Name of Presentation
Total Page:16
File Type:pdf, Size:1020Kb
Scenarios and Stress Testing For risk management, regulation, models and fun Philipp Keller Scenarios Scenarios: Possible, internally consistent events, parameterized by a set of risk factors t=3 t=2 Forecast t=1 t=0 Scenarios {E│Capital(E)<0} Reverse Scenarios Stress Synthetic Scenarios Scenarios Historical Sensitivities Scenarios {E│P(E)<α} Scenarios Multi- Implicit Event Events Single- Scenarios {E│Capital(E)<0} Global Event Events Scenarios Scenarios Single Event X ~ LN(x,α,β) Company- Specific Set of Scenarios Events Uses of Scenarios Crisis Extreme Management Events Recovery and Assessing Resolution Unquantifiable Planning Events Impact of Impact of Contingency Regulation and Uncertain Planning Financial Events Policies Assessing Scenario- ORSA Financial based Planning Soundness Systemic Strategic Emerging Exposure Analysis Risks Analysis Incentivizing Communication Setting Capital Materiality Risk Culture on Risks Requirements Assessment Defining Risk Reverse Stress Model Appetite Testing Assessment Simulation / Scenario Training based Models Model Assessing Enhancement Concentration Risks Scenario Analysis The importance of valuation • The quantification of the impact of a scenario depends strongly on the valuation standard being used. • Illustrative Example: Solvency Ratios pre- and post stress test of a financial risk event for Solvency I, QIS 5 (using liquidity premia, matching adjustments etc.) and market-consistent Solvency Ratios 300% The relative declines in 250% solvency ratios as well as the relative solvency ratios 200% between Solvency I, QIS 5 and 150% market consistent valuation 100% under a realistic financial 50% market event. 0% Solvency I QIS 5 Market Consistent pre-stress post-stress Solvency I ratios barely QIS 5 solvency ratios react In a market consistent react to the stress test more but dampeners mitigate framework, the impact of effects of stress events the scenario is strongest Scenario Analysis Regulatory uses: The IMF Source: IMF Country Report No. 13/124, May 2013: Belgium Financial System Stability Assessment Scenario Analysis Regulatory uses: Swiss Solvency Test 2007 Life - Impact of Scenarios on RBC Stock market fall (2001/2002) Spreads widening - LTCM (1998) High spread risk, which materialized US interest rate crisis (1994) in 2008+ European currency crisis (1992) Nikkei crash (1990) Very high risk to falling interest rates, which Stock market crash (1987) is materializing since the 1990s 10 year Yield Real estate -50% and i.r. +1% 8 Equity fall -60% 6 4 Drastic drop of interest rates 2 -1 -0.5 0 0.5 0 Impact of Scenario / RBC -2 1988 1988 01 1989 07 1991 01 1992 07 1994 01 1995 07 1997 01 1998 07 2000 01 2001 07 2003 01 2004 07 2006 01 2007 07 2009 01 2010 07 2012 01 2013 07 2015 01 Scenarios for Modelling Historical tracks Hurricane North Atlantic Track sampling Historical data set with approximately 1’000 storm events in the last 100 years is enriched by • track sampling: 9 “daughter tracks” from one “mother track” Conceivable tracks • pressure sampling: 9 additional pressure samples • uncertainty modelling: each event loss sampled Pressure sampling five times Generates 500’000 potential events Value distribution Vulnerability Cover conditions Generation of The exposure of a portfolio towards an event is the event loss possible events Where situated? How well built? How covered? Uncertainty modelling • Sums insured Example • Hurricane Cover limits “Charley” Aug 2004 • Deductibles • Exclusions • etc. Source: Swiss Re Scenarios and Catastrophes Compliance Culture Legal Uncertainty Financial Crash Bad Governance Financial Repression Rogue States Limited Capital Mobility Nationalisation Dictatorships Nuclear War Fire Autocratic Rules Corruption Industrial Accident Resource Depletion Demographic Change Dam Collapse Airplane Crash Rapture Marine Catastrophe Simulation Hypothesis Global Causality Violations Doomsday Equation Alien Attack Tornado Earthquake Climate Change Volcano Collapse of Infrastructure Tsunami Artificial Biology Biological Hazards Limnic Eruption AI Uprising Radiological Contamination Stone Fall Artificial Moral Hazard Nanotechnology Solar Flare Black Hole Collision The Big Rip Gammy Ray Burst Asteroid Impact Supernova Entropic Death Brane Collision Vacuum Metastability Event Scenarios: Pandemics Megadeaths by microorganisms The 1918 flu pandemic infected 500 Mio people and killed between 50 - 100 Mio of the infected. It caused a cytokine storm (overreaction of the immune system) which led to younger adults between ages 20 to 40 to die disproportionally while children and older people with weaker immune systems were less affected. Globally approx. 30% of all people were infected, and a total of 3%-6% of the world population died. Deaths per 100’000 in each age group for the World War Z, Paramount Pictures US population for 1911-1917 and for 1918 Massively higher mortality for people aged 15 to 44 A scenario has to be specified properly for it to be useful. Assuming a multiplicative mortality increase might lead to an impact that is positive (due to annuitants dying) while in reality the impact for most life insurers would be negative. Monty Python and the Holy Grail, Python (Monty) Pictures Scenarios: Attack by Aliens Relativistic death Mars Attacks!, Warner Bros Pictures Mars Attacks!, Warner Bros Pictures A 1kg mass impacting at 99% Our galaxy consists of ~100bn stars and the number speed of light would release an of planets is likely an order of magnitude higher, Apparent position from Earth energy of ~132 megatons or twice the larges hydrogen many of which must have the condition for life. Light speed However no signs of intelligent life have been bomb ever exploded. The detected yet (the Fermi paradox). One solution to the detection time would be very Relativistic speed limited if the RKV travels close Fermi paradox is that self-replicating von Neumann to the speed of light. Even if machines destroy all life in order to inhibit RKV detected, it would be futile to competition. Alien civilizations then either are careful Actual position destroy it since even vapour or and do not emit signals or are being destroyed, e.g. small particles would be highly by relativistic impactors (Relativistic Kill Vehicles) or destructive. by other means of planetary and solar destruction. Scenarios: Autonomous Cars Moral Dilemmas Self-driving networked cars will pose unique challenges to insurers. The programming has to cope with situations where the losses will have to be minimized and which require potential moral judgment. In addition, the type and number of accidents will change radically, requiring changes in the business model of insurers. Scenarios help to think about possible consequences, risks and opportunities. B A A ? A Scenario 1: Car A has to choose Scenario 2: Car A will crash Scenario 3: Ruritania requires between crashing into different into child unless car B rams producers of self-driving cars to pedestrians and has make a value into A, killing driver B assign lower values to certain judgment minorities Scenarios: Tambora Eruption Source: Wikimedia Commons Source: Wikimedia Commons Source: Wikimedia Commons Tambora Eruption 1815 1818 return to normal Shift in thermohaline circulation climate situation 1816 Summer Temperature Anomaly 100 km^3 of debris ejected into Snow in June 1816 in North America Shift in jet stream atmosphere (~10 times Vesuvius) Disruption of the Indian monsoon Global cooling, failed crops in Europe and North America and Yunnan, China Immediate local devastation, estimated deaths 70’000+ Floods and low temperatures in Europe Lung infections due to Typhus epidemics in Ireland 1816 – 1819 (40’000+ victims) sulfur acid and ashes Most livestock died during winter 1816/17 in the US Cholera pandemic in India , China and Indonesia: 1817to 1824 Doubling of mortality in Switzerland Catastrophic collapse of ice dam in Switzerland in 1818 (Gietro Glacier) Thank you www.actuaries.org Moving the profession forward internationally .