How to use GARCH(1,1) model?

[luzhu]

Does any one have any ideas on how to use GARCH(1,1) to forecast volatility and hedge risk?
For example, we are given 2 years' stock daily prices, how can we use this model?
I just want to gain a deeper understanding on how this model is used in real projects.

 

[Andy Nguyen]

GARCH is popular in academic because it only needs price data. Many people don't use it in real life.
You can take a look here for a similar discussion
http://quant.stackexchange.com/ques...-for-assessing-the-performance-time-series-fo
http://quant.stackexchange.com/ques...t-volatility-if-residuals-are-often-correlate

 

[zyz]

Andy, what method is mostly used to estimate volatility in risk measurement (VaR, ES, coherent etc)?

 

[luzhu]

GARCH is popular in academic because it only needs price data. Many people don't use it in real life.
You can take a look here for a similar discussion
http://quant.stackexchange.com/ques...-for-assessing-the-performance-time-series-fo
http://quant.stackexchange.com/ques...t-volatility-if-residuals-are-often-correlate

Thanks Andy! That helps a lot. It is also a cool website! Nice day

 

[Andy Nguyen]

Andy, what method is mostly used to estimate volatility in risk measurement (VaR, ES, coherent etc)?

I don't work in risk so I can't tell you from first hand experience. Hopefully, experienced practitioners like @Ken Abbott can give us some guidance on this matter.

 

[Ken Abbott]

I've seen GARCH used to estimate implied vol where there there is no data.

 

[Tsotne]

I've always been saying I didn't like Jorion's VAR but it would be a good read for you. There is 9th chapter devoted to GARCH models. I'm sure you won't get the explanatory details completely since they are not perfect, but you will get the answer to your question. If needed, notify me and I'll send you that specific chapter (non-copyright) to take a look.

@luzhu

 

[luzhu]

I've seen GARCH used to estimate implied vol where there there is no data.

How to do the estimation without data then?

 

[Ken Abbott]

You do the GARCH on the underlying.

 

[Ken Abbott]

• Take the stock, do first differences.
• Feed the differences into the GARCH software of your choice.
• Observe the alpha, beta 1 and beta 2.
• Estimate the unconditional variance of the time series differences. (This is the sigma squared nought.)
• Starting with sigma squared nought, use the coefficients just estimated to generate the time series.

I have not seen GARCH used as a forecasting tool for trading ideas. The original Wall Street papers form Salomon Brothers in the 1990's proposed this, but I think the GARCH estimates are too noisy.