loss aggregate part3

Upload: blimundo

Post on 25-Feb-2018

219 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/25/2019 Loss Aggregate Part3

    1/26

    1

    Modeling Property Treaties withSignifcant Cat Exposure

    Model non-cat & cat LRs separately

    !on Cat LRs ft to a lognor"al cur#e

    Cat LR distri$ution produced $y co""ercialcatastrophe "odel

    Co"$ine %con#olute the non-cat & cat loss ratiodistri$utions

  • 7/25/2019 Loss Aggregate Part3

    2/26

    2

    Con#oluting !on-cat & Cat LRs- Exa"ple

    0% 30% 60% 100%

    LR Prob 60% 20% 15% 5%

    40% 10% 6.0% 2.0% 1.5% 0.5%

    55% 25% 15.0% 5.0% 3.8% 1.3%65% 35% 21.0% 7.0% 5.3% 1.8%

    77% 25% 15.0% 5.0% 3.8% 1.3%

    100% 5% 3.0% 1.0% 0.8% 0.3%

    These probabilities 40% 70% 100% 140%

    correspond to 55% 85% 115% 155%

    these total LR's 65% 5% 125% 165%

    77% 107% 137% 177%

    100% 130% 160% 200%

    Total Loss Ratios

    !isreti"ed #at LR's

    $on cat

  • 7/25/2019 Loss Aggregate Part3

    3/26

    3

    Truncated Loss Ratio'istri$utions

    Pro$le"( To reasona$ly "odel the possi$ility o)high LR re*uires a high lognor"al C+

    ,igh lognor"al C+ o)ten leads to unrealisticallyhigh pro$a$ilities o) low LRs which o#erstatescost o) PC

    Solution( 'ont allow LR to go $elow selected"ini"u" e.g.. /0 pro$a$ility o) LR12/0

    3d4ust the "ean loss ratio used to calculatethe lognor"al para"eters to cause the

    aggregate distri$ution to pro$a$ility weight$ac5 to initial expected LR

  • 7/25/2019 Loss Aggregate Part3

    4/26

    4

    Su""ary o) Loss Ratio'istri$ution Method

    3d#antage(

    Easier and *uic5er than separately "odeling)re*uency and se#erity

    Reasona$le )or "ost pro-rata treaties

    6sually inappropriate )or excess o) loss contracts 'oes not re7ect the hit or "iss nature o) "any

    excess o) loss contracts

    6nderstates pro$a$ility o) 8ero loss

    May understate the potential o) losses "uch

    greater than the expected loss

  • 7/25/2019 Loss Aggregate Part3

    5/26

    5

    Excess o) Loss Contracts( SeparateModeling o) 9re*uency and Se#erity

    6sed "ainly )or "odeling excess o) loss contracts Most aggregate distri$ution approaches assu"e

    that )re*uency and se#erity are independent 'i:erent 3pproaches

    Si"ulation %9ocus o) this presentation !u"erical Methods ,ec5"an Meyers ; 9ast calculating approxi"ation to

    aggregate distri$ution Pan4er Method ;

    Select discrete nu"$er o) possi$le se#erities %i.e. create