Sep 13, 2011

Market Risk Management Tools (I) - VaR

Value-at-Risk (VaR)
VaR is a measure of potential loss on a portfolio which is expected not to be exceeded given a holding period and a defined confidence interval (C.I). It is used by banks for internal risk management as well as for capital and regulatory disclosure purposes. For regulatory reporting and capital computation, 10d VaR at 99% C.I. is adopted.


Common VaR methodology
1. Parametric VaR - assumes normal distribution of historical returns, thus fails to capture fat-tail events. Hence, it's not very suitable for use if  portfolio has a lot of complex/non-linear risks. Relatively easy to compute.
2. Historical VaR - Most commonly adopted by banks. It makes no assumption about historical distribution, thus fat-tail events are captured. Revaluation of current position based on historical changes in the market risk factors. May be a problem if there's not enough historical data.
3. Monte-Carlo  - most robust of the 3 but computationally intensive. It introduces a random sampling distribution technique and can be used when there's not enough data for the historical period used. Hence, this is commonly used for credit spread VaR.


Pros of VaR:
1. Provides an overview of total risk across all portfolios and asset classes as it allows for aggregation and netting.
Limitations of VaR:
1. Losses beyond the confidence interval are not captured by a VaR calculation. Essentially VaR only captures potential loss under normal market conditions and leave out the extreme but plausible events.
2. Limited by relevance of historical data. But the past may not be a good predictor about future events, especially when there are significant changes in market conditions. Eg. Market structure changes, spikes in volality, correlation changes, liquidity dry-up in stressful events.
3. Holding period (usually 1 day) does not capture market risk of positions which cannot be closed out or hedged within the holding period defined in VaR.
4. Intraday risks not captured.


Below table summarises the VaR methodology used by major banks in Singapore:



*Expected Shortfall is the average of all one day hypothetical losses beyond the 95% confidence level.
3W is the average of the three largest one day estimated losses.

No comments:

Post a Comment