how do you guys model it? we are currently using standard default copula approach with abs tables whereever they are available. we are using fitch correlation assumptions.

reason for me asking is that it seems that many traditional ABS people are approaching the cdo market, yet do not seem to apply cdo technology. at leat that was my impression over the last couple of months.

Actually, I found it difficult to obtain reasonable prices using copula based models from the corp synth tranches on tranches of ABS backed cash-flow CDOs. I wonder if the binary prepayment-of-recovery events of those models are appropriate to mimic the dynamics of an ABS backed CDO. In those trades prepayments may, depending on the waterfall reduce the outstanding amount of tranches, excess interest on the asset side may be used to repay principal on the liability side.We played around with a lot of applications out there. If you start modelling cash-flows on the third-level underlying collateral, using intex or applications that link to intex, you would derive cash-flows based on default and recovery assumptions driven by macro-oeconomic parameters (housing-turnover, interest-rates etc.). Then you'd push those cash-flows trough a waterfall and could derive a valuation for your tranche. It would not make sense to simulate defaults again on the second-level collateral, since you already accounted for this by assuming a default rate for the credit-card or mortgage pools underneith the ABS collateral that gets referenced by your CDO-tranche.I'd be very interested to learn if anybody figured out how to skip the step of looking at cash-flows on third-level and just employing somespread based copula model that yields reasonable prices or can describe how they deal with ABS CDOs.

Working with a one factor copula model, some tailor made synthetic CDO positions are valued well using CDS quotes as input for default information if the reference portfolio exclusively contains corporate bonds. Whenever ABS enter the portfolio, it's far less obvious how to extract default information on these positions from readily available data sources such as Bloomberg. Anybody has any suggestions ?Some practitioners, and academics encountered when attending some recent seminars, called it "more of an art than a science", and some suggested to "consider the ABS positions default-free", but that didn't getme much further so far ....Thanks.

we've seen a couple of synth. CDO^2 trades that reference a set of (AAA, AA) ABS and several inner single-tranche CDOs that are defined ona pool of 200-300 corp names. Those trades price fairly well using CDS-spreads on the corporates and the (very low) rating implied default prob on the ABS. This is because the ABS is not where the risk sits in those structures, their just packaged in to divert the attention of the unwary investor.Not sure how to go about low-rated ABS collateral. The market of CDS-on-ABS seems to develop, that may help in future.Corgan, the ABS in your CDO, can they prepay? How does this effect your tranche/structure?

we modelled a cdo square containing several synths inner cdos and a big ABS AAA bucket. we ignored the ABS since its PD-distribution was pretty much neglectable compared to what happened to the inner cdos when we looked at different corr levels. frightening. sp's 30/0 made the whole thing look like AAAA ... i have a bloomberg, but no intex. quite a limitation. i consider to treat all underlying abs in the same way and assume exponential increase in prepayments after year 4. i hope to net out some of the error by the number of relative similar abs. since i do not have the data i cannot truly "model" prepayments ... what is your opinion on the legal aspect of the abs? pay as you go versus bullit and this kind of issue ...what do you think about servicers? i consider to take 10% of each underlying abs (second level) and model that seperately giving each such virtual asset the rating of the servicer and correlate these servicers by 50%. thoughts? stupid idea?peace