Value at Risk
Overview
Value at risk is common risk model. It measures the amount of money
lost in a worst case scenario over a given time interval.
In order to use the model, one needs to
specify a threshold, which is specified as fraction or perecentage,
call it {% \alpha %}.
The VAR model would report the dollar loss in the scenario where only
1 - {% \alpha %} % of the scenarios are worse.
As an example, if {% \alpha %} is specified at the 99% level, VAR would report
the dollar loss of the scenario where only 1% of the scenarios are worse.
Illustration
Portfolio Distribution
The challenging part of constructing a VAR model is to obtain the portfolio
distribution. The method that is used to obtain the distribution
catagorizes the model in the following:
- Parametric models are models where the porfolio distribution is
assumed to be of a given mathematical form, with only a set of
parameters needed to specify it completely. The canonical example
is the assumption that the portfio value is normally distributed.
- Monte Carlo VAR models are a natural extension of parametric models.
The monte carlo method also assumes distributions for the assets in
the portfolio, which are typically fit by a set of parameters. However,
the modeler need not know how to sum these distributions to get the
portfolio distribution analytically. As long as one can simulate
the individual assets, a modeler can siulate the portfolio values
and then calculate the VAR.
- Historical Simulation
Scaling VAR
The definition of a VAR measure nees to specify a time frame. That is, it determines the possible amount of loss
at a certain confidence interval, over a given time frame. Often, it is necessary to specify a different time frame
from that for which the measure was calculated. If the resources are available, the measure could just be re-run at
the new time frame, however, there are assumptions that can be made that make scaling apossilbe without having
to re-run the measure.
As an example, for an asset (or portfolio) that follows i.i.d (independent and identically distributed) normal
returns, the variance is proportional to time. That is, if we know the variance of the portfolio over the time
span of 1 day, then the variance of the portfolio over a 10 day time frame is 10 times the 1 day variance.
(see
geometric brownian motion).
(see
Alexander pg. 21)
VAR Risk Types
Value at Risk is a measure that applies to a portfolio as a whole, regardless of the types of assets within
the portfolio. However, the methods used to calculate VAR are often dependent on the assets within the
portfolio. When a portfolio consists of multiple asset types, it is sometimes possible to compute the VAR separately for
each asset class and then to aggregate the results.
- Market VAR: is the
risk to a portfolio due to changes in the prices of the underlying assets. Measuring market var
is generally a similar exercise for geometric brownian motion assets
such as equities, foreign exchange, and commodities.
Market risk for a fixed income portfolio refers to the risk that arises from changes in market interest rates.
- Credit VAR: is the risk to a portfolio due to the risk
of the fixed income assets in the portfolio defaulting.