I have been receiving a lot of feedback on the Ambac article and the MBIA one as well. Many want more in terms of clarification, assumptions, additional calculations, data, etc. I just want to remind all that this is a blog on my commentary and thoughts on the markets and my investments. My primary occupation is investing, not blogging. I disseminate my research and opinions to provoke discussion and I love to blog on these topics, but I have limited bandwidth to return emails. I do not want the lack of answers to email questions to appear as if I am avoiding them, it is just that after a certain level of volume it distracts me from my day job. Please keep the emails coming, just be aware that I may not be able to answer all of them. That being said, I will present some additional data from my Ambac research, and am considering posting a what-if scenario of Ambac insuring E-trade (I am sure that will garner some interest). After that, we will be moving on to the commercial real estate, investment banking and consumer finance sectors where I am arranging additional bearish positions and will blog on them. I will, of course, enjoy and entertain discourse on these or related topics.
As for Ambac, my analysts have incorporated explicitly discernible calculations on subordination into the valuation model. However, like I’ve mentioned in the comments of the last Ambac post, the default probabilities that we had assigned earlier were after considering an implied 20% subordination into Ambac’s portfolio. In the latest version of the Ambac valuation model, we are explicitly showing the calculation of default rates before as well as after subordination. Please note that even in the riskiest form of collaterals (e.g. ABS CDO Mezzanine), the worst case default probability after adjusting for the given subordination level in our model is 70% compared to Bank of America’s assumption of 100% (which is plausible in some products). In the base case scenario, the average default probability we assumed (after subordination) was 25%. This again highlights that we have been conservative in our assumptions over default rates and potential losses that Ambac could incur. For an anecdotal example of the implications of foreclosure recovery rates in housing markets like CA and FL, see my last post. My findings are actually on par with that guy that writes the well followed Accrued Interest credit blog (very smart guy who came to almost the exact level of losses I did), Bill Ackman from Pershing Capital (prescient guy who got me started following these companies and also believes them to be insolvent soon), and UBS. So, not withstanding my flair for dramatic writing, I do have company in these loss estimates. Be aware that ABK has not opened up its books to any of us, or at least not to me - so everything is pretty much just best guesses based upon publicly available information. Well, now that I have finished handing out compliments...
You can download the 40 page sample Ambac valuation model here, which details the defaults before and after subordination for several categories, as well as providing for proforma financials and relative book valuation. Be aware that this is the third (and probably final) time I have increased the amount of documentation in support of the Ambac post, and we have gone from what I thought was originally a lengthy and well documented post to two lengthy posts, two downloaded pdf files, and about 70 pages of supporting documentation. My day job beckons... In summary:
Has subordination been taken Into consideration?
Yes, it has. To make it clear, it has been broken down so the calculations can be followed without the spreadsheet that made them.
8% Other RMBS
8% ABS CDO High Grade
8% ABS CDO Mezzanine
31% CDO Other
13% Other ABS
13% Average default probabilities (By Collateral) After Subordination
6% Other RMBS
6% ABS CDO High Grade
6% ABS CDO Mezzanine
25% CDO Other
10% Other ABS
Hopefully this will help in clarifying the doubts over subordination.
After running explicit contract by contract subordination calculations, the results remain the same as before (with the blanket subordination assumption) except the structured finance portfolio, where the potential losses have marginally come down to $3.9 billion (exactly same as that arrived at by Accrued Interest and UBS) after adjusting for contract wide subordination from the earlier levels of $4.2 billion. There has been no change in the potential losses in the Subprime RMBS and the Consumer Finance portfolio losses from the earlier levels. So, in essence, roughly a 3.75% difference in overall loss calculated after going through each contract with an explicit subordination calculation for each one. Thus, we stand by the original calculations, as clarified by this most recent one.
Default levels are not too aggressive and recovery rates are not understated
Higher recovery rates could’ve been possible if the credit crunch, real asset depression/recession and turmoil we’re currently witnessing was not as bad as it is today & trending downward. E*trade received anywhere between 11 cents to 27 cents on the dollar from the sale of its $3.1 billion portfolio of prime and investment grade asset-backed securities. 73% of E-trades 74 cent haircut MBS sale was backed by prime mortgages with an average 720 FICO score - more than 50% of that was AA or better (reports are that most of it was AAA). Some say it was a distressed sale and not reflective of normal economic activity. I say that it was normal economic activity for this environment and those assets. You cannot just ignore market transactions and paper them over with guesstimate opinions on price and models. That is what got us here in the first place in terms of structured finance. Lennar (the world's largest builder) sold a large chunk of their land and CIP inventory at about 50 cents on the dollar, after taking over a billion dollars in write downs on their entire inventory. Several big builders are going bankrupt over the next 8 quarters (a few have already), and many small builders have already crossed the line: of of which will dump many billion dollars of distressed properties on an already distressed housing market that is getting hundreds of thousands of homes added through foreclosure, driving prices down and supplies up even farther. Prices are dropping like dead flies across the country and homebuilders and banks pushing REOs are competing to drive prices down even further, reducing the collateral behind much of this structured (and unstructured) stuff (collateral which was significantly overstated by overly optimistic if not fraudulent appraising practices). The macro environment for these assets are getting worse, not better (see A note on mortgages, overly optimistic recovery rates and recent events...). Take a look at Centex's mortgage origination performance as far back as September, and imagine what it is now (this was BEFORE the mortgage crunch hit to prevent refinancing to recapitalize). The builders are (at least were) some of the largest non-bank mortgage underwriters in the country (little known secret) and they were quite aggressive in pushing loans to move inventory that normally would not have been easy to sell (severe understatement). As they sell off these tens of billions of dollars of loans to be included in CDOs and pools, they are poisoning these vehicles even further. These fluctuations in the macro environment is what the Boom, Bust, Bling Blog is about, and I think I know what I am talking about on this topic. In addition, factoring in what could be the losses compared to the actual default rates at this point of time would be cumbersome and this will involve making some unrealistic assumptions. If anyone want to take a stab at it for the community, feel free to have a go at it, send it to me and I will publish it & give you full credit. I may try incorporating more into the valuation after getting some clarity on the subprime rate freeze (which could decrease if not prevent the rate of foreclosures, although I definitely doubt so - )..
No assumptions were given and the post was too wordy
Yes, I actually received this one as a complaint. The original writeup on Ambac was fairly descriptive while explaining the assumptions. For example, Ambac’s consumer finance insured included Countrywide, GMAC, Indymac, Greenpoint Mortgage and Accredited Mortgage Loan amongst others who are in a financial mess and very close to bankruptcy. This is why I was considerably bearish on this group of insureds with sloppy underwriting standards. The Ambac piece was written as a post in my blog, and if you follow my blow my views on the macro environment, the real asset recession/depression and the financial markets have been made quite clear. The latest model has assumptions for quarterly and annual drivers, as well.
The duration of 5 years is inapplicable
The loss tail analysis was given a broad duration of 5 years since I instructed my analysts to consider the financial guaranty business a short tail casualty line, which it is, and many CDOs have a maturity of around 5 years. Pricing software allows selection of maturities from 1 month to a maximum of 5 years, but the average maturity is five years, such as in Australian CDO squareds, It is quite true that this can vary depending on the insured product, and yes, we can adjust the model to vary duration based on individual contract/product/tranche, but as I said earlier, this is a blog of my thoughts on the macro environment, the market and my own investments, and not a paid analysis. The majority of Ambac's losses are expected in structure products (CDOs) where a 5 year duration is most appropriate. For those who have no idea what this is about, here is a good primer that also gives the perspective of recent history. This is not the first time the CDO market hit bumps, the junk bond correction earlier in the decade tripped them up as well. A more complex essay is here (where of course they default maturity to 5 years). Back on topic - I truly believe that I have published an unprecedented amount of work on this topic (for a free blog) as it is. Can it be more accurate? All analyses can be more accurate! Does it convey a meaningfully accurate message? We think it does, and I have put my money behind it, as I have on practically every stock, security or real asset that I have blogged on. Thus far, my track record has been pretty good (knock on wood:-). In addition, I am confident the default rates used by the model are overly conservative. If anyone is truly interested in a more granular analysis, I may be willing to disseminate my own, more detailed proprietary research on a more formal basis.
The charge to capital is overstated due to exclusion of unearned premiums as claims paying capacity
Regarding reinsurance and charges to equity, we worked under the assumption that the company would be reinsuring some of the risks from its books (as mentioned by Ambac’s CFO in the recent conference organized by the Bank of America). Therefore, we did not consider unearned premiums as an additional claim paying capacity that the company will have since it is difficult to estimate how much will be reinsured and what will be the company’s earnings on ceding the premiums. However, we believe that considering the current negative sentiment over monoliners, reinsurers will get a very favorable deal on their part on assuming risk from Ambac’s books. Nevertheless, if one were to take unearned premiums into account, the charge against Ambac’s equity would be lesser by approximately $2.5 billion, which still leaves them in a bind in nearly all scenarios calculated.