Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Seminar
Robert Crampton
Series 1
02/27/2015
Agenda
What is VaR
How is VaR computed on PORT<go>
Different VaR Methodologies used on
PORT<go>
Analyzing VaR on PORT<go>
Settings
VaR Comparison
VaR Terminology
Factor Breakdown
Monte Carlo/Historical VaR Simulation
What is VaR
VaR measures the maximum dollar loss projected
given inputs for the time horizon and confidence level.
VaR can be measured on the portfolio, benchmark, or
active/difference portfolio.
Based on a distribution of returns, whether historical
or derived via a simulation, we can find the specified
dollar value ranges.
PORT calculates VaR based on three different percentile
ranges:
97.5
95%
99%
%
VaR Methodologies on
PORT
Five types of VaR are provided on the PORT<go>
system in Bloomberg, each model utilizes the
Bloomberg Multi-Factor Risk Model in order to
keep VaR consistent with Tracking Error data
You can choose which Factor Model to use while
calculating VaR through your PORT settings
Parametric
VaR
Historical 1y
VaR
Historical 2y
VaR
Historical 3y
VaR
Monte Carlo
VaR
Parametric VaR
Assumes factor returns and non-factor returns are
normally distributed
We first model each factor, and assume each of these has
returns are normally distributed.
We next look at our exposures to each of these factors.
Each factor will have different exposures calculations.
HINT!: To see the exposures to each of the factors, we can
look in the Tracking Error tab on PORT. The factors sub tab is
going to show us our portfolios exposure to the said factor
model.
Parametric VaR
Benefits /
Drawbacks with
Simplest of all three
Parametric
VaR
of the
methodologies.
Underestimates VaR
since returns are
usually
fat tailed
Used for linear
securities
since we assume
normal
distribution we are
fully
dependent on linear
factor model
Historical VaR
Historical VaR builds on Parametric VaR, as it does not
assume a normal distribution for all of the factor returns.
We build a distribution of factor returns based on the
historical performance of those factors.
Given historical factor returns and current portfolio
exposures, (again, seen on the Tracking Error tab) we can
model the corresponding returns for all securities in a
portfolio and aggregate these returns on the portfolio level.
Given the aggregated portfolio returns we create a historical
distribution of said returns and compute the desired
percentile of the distribution for VaR
PMs currently have the choice to use 1-3 years of factor
return history on PORT<go> for historical VaR.
Thus, we allow for flexibility on how to construct the
distribution of historical factor returns.
Full Valuation
Used for securities with highly non-linear pricing
Use actual realized prices for each simulation
This is computationally demanding and cannot be realistically
implemented for a multi-asset risk system that updates daily. To expedite
the computation while faithfully representing the risk profiles of nonlinear
instruments, many methods have been developed such as stress matrix
pricing below.
SMP
The basic idea of stress matrix pricing (SMP) is to compute full valuation
on a low dimensional grid. The scenario P&L is then approximated by
interpolating on the grid during simulation.
VaR Methodology
VaR-Main View
VaR-Main View
Customization
VaR-Settings
Settings
effect your
Distribution
and Factor
Breakdown
sub tabs
Remember
that your VaR
value is
effected by
the risk
model you
have loaded
VaR-Main View
VaR %
Compone
nt VaR
VaR-Main View
VaR %
Contributio
n
VaR Ratio
VaR Comparison
VaR Comparison
Marginal
VaR
Partial
VaR
Condition
al VaR
VaR Distribution
VaR Simulations
BEFORE I
Decrease my
Exposure
AFTER I
Decrease my
Exposure
Custom Reporting
ActionsCreate/Edit TemplatesVaR
Choose PDF or Excel
Choose the Subtabs you are interested in
Create charts based on Main View subtab
Custom Reporting
Custom Reporting
THANK YOU!