Why GARCH and Stochastic Models Serve Different Risk Tasks

Summary

The answer distinguishes statistical risk analysis from no-arbitrage derivative pricing. For portfolio risk measures such as value at risk or expected shortfall, analysts generally model outcomes under the real-world probability measure, informed in some way by historical observations. Derivative valuation instead commonly uses a risk-neutral measure derived from prices of traded instruments so that valuations fit the market and avoid arbitrage.

The response uses this distinction to explain why GARCH is common in risk forecasting while stochastic calculus is often associated with derivatives pricing. It notes that GARCH forecasts volatility from past data, whereas local and stochastic volatility models are often framed around market calibration. These are broad tendencies, not strict boundaries: the document does not claim that stochastic methods cannot be used for risk analysis, nor does it give a credit-risk model or empirical comparison. Measure choice and calibration target depend on the question being answered.

Key ideas

• Portfolio risk analysis typically estimates behavior under the real-world probability measure.
• Derivative pricing commonly uses a risk-neutral measure calibrated to traded market instruments.
• GARCH is presented as a statistical approach for forecasting volatility from historical data.
• Local and stochastic volatility models are commonly associated with derivatives pricing in this explanation.
• The distinction reflects different objectives and calibration sources, not an absolute separation between methods.

Tags

# Why is GARCH more often applied in risk analysis than stochastics?


# Why is GARCH more often applied in risk analysis than stochastics?












I am trying to look out for something I can engage in for my final year project (M.Sc) but my interests lie more in risk analysis (specifically credit risk). I have tried searching the web but really failed to get a good answer to my question.

Question: Can risk analysis be done using stochastic calculus? If yes, why is it that most work on risk analysis is done using GARCH models? (I dont hate GARCH but I am more interested in stochastics)

## Answer by Richi Wa (score 8, accepted)

https://quant.stackexchange.com/a/21324

Stochastics are usually applied in the field of derivatives pricing. In this setting the task is to price a derivative such that it fits into the landscape of tradable instruments (no-arbitrage). We work using the risk-neutral measure - usually denoted by $Q$. The measure is derived from other traded instruments.

In risk analysis (e.g. calculate the VaR, ES of this portfolio of stocks or credits) we work in the real world measure $P$. Usually $P$ is in some sense derived from history. This approach is rather statistical.

So the answer is: because this are 2 connected but somehow quite different fields of (quantitative/financial) mathematics.

One important EDIT: The GARCH approach tries to forecast volatility. This is doen by local volatility and stochastic volatility in the world of stochastic analysis/derivatives pricing. Still: GARCH is rather a P-thing (and it is calibrated on the past) and the others are Q-things (calibrated in the market).

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.