standardized method for market capital charge
- Thread starter ajsa
- Start date
Hi asja,
Re double counting, you may be right about that, I cannot recall that specific criticism against standardized market; in theory, if an instrument transomes multiple risk types (e.g., equity and currency) and those are distinct risks, it doesn't prima facie sound like double-counting.
...but *definitely* a criticim of building block is the simple adding which implicity assumes perfect correlations (+1.0), so that's overcharging but i wouldn't call it double counting per se (more like failing to credit risk diversification benefits ... hmm....maybe that is a sort of double-counting?)
re specific risk, yes, as i mentioned in tutorial. Here is XLS I used in the tuorial example:
http://sheet.zoho.com/public/btzoho/buildingblockcharge
long $100 in liquid stock XYZ
short $60 in same
so you can see, the XLS very simply has:
8% generic of the NET position, plus
4% specific of each GROSS position (i.e., yes to absolute value here)
David
Re double counting, you may be right about that, I cannot recall that specific criticism against standardized market; in theory, if an instrument transomes multiple risk types (e.g., equity and currency) and those are distinct risks, it doesn't prima facie sound like double-counting.
...but *definitely* a criticim of building block is the simple adding which implicity assumes perfect correlations (+1.0), so that's overcharging but i wouldn't call it double counting per se (more like failing to credit risk diversification benefits ... hmm....maybe that is a sort of double-counting?)
re specific risk, yes, as i mentioned in tutorial. Here is XLS I used in the tuorial example:
http://sheet.zoho.com/public/btzoho/buildingblockcharge
long $100 in liquid stock XYZ
short $60 in same
so you can see, the XLS very simply has:
8% generic of the NET position, plus
4% specific of each GROSS position (i.e., yes to absolute value here)
David
Hi ajsa, yes, if they need the specific charge in IMA is the *same* as per standardized; but they can also "demonstrate" their VaR model incorporates specific so they can qualify out...in case you are going to ask if the stressed VaR, by adding, double-counts, then yes, that was my first impression when I saw it and many firms have criticized the stressed VaR double-count. The stressed VaR add is still totally weird to me. In my view that is the most flagrant doulble count...David
Hi asja, correct, that's under's Crouhy's RAROC ... which is about economic capital not regulatory captial, so this is one version of an *internal* approach to sizing the market risk component of EC ... I would venture: low testability ... David
frm_daniel
Member
Hi David,
I refer to the screencast page. 6 and page.13 for operational risk 7a. How will the market risk capital charge fall on the graph of x-axis as shown in page 6?
Regards,
Daniel
I refer to the screencast page. 6 and page.13 for operational risk 7a. How will the market risk capital charge fall on the graph of x-axis as shown in page 6?
Regards,
Daniel
Hi Daniel,
That's a good question, I'd like to improve on that graph to incorporate a "unifiying" credit/market/operational perspective. Currently, that graph refers to a credit risk distribution but it can be viewed from a market risk perspective if you think of EL as the credit risk version of drift (negative for credit, positive for market risk) that informs the expected future value.
So although it's not meant for market risk, as the point of EC/RAROC is a common yardstick, it could proxy for market risk if you replace EL with expected return (drift) on the market for the market risk component (portfolio). In that case, you tend to see different terminology, like: UL = FV - P(c) = Future expected portfolio value (i.e., the future mean) - wost portfolio value @ confidence level (i.e., the VaR quantile). So they are essentially similar:
1. Credit risk (p 6, displayed): capital charge for UL = VaR (confidence) - EL; is analogous to:
2. Market risk: capital charge for UL = VaR (confidence; i.e., worst expected future portfolio value) - (portfolio drift)
... so, typically, with a daily VaR the drift is assume ~ 0, so you can replace the drift (EL in the chart) with 0.
... the "complicating" difference is that, if you want to include drift, the difference between market risk and credit risk is that EL is a loss and drift is a gain (positive expected return). So, the chart 6 (b/c it has losses to the right) would replace EL with drift that is to the left (!) of the current portfolio value, such that UL = VAR() - (-drift+ = VaR() + drift (i.e., what Jorion call's the relative VaR).
It's a long way of saying: if you take the chart, shift the EL to the left (to reflect a positive expected market return rather than a negative expected credit loss), then the market risk charge would similarly cover the UL but now UL would be drift + VaR().
Hope that helps, David
That's a good question, I'd like to improve on that graph to incorporate a "unifiying" credit/market/operational perspective. Currently, that graph refers to a credit risk distribution but it can be viewed from a market risk perspective if you think of EL as the credit risk version of drift (negative for credit, positive for market risk) that informs the expected future value.
So although it's not meant for market risk, as the point of EC/RAROC is a common yardstick, it could proxy for market risk if you replace EL with expected return (drift) on the market for the market risk component (portfolio). In that case, you tend to see different terminology, like: UL = FV - P(c) = Future expected portfolio value (i.e., the future mean) - wost portfolio value @ confidence level (i.e., the VaR quantile). So they are essentially similar:
1. Credit risk (p 6, displayed): capital charge for UL = VaR (confidence) - EL; is analogous to:
2. Market risk: capital charge for UL = VaR (confidence; i.e., worst expected future portfolio value) - (portfolio drift)
... so, typically, with a daily VaR the drift is assume ~ 0, so you can replace the drift (EL in the chart) with 0.
... the "complicating" difference is that, if you want to include drift, the difference between market risk and credit risk is that EL is a loss and drift is a gain (positive expected return). So, the chart 6 (b/c it has losses to the right) would replace EL with drift that is to the left (!) of the current portfolio value, such that UL = VAR() - (-drift+ = VaR() + drift (i.e., what Jorion call's the relative VaR).
It's a long way of saying: if you take the chart, shift the EL to the left (to reflect a positive expected market return rather than a negative expected credit loss), then the market risk charge would similarly cover the UL but now UL would be drift + VaR().
Hope that helps, David
The power of search..! For a niggling question**
I just found this answer above about the specific risk charge wrt how it is actually calculated. What I understand (imagine this as a flow/process diagram - in fact, Basel in a decision tree format would be fairly awesome):
A. Standard approach - just add up all the specific and general risks to get the charge. Grossly penalizes: no diversification, etc.
B. IMA - you calculate the general market charge with your VaR model (the VaR + stressed VaR). Then either:
(B1) you don't get approval to model specific risks, and so you must add-on the SA specific risk charges from A1, or;
(B2) you get permission to model specific risk, and then have to model it somehow. This permission comes with the added effort of the IRC calculation. So you are no longer grossly penalized by the SA specific risk calc, but they take some of the joy away by making you do IRC.
Can someone verify my understanding ? Basically, wrt B2 above, if you can show that your VaR model already covers specific risks, then effectively you don't need to do extra work in B2, except the IRC calc (which is not trivial).
Maybe someone will find it in three years, like I did!
And it will be semi-useful (and not superceded by Basel 4!)
Thanks!
**I had a hunch all along that if the VaR model is robust enough to capture specific risk, it would just get subsumed into the VaR+stressedVaR part using IMA (and not using the standardised approach). I couldn't find the wording in the Basel docs (or have the time to pick through it) to verify it, especially among the voluminous chapters of admendments and red-cross outs. A tracked-changes MS Word file would have been better
I just found this answer above about the specific risk charge wrt how it is actually calculated. What I understand (imagine this as a flow/process diagram - in fact, Basel in a decision tree format would be fairly awesome):
A. Standard approach - just add up all the specific and general risks to get the charge. Grossly penalizes: no diversification, etc.
B. IMA - you calculate the general market charge with your VaR model (the VaR + stressed VaR). Then either:
(B1) you don't get approval to model specific risks, and so you must add-on the SA specific risk charges from A1, or;
(B2) you get permission to model specific risk, and then have to model it somehow. This permission comes with the added effort of the IRC calculation. So you are no longer grossly penalized by the SA specific risk calc, but they take some of the joy away by making you do IRC.
Can someone verify my understanding ? Basically, wrt B2 above, if you can show that your VaR model already covers specific risks, then effectively you don't need to do extra work in B2, except the IRC calc (which is not trivial).
Maybe someone will find it in three years, like I did!
And it will be semi-useful (and not superceded by Basel 4!)Thanks!
**I had a hunch all along that if the VaR model is robust enough to capture specific risk, it would just get subsumed into the VaR+stressedVaR part using IMA (and not using the standardised approach). I couldn't find the wording in the Basel docs (or have the time to pick through it) to verify it, especially among the voluminous chapters of admendments and red-cross outs. A tracked-changes MS Word file would have been better
Similar threads
