[R-SIG-Finance] Modeling sharp drops in volatility

Valapet Badri valapet.badri at gmail.com
Wed Aug 8 19:20:08 CEST 2012

	Your question falls between theory and practical use of GARCH and
its stated and unstated use(vol forecasting, Risk, pricing prediction) . A
practical and common sense  approach followed in the industry is to look at
volatility for past X periods (days) and extrapolate factoring in news
impact (keep models simple) , MACRO scenarios as viewed by the Trader. This
is used commonly in systematic and manual trading scenarios. This type of
simplicity may not sit well in theoretical world -:)

As always, after all the theory crammed in MBA, FE or PhD program we should
not forget at the end of day,  we are trying to model and  "FORECAST  some
random variable". Please remember your basic assumption from your
"Stochastic process classes" in deriving your pricing equation of
derivatives, while performing modeling in the real world.

1, Price of an equity is modeled as a "stochastic process" 
2, Its not moving $0 - $Infinity every a day but are range bound behavior,
with varying ranges.
3, Enter the concept regime switching models and behavioral economics 

You see the basic assumption in theory falling apart already -:) Guess I
digressed quite a bit ..

My 2 cents worth...!


-----Original Message-----
From: r-sig-finance-bounces at r-project.org
[mailto:r-sig-finance-bounces at r-project.org] On Behalf Of stoyan.stoyanov
Sent: Monday, August 06, 2012 10:11 AM
To: r-sig-finance at r-project.org
Subject: [R-SIG-Finance] Modeling sharp drops in volatility

Hi all,

This is less of a programming question and more of a theoretical issue, but
hopefully someone can help. How would you suggest going about modeling a
sharp drop in volatility using a univariate GARCH model? What I am trying to
capture is the overstatement of volatility forecasts that occurs after
earnings announcements. As you surely know, the memory of GARCH models
incorporates recent (and not so recent) swings in volatility into the
variance of the stock price process. As a result, in the periods following
an earnings event volatility forecasts will be severely overstated.
I have been able to predict future increases in volatility owed to an
earnings event, through a binary event day dummy. However, the effects of
such single big moves persist in my process much longer than they should.
Instead of having my GARCH model take an isolated big move as significant
information affecting volatility in future periods, I am trying to model a
very sharp reversal back to levels observed before the earnings event.
I guess this is to be expected with GARCH models, but I am wondering if
there is a common technique to deal with this.

I am using the rugarch package to do build my model.

Thanks in advance for any suggestions.


Stoyan Stoyanov
The University of Chicago Booth School of Business MBA Class of 2013
(312) 532-0120 | stoyanov at chicagobooth.edu
View this message in context:
Sent from the Rmetrics mailing list archive at Nabble.com.

R-SIG-Finance at r-project.org mailing list
-- Subscriber-posting only. If you want to post, subscribe first.
-- Also note that this is not the r-help list where general R questions
should go.

More information about the R-SIG-Finance mailing list