SERVING THE QUANTITATIVE FINANCE COMMUNITY

 
User avatar
rmax
Posts: 6080
Joined: December 8th, 2005, 9:31 am

Time Horizont for measuring market Risk

September 12th, 2014, 6:56 am

QuoteOriginally posted by: tulaQuoteOriginally posted by: DavidJN"how do you define antithetic scenarios ?"Say you have N observations of historical price changes.Extend the vector by multiplying each one of those changes by -1.While this is useful to reduce noise in a Monte Carlo context, I dont see why this would make sense with historical data. Wouldn't it just distort your dataset?It makes sense if you are looking at times of market stress. In essence you are generating a set of histoical bumps: i.e. credit spreads increasing hugely due to the credit crisis. However if you are looking at stress, then also looking at your portfolio valuation due to credit spreads collapsing also makes sense.
 
User avatar
tula
Posts: 46
Joined: December 1st, 2007, 10:35 pm

Time Horizont for measuring market Risk

September 12th, 2014, 6:13 pm

QuoteOriginally posted by: rmaxIt makes sense if you are looking at times of market stress. In essence you are generating a set of histoical bumps: i.e. credit spreads increasing hugely due to the credit crisis. However if you are looking at stress, then also looking at your portfolio valuation due to credit spreads collapsing also makes sense.My experience with credit markets is limited, but there seems to be an inherent asymmetry, ie credit spreads don't typically collapse with the same speed as they blow out. The same generally applies to other markets as well, to varying degrees. Why would you be interested in the sensitivity of your portfolio to scenarios that don't happen?
 
User avatar
DavidJN
Posts: 1765
Joined: July 14th, 2002, 3:00 am

Time Horizont for measuring market Risk

September 13th, 2014, 1:32 pm

"Why would you be interested in the sensitivity of your portfolio to scenarios that don't happen? "Because those scenarios could happen in the future, for starts.If you randomly generated scenarios, such realizations could also happen, no?
 
User avatar
tula
Posts: 46
Joined: December 1st, 2007, 10:35 pm

Time Horizont for measuring market Risk

September 13th, 2014, 8:17 pm

QuoteOriginally posted by: DavidJN"Why would you be interested in the sensitivity of your portfolio to scenarios that don't happen? "Because those scenarios could happen in the future, for starts.I'm talking about stress events, not regular day-to-day fluctuations. Do you really think the opposite of Sept-Oct 2008 is likely to happen? Or the opposite of Black Monday, with S&P rising +20% in one day?QuoteIf you randomly generated scenarios, such realizations could also happen, no?No, if the model/distribution you generate scenarios from is asymmetric (as I believe reality is).
 
User avatar
bearish
Posts: 5891
Joined: February 3rd, 2011, 2:19 pm

Time Horizont for measuring market Risk

September 13th, 2014, 9:54 pm

QuoteOriginally posted by: tulaQuoteOriginally posted by: DavidJN"Why would you be interested in the sensitivity of your portfolio to scenarios that don't happen? "Because those scenarios could happen in the future, for starts.I'm talking about stress events, not regular day-to-day fluctuations. Do you really think the opposite of Sept-Oct 2008 is likely to happen? Or the opposite of Black Monday, with S&P rising +20% in one day?QuoteIf you randomly generated scenarios, such realizations could also happen, no?No, if the model/distribution you generate scenarios from is asymmetric (as I believe reality is).You presumably use your beliefs to create your portfolio. Are you sure you also want to rely on them to make the risk management decisions?
 
User avatar
acastaldo
Posts: 1416
Joined: October 11th, 2002, 11:24 pm

Time Horizont for measuring market Risk

September 13th, 2014, 11:42 pm

QuoteDo you really think the opposite of Sept-Oct 2008 is likely to happen? Or the opposite of Black Monday, with S&P rising +20% in one day?I don't know.However, the S&P rose by 11.58% on October 13, 2008 and by 10.79% on October 28
 
User avatar
tula
Posts: 46
Joined: December 1st, 2007, 10:35 pm

Time Horizont for measuring market Risk

September 14th, 2014, 9:15 am

QuoteOriginally posted by: bearishYou presumably use your beliefs to create your portfolio. Are you sure you also want to rely on them to make the risk management decisions?I'd say it is best to rely on undistorted historical data to make risk management decisions, whatever that data is. Augmenting historical data with antithetic scenarios represents an implicit (and I believe unjustified) assumption that the data is symmetric. I'd rather not make such assumptions in general.There are types of financial data (such as certain FX rate changes) that exhibit a higher degree of up/down symmetry than, say, stock index returns or credit spread changes. In such cases using antithetic scenarios could be justified.Going back to historical data vs Monte Carlo: In the context of simulating a model, the symmetry assumption is often applied to latent, unobservable factors. I believe that assuming these to be symmetric is more easily justified (if nothing else, by Occam's razor). What we observe empirically though are nontrivial combinations of these factors, and there is no need for these to be symmetric in general.QuoteOriginally posted by: acastaldoHowever, the S&P rose by 11.58% on October 13, 2008 and by 10.79% on October 28Sure, and these points are all present in the data set without employing antithetic scenarios
 
User avatar
DavidJN
Posts: 1765
Joined: July 14th, 2002, 3:00 am

Time Horizont for measuring market Risk

September 14th, 2014, 1:34 pm

"Do you really think the opposite of Sept-Oct 2008 is likely to happen? Or the opposite of Black Monday, with S&P rising +20% in one day?"Let's turn the question back to you. Do you think Sept-Oct 2008 is likely to happen again? Or will Black Monday happen again? If not, why use those scenarios either?
 
User avatar
tula
Posts: 46
Joined: December 1st, 2007, 10:35 pm

Time Horizont for measuring market Risk

September 14th, 2014, 6:07 pm

QuoteOriginally posted by: DavidJNLet's turn the question back to you. Do you think Sept-Oct 2008 is likely to happen again? Or will Black Monday happen again? If not, why use those scenarios either?Fair question. I could say clichés like crises will happen again, but they won't be like past ones, yet we only have data about past ones, so we may at least want to see the sensitivity of our portfolio to history repeating itself, and maybe even make scenarios a bit more conservative. (That's what the Fed seems to be doing with CCAR.) While the benefit of this is questionable, think generals fighting the last war, in my view it's still much more justifiable than constructing stress scenarios by flipping signs of historical data. I have yet to see a crisis that starts with a sharp upward move in prices.
 
User avatar
DavidJN
Posts: 1765
Joined: July 14th, 2002, 3:00 am

Time Horizont for measuring market Risk

September 15th, 2014, 8:01 pm

"While the benefit of this is questionable, think generals fighting the last war, in my view it's still much more justifiable than constructing stress scenarios by flipping signs of historical data."Here's what a 1999 Australian regulator paper on comparative VaR methods by James Engel and Marianne Gizycki said about this:"Another problem with simple the historical-simulation approach is that asset prices often exhibit trending behaviour. One solution to the trend problem is to impose symmetry on the portfolio-value distribution by taking the negatives of the profits and losses used in standard historical-simulation and augmenting the original data with these results. This effectively doubles the data used in calculating the percentiles and eliminated the trend."They refer the reader to a 1998 paper by Holton.This view may have evolved over time.
 
User avatar
rmax
Posts: 6080
Joined: December 8th, 2005, 9:31 am

Time Horizont for measuring market Risk

September 16th, 2014, 7:10 am

QuoteI have yet to see a crisis that starts with a sharp upward move in prices.Why does it matter when a crisis starts and what happens then? Doesn't it matter that prices have moved a certain amount? If prices started going up, and you were short the asset that would be a crisis for the participant.
 
User avatar
Traden4Alpha
Posts: 23951
Joined: September 20th, 2002, 8:30 pm

Time Horizont for measuring market Risk

September 16th, 2014, 2:15 pm

Although the overall market may not have 20% up-moves, individual equities certainly can surge upward under the influence of M&A, major announcements, or short-squeezes. Also, if one adds FX effects to ADRs, then violent bidirectional moves are possible.That said, the doubling of data by appending an antitheitical copy makes a very strong assumption that all of the odd moments of the price movement distribution are exactly zero and it also increases the magnitudes of the even-numbered moments relative to the original dataset. That qualitative and quantitative structure may not be truely "representative" although is is arguable more stressful.
 
User avatar
daveangel
Posts: 17031
Joined: October 20th, 2003, 4:05 pm

Time Horizont for measuring market Risk

September 16th, 2014, 2:21 pm

Most people are usually quite relaxed about rising asset prices.
knowledge comes, wisdom lingers
ABOUT WILMOTT

PW by JB

Wilmott.com has been "Serving the Quantitative Finance Community" since 2001. Continued...


Twitter LinkedIn Instagram

JOBS BOARD

JOBS BOARD

Looking for a quant job, risk, algo trading,...? Browse jobs here...


GZIP: On