SOA 2012 Life & Annuity Symposium May 21-22, 2012 Session 28 PD, Using an Artificial Society (a Complexity Science Tool) to Project Life Insurance Sales Moderator: Benjamin Steward Wadsley, FSA, MAAA Presenters: Ben H Wolzenski, FSA, MAAA Walter H. Zultowski, Ph.D. Primary Competency Technical Skills & Analytical Problem Solving
MAIN EVENTS in the Life Insurance Marketplace Walter H. Zultowski, Ph.D. WZ Research + Consulting, LLC SOA Life and Annuity Symposium May 21, 2012 Business 101: Successful businesses are those that respond efficiently and effectively to changes in their external environment. 2 1
External Environmental Influences Demographics The Economy The Life Insurance Marketplace Technology Regulation More Predictable Less Predictable Demographics 1) Continued Aging of the Baby-Boomers Retirement Security Health Care and Long-Term-Care 2) Gen X and Gen Y Coming to the Forefront Theoretically, should be a boom for life insurance May still be a positive Delayed household formation; different buying behavior 4 2
Demographics (cont d) 3) Changing Family Structures Continued decline of husband-wife families Growth of non-traditional families of all types 4) Increasing Diversity Hispanics and women Immigration and population heterogeneity 5 Demographics (cont d) 5) Impacts on the Field Force Also Greying of the sales force; succession planning Difficulty attracting Gen X and Y to the business A force driving interest in non-agent distribution channels 6 3
Technology 1) Continued Pervasiveness of the Internet A tool for information vs. direct selling? High tech vs. high touch 2) Social Networking A tool for recruiting vs. direct selling? Democratization of information and opinion The need to get marketing and compliance on the same side of the table 7 Technology (cont d) 3) Hybrid Technologies Would seem to address the high tech/high touch challenge Not new: Is the agent of the future human? (Zultowski) Comp-phone-a-vision (Baranoff) An issue of critical mass of the technology itself 4) The Underwriting Revolution Simplified issue (agent vs. consumer driven?) Predictive modeling and analytics 8 4
Technology (contd ) 5) Medical Technology Longevity (has it peaked?) Genetic testing 9 The Economy 1) Hangover from the Past Decade (at the company level) Uncertainty Continued low interest rate environment Market volatility; cost of hedging 2) Hangover from the Past Decade (at the consumer level) Job instability, wage stagnation, loss of confidence, desire for guarantees Disappearance of government and employer-based safety nets Changing views on investments (a positive?) The New Retirement: Working 1 1 12 th Annual Transamerica Retirement Survey. Transamerica Center for Retirement Studies. May, 2011 10 5
The Economy (cont d) 3) Rationalization of the Business At the corporate level (e.g., exiting certain lines of business) At the product level Stagnant product lines (e.g., fixed annuities) Disappearance of product lines (e.g., financial guarantees, standalone LTC) Growth in hybrid products (e.g., life/ltc combos) 4) Results of Industry Globalization Internationalization of regulation (e.g., Solvency II) New market opportunities 11 The Technology (cont d) 5) The Geo-Political Environment Debt crises, defaults of economies, wars, political upheavals, new international powers, nuclear proliferation, price of oil, etc., etc., etc. 12 6
Regulation 1) Political Uncertainty Gridlock in Washington Undoing of actions taken 2) Budget Deficits New sources of revenue: taxation of the inside build-up 13 Regulation (cont d) 3) National Health Care Cost Impact on the health insurance business Potential source of new life agents? 4) Estate Tax Legislation 2010 tax law expires in 2012 If no action, returns to $1M exemption and 55% tax rate Unlikely, but stranger things have happened 14 7
Regulation (cont d) 5) Commission Disclosure European models New York State Regulation 194 15 860-471-3692 www.wzresearch.com 8
USING AN ARTIFICIAL SOCIETY (A COMPLEXITY SCIENCE TOOL) TO PROJECT LIFE INSURANCE SALES Presenter: Ben Wolzenski,FSA, MAAA Managing Member -Actuarial Innovations, LLC Primary Competency: Technical Skills & Analytical Problem Solving From Complexity Science to Life Insurance Sales Complexity Science Agent-based models Artificial societies Sugarscape Original and latest online versions Interpretative adaptation to model life insurance sales Practical considerations 1
Complexity Science Complexity Science An Introduction (and invitation) for actuaries, by Alan Mills, FSA, ND http://www.soa.org/research/research-projects/health/researchcomplexity-science.aspx Agent-based models Artificial societies Schelling segregation model Sugarscape Artificial Societies Also known as virtual worlds A form of agent-based computer model Agents Behavior rules Relationship rules Environment Attributes & change rules Rules for interaction with agents 2
Artificial Societies Modeler defines Agent behavior and relationship rules Environmental rules then lets the model run At successive time periods Agents may prosper, die, engage in specific activities Environment changes The Sugarscape model Growing Artificial Societies Social Science from the Bottom Up (1996), by Joshua M. Epstein & Robert Axtell Simplest version Environment: grid with two piles of sugar Agents distributed [randomly] 3
The Sugarscape model Environment Sugar grows Agents Eat sugar See nearby Move Accumulate wealth (sugar) Die 4
by Abraham Kannankeril http://sugarscape.sourceforge.net Applet to demo the program Documentation Link to download the source code Released under the GNU General Public License by Abraham Kannankeril Enhanced environment Variable size of grid (environment) Sugar and spice Pollution created and dispersed Summer and winter seasons Variable duration Variable fertility 5
by Abraham Kannankeril Enhanced agents (citizens) Random life expectancy Age groups Inherited and acquired characteristics Cultural group membership Risk taking/aversion by Abraham Kannankeril Enhanced agents, continued Barter goods Diseases and immunities Find mate(s) Have children Inheritance options 6
by Abraham Kannankeril Interpretative Adaptation of Interpretation of environment Environment (grid) = area containing citizens who may purchase life insurance Sugar + spice = financial resources Summer / winter = economic cycles Pollution = economic inefficiency or friction 7
Interpretative Adaptation of Interpretation of agents (citizens) Realistic age, life span Mate, procreate, accumulate wealth Diseases and immunities Characteristics and group membership Engage in financial transactions (barter) Leave inheritance Interpretative Adaptation of Newborn child initial wealth options #1 Newborn children are given half of the wealth that each parent had at his or her own birth. Since this applies from the outset, all new citizens start with wealth in a narrow range #2 Newborn children are given half of the wealth that each parent has at the new child s birth, which could be more or less. 8
500 450 400 350 300 250 200 150 100 50 0 Distribution of wealth under Option #1 vs. Option #2 0 1 2 3 4 5 6 7 8 9 10 Option #1 Option #2 9
Interpretative Adaptation of Number of financial transactions Influenced by environment Economic cycles Degree of economic inefficiency Influenced by population Absolute size Density Interpretative Adaptation of Some financial transactions are life insurance purchases A percentage of citizens, varying by cultural group, are life insurance agents 10
Interpretative Adaptation of The chance that any financial transaction is a life insurance purchase varies based on the citizen s: Age bracket 0-17,18-24,25-34,35-44,45-54,55-64,65+ Wealth bracket Cultural group membership and Interpretative Adaptation of The chance that any financial transaction is a life insurance purchase varies based on: Whether the client has children Whether the client has any diseases Whether the client already has life insurance Whether the other party to the transaction is a life insurance agent 11
Individual life insurance purchases in the U.S. per 100 residents 5.00 4.50 4.00 3.50 3.00 2.50 2.00 1.50 1.00 0.50 0.00 Policies per 100 population 2003 2004 2005 2006 2007 2008 2009 2010 Policies per 100 population Interpretative Adaptation of Baseline model: establish model parameters so that: The citizen population is relatively stable after the start-up period The average number of policies purchased is fairly stable over the extension periods 3.4 per 100 population in 1 st extension period 3.5 per 100 population in 2 nd extension period 12
Interpretative Adaptation of First variation: Increase unemployment by 25% Over the extension periods, how would you expect this to change: the average population, the average # of life insurance purchases per capita, and the variability of the # of life insurance purchases per capita? 13
Answer to audience question #3 3. The average population decreased 14% AND the average # of life insurance purchases per capita decreased 3% AND the variability of outcomes increased 19%. Interpretative Adaptation of Second variation: Increase the age at which citizens can have children from 18 to 25 How would you expect this to change: the average population, the average # of life insurance purchases per capita. 14
Answer to audience question #4 The average population was lower for a while but increased more steadily and was just over 1% higher over the entire period. The average # of life insurance purchases per capita increased 12%. 15
Interpretative Adaptation of Third variation: Increase productivity (of goods) by 50% Effect on average population: None (less than 0.1% change) Effect on per capita life insurance purchases: Increased by 8% Interpretative Adaptation of Fourth variation: Combine all 3 prior variations Increase unemployment by 25% Increase the age at which citizens can have children from 18 to 25 Increase productivity (of goods) by 50% 16
Interpretative Adaptation of Population & purchase effects combined outside the model: Average population (down 14%, up 1%, no change) = down 13% Average # life insurance purchases per citizen (down 3%, up 12%, up 8%) = up 17% Average # life insurance purchases per population (down 13%, up 17%) = up 2% Interpretative Adaptation of Population & purchase effects combined inside the model: Average population: down 14% Average # life insurance purchases per citizen: up 18% Average # life insurance purchases per population: up 2% 17
Interpretative Adaptation of Population & purchase effects combined inside the model: Also of interest The variability of the # of life insurance purchases per capita more than doubled an increase of 142%! Individual life ownership by Ethnicity - Guaranteed Uncertainty, LIMRA & SOA 80% 70% 60% 50% 40% 30% 20% 10% 0% % households with income >= $25,000 % households with income >= $25,000 18
Practical Considerations Today s results illustrative Artificial society applications Non-equilibrium economics, epidemic dynamics, healthcare decisions Requirements starting with Sugarscape Java programming Research regarding environmental and behavioral factors 19
From Complexity Science to Life Insurance Sales Questions for the audience What enhancements do you think would be needed to make this example a practical model? What other applications of artificial society models do you see? What are your questions? 20