BOLTON WED 330PM VEN IV.pdf

download BOLTON WED 330PM VEN IV.pdf

of 26

Transcript of BOLTON WED 330PM VEN IV.pdf

  • 8/12/2019 BOLTON WED 330PM VEN IV.pdf

    1/26

    US Earthquake Model:Its in the Details

    RAA Cat Modeling 2014

    Februar 12 2014

    Product Manager, Global Earthquake Products

  • 8/12/2019 BOLTON WED 330PM VEN IV.pdf

    2/26

    Introduction

    Its in the detailsHazard

    Financial Loss

    Aggregation Output

  • 8/12/2019 BOLTON WED 330PM VEN IV.pdf

    3/26

    First ste : Define stochastic event set

    Hazard Modeling

    How Big?

    Where?

  • 8/12/2019 BOLTON WED 330PM VEN IV.pdf

    4/26

    Gutenber Richter Recurrence Relationshi

    Defining Stochastic Event Set

    BIAS

    Too High

    oo ow

  • 8/12/2019 BOLTON WED 330PM VEN IV.pdf

    5/26

    Understanding Bias:Probability of Occurrence

    . n years

    is1-in-30,000 years

    In a 300,000 yearsimulation is sam led10 times

    In 10 000 ear simulation?

    If included = model too high!

    Not included = model too low!

  • 8/12/2019 BOLTON WED 330PM VEN IV.pdf

    6/26

    Why is Sampling Important?

    https://www.facebook.com/pages/Geomorphology-Rules/

  • 8/12/2019 BOLTON WED 330PM VEN IV.pdf

    7/26

    Minimizing Bias:California Example

    http://upload.wikimedia.org/wikipedia/commons/c/cf/Flat_eq_map_anotated.png

  • 8/12/2019 BOLTON WED 330PM VEN IV.pdf

    8/26

    Quantifying Bias:Variability of 10,000 Years

    Catastrophe modeling uncertaintyis significant at decision points.

    Increasin event set size reduces

    modeling uncertainty

  • 8/12/2019 BOLTON WED 330PM VEN IV.pdf

    9/26

    Stability of 300,000 Yearsof Simulation

    Additional samples in a 300,000year simulation minimizes bias

    and lessens uncertaint

  • 8/12/2019 BOLTON WED 330PM VEN IV.pdf

    10/26

    Stability of 300,000 Yearsof Simulation

    beyond decision points

    Stability for decision

    making return periods

  • 8/12/2019 BOLTON WED 330PM VEN IV.pdf

    11/26

    Capturing AllLevels of Hazard

    .(states of AR, IL, IN, KY,MS, MO, TN)

    USGS

    Earthquake Shaking Perilfire not included

    Coverages includeBuilding, Contents, andTime Element

  • 8/12/2019 BOLTON WED 330PM VEN IV.pdf

    12/26

    Historicall

    Financial Loss Modeling

    StructureContents

    BuildingA

    Mean &Standard Deviation

    Structure

    ContentsTime Element

    u ngB

    Mean &Standard Deviation

    StructureContentsBuilding Mean &

    Standard Deviation

  • 8/12/2019 BOLTON WED 330PM VEN IV.pdf

    13/26

    Financial Loss Modeling the old way

    ? ??

    UNCERTAINTY x UNCERTAINTY!

  • 8/12/2019 BOLTON WED 330PM VEN IV.pdf

    14/26

    Financial Loss Modeling the old way

    Convolution??

  • 8/12/2019 BOLTON WED 330PM VEN IV.pdf

    15/26

    Example: Modeling Sub-Perils

    Earthquake policies often encompasses several sub-perils:

    the new way

    EQSH - Earthquake ground shaking EQFF Earthquake fire following

    Limits/Deductibles can be different by sub-peril.

    Results are determined for each sub-peril and bycombination. Care is taken not to double/triple countexposures & losses

    Ground up values dont burn rubble

    Each calculation is Incremental

    Sub-perils are correlated

  • 8/12/2019 BOLTON WED 330PM VEN IV.pdf

    16/26

    Calculating Sub-Perils the new way

    The calculation begins at the insured coverage level.amage s a e re never excee s

  • 8/12/2019 BOLTON WED 330PM VEN IV.pdf

    17/26

    Allocate Deductible

    deductible

    deductible

    Blanket deductible = 40

  • 8/12/2019 BOLTON WED 330PM VEN IV.pdf

    18/26

    Apply EQ Shake Sub-Limit

    shake sub-limitdeductible

    deductible

    Earthquake ground shaking sub-limit = 20

  • 8/12/2019 BOLTON WED 330PM VEN IV.pdf

    19/26

    Apply Blanket Limit

    shake sub-limitdeductible

    deductible

    Blanket limit = 45

  • 8/12/2019 BOLTON WED 330PM VEN IV.pdf

    20/26

    he Event Loss Table ELT re resented the 1st

    Catastrophe Model Output

    generation of model output format The Year Loss Table (YLT) provides a more consistent

    an co eren me o o aggrega e oss

  • 8/12/2019 BOLTON WED 330PM VEN IV.pdf

    21/26

    Portfolio Analysis:California Earthquake

    Simple portfolio of~80,000 risks inSouthern California

  • 8/12/2019 BOLTON WED 330PM VEN IV.pdf

    22/26

    Transparency in Model Output

    ModelingTransparency: EventRecurrencemodels

    YLT Statistics: California EQ

    51,587

    5,411 370 18 2 0

    250,000

    300,000

    rsEQECATs YLT produces high-

    150,000

    200,000

    fsim

    ulationyea reso u on ca as rop e r s s a s cs

    Outputs are direct simulations therecurrence model can be directly

    242,612

    100,000Numbero

    evaluated by reviewing model results

    0

    50,000

    Zero One Two Three Four Five Six

    Num ero Eart qua eswit ossinayear

  • 8/12/2019 BOLTON WED 330PM VEN IV.pdf

    23/26

    Example of YLT Statistics:

    $250

    llions

    California EQ

    $200

    Mi

    The YLT can be used

    to create theexceedance probability

    $150

    Loss,$

    OEP_Value

    -from the same data!

    $50

    $100 _

    $0

    1 10 100 1,000 10,000 100,000 1,000,000

    eturn er o , ears

  • 8/12/2019 BOLTON WED 330PM VEN IV.pdf

    24/26

    $250

    llions

    Decomposition of the TVAR curveTVAR250FaultDriversofRisk

    $200

    Mi

    18%

    SantaMonica

    S.SanAndreas

    SierraMadre

    PuenteHills

    ElysianPark

    PalosVerdes

    ElsinoreFault

    n ng a po n s n

    the risk curves to thestochastic event libraryenables more insi hts

    $150

    Loss,$

    OEP_Value

    14%

    12%8%

    7%

    6%

    AnacapaDume

    NewportInglewood

    Northridge

    Verdugo

    Hollywood

    Raymond

    OakRidge

    into risk

    $50

    $100 _

    11% SanGabrielSantaSusana

    5%

    TVAR250EQMagnitudeDriversofRisk

    $0

    1 10 100 1,000 10,000 100,000 1,000,000

    23%

    5.5

  • 8/12/2019 BOLTON WED 330PM VEN IV.pdf

    25/26

    E ECAT rovides leadin -ed e solutions to hel measure

    he Future of Cat Modeling

    your risk

    Eliminating inherent biases

    e er con ro an ransparency

    Measurable solutions

    ,

    if you cant measure it, cant manage it

  • 8/12/2019 BOLTON WED 330PM VEN IV.pdf

    26/26

    Its in the details

    rProduct Manager, Global Earthquake [email protected]

    www.eqecat.com/software/rqe