when flexibility exists Why is it important Possible approaches Defined premium levels/periods Last 12 months collected Percent of target Funding levels? If desired, how to define? Is resulting model expansion worth it? How else can it be dynamic? Response to product performance Keep guarantee inforce 6 3

4 Policy Surrender This refers to requested surrender, not value lapse Economic dynamic lapse formula likely similar mechanics to annuity Dynamic lapse formula for end of level term period Dynamics due to guarantees Experience studies Common formula for economic lapses 7 Multiple * (MR CR)^Exponent SC * SCfac MR = Market Rate CR = Credited Rate SC = Surrender Charge (%) SCfac = Multiple Annuity Product Dynamics Hedgeable Risks Market risk Non-Hedgeable Risks Longevity Policy surrender Partial withdrawal Annuity: Required minimum distributions Sub account or crediting method transfer Policy loans Guarantee utilization 8 4

5 GLWB Policyholder Behavior 100% 90% 80% Total Elected After 25 Years 70% 60% 33% 31% 50% 40% 30% 36% 20% Lapsed Died Active 10% 0% Lapses Before Elected Deaths Before Elected In-Force But Not Elected GLWB Elected Deaths After Elected Lapses After Elected 9 PB Assumption Consideration Setting dynamic assumptions the first time can be difficult Governance of static assumption setting/acceptance may be welldefined Static assumptions are often easier to back with experience & more easily understood Assumption credibility Use of more advanced statistical techniques Predictive Modeling 10 5

6 Introduction to Predictive Modeling Predictive Modeling (PM) Use of advanced statistical algorithms to predict a chosen target variable from a set of independent variables Often used in P&C pricing/underwriting, marketing, customer retention and fraud detection Use of PM by life insurers is limited but growing Underwriting Mortality assumptions Annuity assumptions Marketing and retention Source: Guillaume Briere-Giroux, Towers Watson 11 Advantages of Predictive Modeling More easily isolate Base lapses from Dynamic lapses Less data necessary to extract relationships Easier to allow for additional variables Correlations and interactions between variables are captured Source: Guillaume Briere-Giroux, Towers Watson 12 6

7 Impact of Premium Persistency Pricing Example Single ULSG Pricing Cell Shadow Account Design with Competitive Premiums Age 50 Male Standard Nonsmoker Level Lifetime Premium to Age 121 $500k Face Amount Best Estimate Lapse, Mortality, Expense, etc. Pricing Under Three Premium Persistency Assumptions 100% Premium Persistency 85% Premium Persistency (15% Annual Premium Suspension) 85% Stochastic Premium Persistency 13 Impact Pricing Example Premium Assumption IRR 100% Premium Persistency 10.5% 85% Premium Persistency 8.2% 85% Stochastic PP Minimum Scenario -6.0% 25 th Percentile 2.3% Median 7.6% 75 th Percentile 10.4% Max 10.7% 14 7

8 Impact Pricing Example Full Stochastic Pricing Distribution 12% IRR By Scenario - 85% Stochastic Premium Suspension 10% 8% 6% I R R 4% 2% 0% % -4% -6% -8% Scenario 15 Resources for Policyholder Behavior Data SOA Premium Persistency Study of FPUL Products U.S. Individual Life Persistency Update U.S. Individual Life Persistency Update Canadian) Lapse Experience under Term-to-100 Insurance Policies SoA/RGA Post-Level-Term Study Phase 1: Assumptions used for pricing and modeling Rozar and Rushing, Product Matters! 77, June 2010 SoA/RGA Post-Level-Term Study Phase 2: Assumptions used for pricing and modeling 16 8

9 More Resources for Policyholder Behavior Data Interactions Between Dynamic Lapses and Interest Rates in Stochastic Modeling Includes results of GMWB model Xue, Product Matters! 77, June 2010 LIMRA/SoA Deferred Annuity Persistency Study (2007) Predictive Modeling for Life Insurers Measuring PB in VA Contracts SOA/Towers Watson Research Paper August 2010, Guillaume Briere-Giroux, et al, 17 Links for PHB Resources Insurance/research-premium-persist-assumptions.aspx Life/Persistency/ ind-life-persistency.aspx

10 Recap More products than ever have dynamic behaviors Regulatory scrutiny is higher than ever Consider experience studies, credibility and predictive modeling When available company data is insufficient, industry resources are available 19 10