After ABS simulations I have found that for one class of the ABS issue we do not have enough credit enhancement. The security is externally rated and have been recently put on negative outlook, though it has been active only for a year. One of our analysts proposed to add excess spread to credit enhancement for a consumer loan ABS. I do not mind to add one year spread since I use an annualized loss ratio and st. deviation to calculate required subordination. Besides, excess spread is indeed first loss protection. However, that analyst wants to multiply minimum excess spread by the number of years to expected maturity, that is, by 5 years it will be 5*3%=15% to add as additional credit enhancement. The problem is that by adding the excess spread in such a way, almost all ABS securities will be triple A. The other “problem” is that if I add just one year spread, my rating for that security is going to be 3 notches lower than the external one. I think in current market it is not a big problem since rating agencies are usually slow, however, it may be difficult to convince management to downgrade that security given its higher external rating. At the same time, investors should be faster to get rid of bad assets, therefore they should downgrade earlier. IMO. The question is - how can available excess spread be added to credit enhancement? Is it a real problem to have lower rating than external for ABS? It is said that excess spread of 3% is guaranteed by the eligibility criteria to purchase new receivables, but I think it does not mean that it will be always 3% since if 10% of borrowers do not pay then they will contribute less to the excess spread. I have talked to Fitch’s analyst and he said they do not run ABS simulations at all, they just look at different performance metrics and scenarios. The only time when they run a simulation is when they rate it at inception.
Luchano Wrote: ------------------------------------------------------- > After ABS simulations I have found that for one > class of the ABS issue we do not have enough > credit enhancement. The security is externally > rated and have been recently put on negative > outlook, though it has been active only for a > year. One of our analysts proposed to add excess > spread to credit enhancement for a consumer loan > ABS. I do not mind to add one year spread since I > use an annualized loss ratio and st. deviation to > calculate required subordination. Besides, excess > spread is indeed first loss protection. However, > that analyst wants to multiply minimum excess > spread by the number of years to expected > maturity, that is, by 5 years it will be 5*3%=15% > to add as additional credit enhancement. The > problem is that by adding the excess spread in > such a way, almost all ABS securities will be > triple A. The other “problem” is that if I add > just one year spread, my rating for that security > is going to be 3 notches lower than the external > one. I think in current market it is not a big > problem since rating agencies are usually slow, > however, it may be difficult to convince > management to downgrade that security given its > higher external rating. At the same time, > investors should be faster to get rid of bad > assets, therefore they should downgrade earlier. > IMO. > > The question is - how can available excess spread > be added to credit enhancement? Is it a real > problem to have lower rating than external for > ABS? > > It is said that excess spread of 3% is guaranteed > by the eligibility criteria to purchase new > receivables, but I think it does not mean that it > will be always 3% since if 10% of borrowers do not > pay then they will contribute less to the excess > spread. > > I have talked to Fitch’s analyst and he said they > do not run ABS simulations at all, they just look > at different performance metrics and scenarios. > The only time when they run a simulation is when > they rate it at inception. What did he say about the inception run? They should reasonably run an analysis of the transaction upon a cash trap event and stress the excess spread. I am not sure how the mechanics work, but you are saying that repurchase of new assets will prop up the spread. Are new assets funded despite stressed defaults? How are the triggers working? If you are trapping cash it usually means that, at least on a revolving deal, you are into rapid am where no more funding/purchases occur. Usually, in that case, you just work on the assumption that you have a pool under distress and you model it out with a fixed pool, no other loans are purchased and no other spread is had by new loans. Your deal could be different, but at the same time, it sounds like you need to be a bit more conservative. I would assume that, even if you haven’t stopped purchasing, that you can’t get any more assets out of the company, since it’s already under duress and things probably have fallen apart. Normally, when we look at ABS conduit transactions, we give them some credit for spread. Usually you give them spread * WAL, then stress it by delinquencies and stressed losses. It wouldn’t be hard to build a cashflow model that includes delinquency/default assumptions, how they flow in the waterfall, and then perhaps stress it by the required rating stress (say 3x for a “A”). It also depends on your credit department. We always get a lot of pushback about spread, our credit guys say that, in the scenario you describe, that you cannot count on spread, especially since, if it’s a cost of funds deal, rates can be stressed pretty quickly. That’s why I usually counter with stressing pay rates, but cost of funds is too difficult to determine or measure, especially with the different rates that would occur, straight CP, LIBOR, default pricing, prime pricing…etc. It’s amazing how quickly some of these deals can unwind. It doesn’t take much for a security that seemed to have a lot of enhancement, whether from cash accounts, subordination, OC, or lots of spread, to go to nothing at all. When the hammer falls it doesn’t take much to have a busted deal.
Thanx, spierce! I am in credit risk. >>>What did he say about the inception run? Are new assets funded despite stressed defaults? How are the triggers working? The issue is ramping up, it is pretty quickly amortizing, therefore it has a 3 year revolving period when the originator can add new receivables to keep collateral stable. These receivables have to be of a good quality. There are some triggers that when breached stop all new purchases and the structure beging just amortizing: excess spread is less than 3%, ratcheting cumulative loss schedule etc. I upgraded one notch both classes for the structure. >>>Normally, when we look at ABS conduit transactions, we give them some credit for spread. Usually you give them spread * WAL, then stress it by delinquencies and stressed losses. It wouldn’t be hard to build a cashflow model that includes delinquency/default assumptions, how they flow in the waterfall, and then perhaps stress it by the required rating stress (say 3x for a “A”). Ok, that is interesting. That is what I have to input into the ABS model: 1. annualized average incremental quarterly loss % (from collateral at each quarter) 2. standard deviation (from quarterly % changes) 3. current collateral 4. number of borrowers 5. number of months to expected maturity Then I get required subordination that is greater than my credit enhancement: reserve fund, subordination and 1year excess spread. Our analyst (front office) suggest to take current excess spread and multiply by WAL. That is, 3%*5=15%. Then my available credit enhancement is surely larger than required. It reminds me a corporate finance analogy when one takes projected NIs over the next five years, add them to retained earnings and drive a rating up. Now what should I do - take an estimation of the net loss over the next 5 years and st. deviation and test whether that credtit enhancement will be breached? And again - what about the descrepancy between my rating and external - how can it be substantiated? Since someone may say that my simulation is unrealistic if agencies keep their ratings higher.
Besides, over one week, these classes lost 5% and 10% three weeks after the performance announcement. My point that the price deterioration is connected with the performance report publication is laughed at. However, the index and other securities did not drop that fast on that week. I have not found any news in Bloomberg to find the reason for such a drop. I suppose that since these mezzanine tranches are not big they are relatively illiquid, and may be there are just a couple of holders, one of whom had decided to get rid of the asset later and not right after the announcement. With other issue I have gotten a confirmation of my hypothesis since the price dropped right after the announcement. Our front office guys showed similar tranches from the same country, which were also performing badly, but some of them did not loose anything. In my opinion, it means that investors sell weaker securities and that the main reason for that is the credit quality of the pool.
Did not lose anything based on what basis? Your pricing service or what you’re actually getting bid?
The discussed security went down from 100.14 to 89.37 (>10%), comparable front office security went down from 91.12 to 91.75 (>.4%). Front office think it is not a credit issue… I disagree. This is the pricing service obviously. I presume the bid price is even less.
any thoughts? I am still struggling to undestand who one can multiply excess spread by WAL. If our excess spread is 5% and 5% of borrowers do not pay then we will have nothing left at all.
Luchano Wrote: ------------------------------------------------------- > any thoughts? > > I am still struggling to undestand who one can > multiply excess spread by WAL. > > If our excess spread is 5% and 5% of borrowers do > not pay then we will have nothing left at all. Those 5% of borrowers only pay 5% of 5%, not 100% of 5%.
Luchano Wrote: ------------------------------------------------------- > Thanx, spierce! I am in credit risk. > > >>>What did he say about the inception run? Are > new assets funded despite stressed defaults? How > are the triggers working? > > The issue is ramping up, it is pretty quickly > amortizing, therefore it has a 3 year revolving > period when the originator can add new receivables > to keep collateral stable. These receivables have > to be of a good quality. There are some triggers > that when breached stop all new purchases and the > structure beging just amortizing: excess spread is > less than 3%, ratcheting cumulative loss schedule > etc. I upgraded one notch both classes for the > structure. > > >>>Normally, when we look at ABS conduit > transactions, we give them some credit for spread. > Usually you give them spread * WAL, then stress it > by delinquencies and stressed losses. It wouldn’t > be hard to build a cashflow model that includes > delinquency/default assumptions, how they flow in > the waterfall, and then perhaps stress it by the > required rating stress (say 3x for a “A”). > > Ok, that is interesting. That is what I have to > input into the ABS model: > > 1. annualized average incremental quarterly loss % > (from collateral at each quarter) > 2. standard deviation (from quarterly % changes) > 3. current collateral > 4. number of borrowers > 5. number of months to expected maturity > > Then I get required subordination that is greater > than my credit enhancement: reserve fund, > subordination and 1year excess spread. > > Our analyst (front office) suggest to take current > excess spread and multiply by WAL. That is, > 3%*5=15%. Then my available credit enhancement is > surely larger than required. It reminds me a > corporate finance analogy when one takes projected > NIs over the next five years, add them to retained > earnings and drive a rating up. > > Now what should I do - take an estimation of the > net loss over the next 5 years and st. deviation > and test whether that credtit enhancement will be > breached? > > And again - what about the descrepancy between my > rating and external - how can it be substantiated? > Since someone may say that my simulation is > unrealistic if agencies keep their ratings higher. Ruh roh! Credit! I don’t think I can talk to you anymore, might be screwing over a fellow originator :). The revolving trigger is about what I assumed. From my experience, credit departments look at the worst-case scenario that could happen within the parameters of the deal upon termination of revolving. For example, if you have a concentration amount of 5% in a 450-550 FICO, 5% in 550-600 FICO, and unlimited upwards and historical numbers were 1%/1% in each bucket, they would “stress” the pool to assume that each of those concentration buckets would be maxed out. Just the same, when you have excess spread, you need to assume that some of that spread will be eliminated through defaults and delinquencies. It all depends on how bad your losses are and your default rates. The agencies don’t always give full credit for excess spread, sometimes they stress the def/delq numbers inside the cashflow model to come up with a “stressed” version of what excess spread can be expected. I’d do the same. If the deal is over-enhanced, it’s over enhanced. There’s nothing wrong with that, many deals go out that way. I would check the loss proxy to make sure you are using the ones the agencies used. The difference can be major. If you are using default curves that are too high, or if you are using gross vs net, or you aren’t taking into account that recoveries take a while to roll through the waterfall, newer vintages may be penalized for low recoveries. I am working on a lease transaction right now. According to their cume net loss curves (extrapolated to terminal losses using a timing curve) and a 3x stress for a “A” structure, their advance rate should be 89% max. However, given current performance, which deteriorated slightly and the corporate position, our max advance rate will be 85%. Once you include excess spread in there, it gets even better. The price will not decline just because of rate changes or even deteriorating trust performance. It’ll decline also because the liquidity has declined overall on the secondary market for many types of bonds. People will want you to “pay up” for them taking the bond. This is not only because money is tighter overall, but because nobody wants to step on a landmine. Thus, your risk spread is much wider than it was 3 or 4 months ago. There’s a lot more to bond pricing than just the rates. I don’t work on a sales desk or have a trading book, but I know, from talking to people and seeing how that stuff works, that it’s a lot more art than science. You need to have a good nose for the overall market, if you don’t, you lose.
Pricing services, historically, have no idea what they’re talking about…Go out and get bids for both pools and then see them both get priced @ $85.
skiloa - what is the current price for example for Ducato Consumer Class B and C? I do not have these prices in BB. spierce - thanx for the input. The main question now is whether I can reasonably multiply the available excess spread by WAL and whether it is indeed guaranteed. I just can’t accept adding spreads over the WAL since my risk measures are annualized, id est, not taken over the WAL period. It is like adding net profit for the next years to retained earnings, artificially incresing the rating. I would understand it, if I calculate 7-year loss and deviation and then add a 7-year cumulative excess spread to subordination, but I have only one year of loss data that I should use.
Luchano Wrote: ------------------------------------------------------- > skiloa - what is the current price for example for > Ducato Consumer Class B and C? I do not have these > prices in BB. > > spierce - thanx for the input. The main question > now is whether I can reasonably multiply the > available excess spread by WAL and whether it is > indeed guaranteed. I just can’t accept adding > spreads over the WAL since my risk measures are > annualized, id est, not taken over the WAL period. > It is like adding net profit for the next years to > retained earnings, artificially incresing the > rating. I would understand it, if I calculate > 7-year loss and deviation and then add a 7-year > cumulative excess spread to subordination, but I > have only one year of loss data that I should use. Why can’t you add in the excess spread over the WAL? It’s a contractual interest rate, is it not? You *will* get that spread from the obligors if it is. That is, unless it’s an unswapped LIBOR tranche, or your delinquencies and defaults increase so much that your spread significantly declines. How can you only have one year of loss data? Does the company not have any static pool loss curves of similar assets that can be customized to your current, or worst case pool? It seems like you are missing a lot of data. How can you be possibly rating this transaction without any data surrounding the performance of an analagous trance in the long-term? Even if you can’t find data from that specific servicer/seller you can find data of a similar company with similar assets and project it. If you can’t even find that, then ask the agency analyst what an approximate timing curve would be for that asset. Take your cumulative defaults up to that point and project them out for the remainder of the curve. Then use that as your basis for the default curve for the *whole* WAL term, which then will allow you to use the excess spread over the whole WAL term. To me, it looks like you’re trying to conform a rating to agency standards with only a fraction of the data available to you. Unless I am missing something, a key part of the story. That’s the wrong approach. In order to pass a transaction through my credit department I build the transaction from step 1 all of the way up to a working cashflow model.
Subordinated European Consumer Loans? Wow. I’d need a prospectus for the deal as there’s hardly any info available on bbg. I don’t know how it would trade better than say US home eq…Your B tranche may price @ 800dm and your C tranche @ 1500dm…that’s 67-16 and 50… I don’t understand how these would trade on top of home eq bonds or even 2nd liens…they’re not collateralized, it’s a euro deal (Ducato’s first) and there’s very little information, so it would trade like a 144a.
spierce Wrote: ------------------------------------------------------- > Why can’t you add in the excess spread over the > WAL? It’s a contractual interest rate, is it not? > You *will* get that spread from the obligors if > it is. That is, unless it’s an unswapped LIBOR > tranche, or your delinquencies and defaults > increase so much that your spread significantly > declines. > How can you only have one year of loss data? Does > the company not have any static pool loss curves > of similar assets that can be customized to your > current, or worst case pool? As you will see this is a wrong approach, my security performs much worse than the similar previous ones. Besides, the macroeconomic environment is different. According to procedures I use only actual loss data from this security and since it has been traded only for one year - I have one year loss data. If I extrapolate the performance, the results would be just devastating. > It seems like you are missing a lot of data. How > can you be possibly rating this transaction > without any data surrounding the performance of an > analagous trance in the long-term? Even if you > can’t find data from that specific servicer/seller > you can find data of a similar company with > similar assets and project it. If you can’t even > find that, then ask the agency analyst what an > approximate timing curve would be for that asset. > Take your cumulative defaults up to that point and > project them out for the remainder of the curve. > Then use that as your basis for the default curve > for the *whole* WAL term, which then will allow > you to use the excess spread over the whole WAL > term. Again this is irrelevant, it is like to compare new tech stock projected performance with… Apple… five years ago. Now the market is different, consumers do default and recoveries are not impressive. > To me, it looks like you’re trying to conform a > rating to agency standards with only a fraction of > the data available to you. Unless I am missing > something, a key part of the story. That’s the > wrong approach. > > In order to pass a transaction through my credit > department I build the transaction from step 1 all > of the way up to a working cashflow model. According to the procedures I take net incremental loss over the considered period, stdev etc and just run the model. I still do not have around 1% of subordination to support the rating. However the security is pretty young, could it be a mitigating factor? I spoke to Fitch’s analysts, they just do not run simulations. However, they produce dynamic performance ratios live GDR anf NDR, they take the last quarter result and then annualize, thus getting pretty concervative metrics if the security is underperforming.
skiloa Wrote: ------------------------------------------------------- > Subordinated European Consumer Loans? Wow. I’d > need a prospectus for the deal as there’s hardly > any info available on bbg. I don’t know how it > would trade better than say US home eq…Your B > tranche may price @ 800dm and your C tranche @ > 1500dm…that’s 67-16 and 50… > > I don’t understand how these would trade on top of > home eq bonds or even 2nd liens…they’re not > collateralized, it’s a euro deal (Ducato’s first) > and there’s very little information, so it would > trade like a 144a. Can you see the size of the tranches in BB? - they are pretty tiny. I doubt that the price could have dropped that much. I want to ask Fitch’s analysts why the tranches have officially lost 5-10% and not been downgraded. It usually happens with shares.
Luchano Wrote: ------------------------------------------------------- > spierce Wrote: > -------------------------------------------------- > ----- > > Why can’t you add in the excess spread over the > > WAL? It’s a contractual interest rate, is it > not? > > You *will* get that spread from the obligors > if > > it is. That is, unless it’s an unswapped LIBOR > > tranche, or your delinquencies and defaults > > increase so much that your spread significantly > > declines. > > > How can you only have one year of loss data? > Does > > the company not have any static pool loss > curves > > of similar assets that can be customized to > your > > current, or worst case pool? > > As you will see this is a wrong approach, my > security performs much worse than the similar > previous ones. Besides, the macroeconomic > environment is different. According to procedures > I use only actual loss data from this security and > since it has been traded only for one year - I > have one year loss data. If I extrapolate the > performance, the results would be just > devastating. > > > It seems like you are missing a lot of data. > How > > can you be possibly rating this transaction > > without any data surrounding the performance of > an > > analagous trance in the long-term? Even if you > > can’t find data from that specific > servicer/seller > > you can find data of a similar company with > > similar assets and project it. If you can’t > even > > find that, then ask the agency analyst what an > > approximate timing curve would be for that > asset. > > Take your cumulative defaults up to that point > and > > project them out for the remainder of the curve. > > > Then use that as your basis for the default > curve > > for the *whole* WAL term, which then will allow > > you to use the excess spread over the whole WAL > > term. > > Again this is irrelevant, it is like to compare > new tech stock projected performance with… > Apple… five years ago. Now the market is > different, consumers do default and recoveries are > not impressive. > > > To me, it looks like you’re trying to conform a > > rating to agency standards with only a fraction > of > > the data available to you. Unless I am missing > > something, a key part of the story. That’s the > > wrong approach. > > > > In order to pass a transaction through my > credit > > department I build the transaction from step 1 > all > > of the way up to a working cashflow model. > > According to the procedures I take net incremental > loss over the considered period, stdev etc and > just run the model. I still do not have around 1% > of subordination to support the rating. However > the security is pretty young, could it be a > mitigating factor? > > I spoke to Fitch’s analysts, they just do not run > simulations. However, they produce dynamic > performance ratios live GDR anf NDR, they take the > last quarter result and then annualize, thus > getting pretty concervative metrics if the > security is underperforming. Extrapolation is off of a timing curve. If you assume performance will be timed similar to historical performance, then it shouldn’t be any different, except for the amplitude of the losses. You seem to keep making comparisons to equities. You should know that while structured products do differ, they’re a bit more grounded in reality and data, as opposed to massive market swings and investor sentiment, at least as far as projecting losses and enhancement. If you want to project out you need a proxy, you need to find out the proxy that was used and increase that for current deviations. It won’t be perfect but it’ll be a decent assumption. Frankly, it really seems like you aren’t looking at the security correctly. Your whole point, as far as 5% delinquencies = 5% loss of spread, was a bit…odd. You can easily provide that what you said is false. What type of model is this? Who created it? Is this related to the bafin audit happening for German banks? Frankly, models not designed *by* structured credit people *for* structured credit people suck ass. They are full of gaps that attempt to be bridged by half-asset work, assumptions, and little knowledge regarding how each security is unique and the availability of data. As far as your quarterly loss, it’s going to introduce a massive amount of variance and will not come close to being able to correctly model the subordination. If I were you, I’d take a hard look at thinking about how to tackle your different environment on a cume loss basis. I could think of a few things to try and come up with a decent proxy. No offense, but this seems like the typical discussion I have with credit every time I close a deal. “no, your loss proxy is wrong!”, “No, your data is wrong!”, “No, the model says this, you say that, you are wrong!”. Ahhh, credit, how I love contentious relationships.
Luchano Wrote: ------------------------------------------------------- > skiloa Wrote: > -------------------------------------------------- > ----- > > Subordinated European Consumer Loans? Wow. > I’d > > need a prospectus for the deal as there’s > hardly > > any info available on bbg. I don’t know how it > > would trade better than say US home eq…Your B > > tranche may price @ 800dm and your C tranche @ > > 1500dm…that’s 67-16 and 50… > > > > I don’t understand how these would trade on top > of > > home eq bonds or even 2nd liens…they’re not > > collateralized, it’s a euro deal (Ducato’s > first) > > and there’s very little information, so it > would > > trade like a 144a. > > Can you see the size of the tranches in BB? - they > are pretty tiny. I doubt that the price could have > dropped that much. I want to ask Fitch’s analysts > why the tranches have officially lost 5-10% and > not been downgraded. It usually happens with > shares. Size of tranche has absolutely no bearing on price or price movement…If anything, a smaller tranche would contribute to greater price volatility. Heard of law of large numbers? The rating agencies don’t know squat…think about it: what motive do they have to downgrade deals? They get paid only to rate new deals…also, price has no bearing on the true value of the bond…