montecarlo - Portfolio simulation sensitivity analysis -
some context: doing project dissertation master's in mortgage bank.
description:
the bank wants me sensitivity analysis on portfolio of mortgage accounts. have simulation (pretty high - level) simulates portoflio @ end of each of next 4 years , outputs capital requirements , position against risk appetites. inputs of simulation lending plan (how % of each type of account , how ltv, redemption rate, probability of default , cure rate predict expected losses, hpi, , desired advances). want me provide them framework testing portfolio against different economic scenarios , lending plans, playing around inputs pretty much.
my approach:
in order that, need explore solution space in systematic way. looking doe methods derive solutions @ levels of each variable. however, have 10 inputs full factorial 3 levels example not feasible (the simulation runs once in 8''). designs should for? also, % of ltv must sum one, means inputs dependent on others. how deal that? once have sample of inputs , solutions, plan divide "good" , "bad" according results (read "good position against risk appetites") , perform smirnov tests see values drive performance, along correlation analysis between pairs of inputs.
any suggestions, comments, feedback , should answers on above appreciated!!!
Comments
Post a Comment