Ambac Financial Group Inc (AMBC) 2007 Q2 法說會逐字稿

Greetings, ladies and gentlemen, and welcome to the Ambac Financial Group Inc. second-quarter earnings conference call.

At this time, all participants are in a listen-only mode. A brief question-and-answer session will follow the formal presentation. (OPERATOR INSTRUCTIONS). As a reminder, this conference is being recorded.

It is now my pleasure to introduce your host, Mr. Sean Leonard, Chief Financial Officer of Ambac Financial Group, Inc. Thank you. Mr. Leonard, you may begin.

Thank you. Good morning, everyone, and welcome to Ambac's second-quarter conference call. I'm Sean Leonard, Chief Financial Officer of Ambac. With me today are Robert Eismann, Controller, Peter Poillon, Investor Relations, and Tom Gandolfo, Senior Manager Director responsible for structured credit as well as the investment portfolio and our financial services activities. At the end of my second-quarter earnings discussion, Tom will give a brief update on our CDO portfolio and he and I will both be available to answer questions after the prepared remarks.

Our earnings press release, quarterly operating supplement and a short slide presentation that summarizes the quarter's results are available on our Web site. Also note that this call is being broadcast on the Internet. Third-quarterly 2007 earnings will be released on October 24, 2007 at 6 AM with a conference call at 11 AM.

During this conference call, we may make statements that would be regarded as forward-looking statements. These statements are based on management's current expectations. I refer you to our press release for factors that could change actual results.

Highlights of the second quarter are as follows. Net income (technical difficulty) $173 million or $1.67 per diluted share. That's down 25% on a per-diluted-share basis from the second quarter of 2006. The current quarter was impacted by a negative mark-to-market adjustment amounting to $56.9 million, primarily related to our CDO exposure to subprime or MBS written under credit (inaudible) swap contracts. Ambac's portfolio of CDOs totaled $71.1 billion at June 30. Of the total, approximately $60.2 billion has been executed under credit default swap contracts and the remainder under insurance policies.

The credit characteristics under both types of executions are essentially the same, including the fact that, in the unlikely event of default, claim payments are paid over time as scheduled principal and interest comes due. However, since CDS transactions may not involve a cash bond or do not require the guaranteed party or beneficiary to hold the bond, these contracts must be marked to market under Generally Accepted Accounting Principles.

In the second quarter of 2007, the negative mark-to-market was driven primarily by our exposure to CDOs of asset-backed securities that are backed by subprime collateral. Tom will discuss that exposure later in the call.

Now, the negative mark-to-market should come as no surprise to those who follow this structured finance market, as the lack of liquidity in CDOs of ABS in recent weeks has been well-publicized and discussed throughout the market. Ambac's RAP transactions all attach at the super senior level, which means that there is a significant amount of AAA protection (inaudible) exposure.

On these transactions, as with all of our CDO exposure, Ambac expects that mark-to-market adjustments in either direction will reverse through the income statement over time as the transactions move towards maturity.

Also, with regard to decline in net income, quarter on quarter, as a reminder, the second quarter of 2006 results included rather large net realized gains amounting to $44.4 million. Most of that was related to cash recoveries received during the period for security and the financial services investment portfolio that had been written off in prior years.

In 2007, there were no similar gains or recoveries in the quarter.

While Ambac reports net income in accordance with Generally Accepted Accounting Principles, or GAAP, research analysts make certain adjustments to net income to calculate their reported estimates. Therefore, to enhance investors' understanding of our financial results, we continue to provide information on the items that analysts adjust out of GAAP net income to arrive at their current estimates. Those items are as follows -- net after-tax gains and losses from investment securities and mark-to-market gains and losses on credit, total return and non-trading derivative contracts.

In the second quarter of 2007, Ambac reported net after-tax losses amounting to $34.6 million or $0.34 per diluted share. Those amounts are added back to our GAAP results. This compares to the second quarter of 2006 when the recorded net gains of $32.9 million or $0.31 per diluted share that were backed out of our GAAP results. On this operating basis, earnings per diluted share were up 5%.

Some analysts also back out the after-tax effect of accelerated premiums earned on obligations that have been refunded and other accelerated premiums. Total after-tax accelerated premiums amounted to $25.6 million or $0.25 per diluted share in the second quarter of 2007, which compares to $22.8 million or $0.21 per diluted share in the second quarter of 2006. On this basis, our core earnings per diluted share were up 4%.

Turning to highlights on our credit enhancement production, or CEP, CEP, which represents gross upfront premiums plus the present value of estimated installment premiums on insurance policies and structured credit derivatives issued or assumed in the period, came in at $367.8 million, down 31% from our all-time record production of $531 million in the second quarter of 2006.

Now, I will take you through some details of the production by business segment. Public-financed CEP was $114.5 million, down 13% from the second quarter of 2006. Overall market issuance of approximately $124 billion was up 13%, while total market penetration, the percentage of bonds issued during the period with financial guaranty insurance, was approximately 45%, down slightly from the 46% we saw in the second quarter of 2006. As has been the trend recently, the mix of issuance and the competitive environment have combined to (technical difficulty) production and public finance. Ambac continues to be focused on the more highly structured asset classes within this sector.

Turning to structured finance, structured finance CEP was $159.1 million, down 25% from the second quarter of 2006. As you may recall, in the second quarter of 2006, Ambac closed two very large transactions included in Dunkin's Brands transactions.

We continued to see a very active market across the broad mix of asset classes. However, the business environment remains fairly challenging in all but the MBS-related areas. During the quarter, our strongest writings were in CDOs of ABS where we have been cautious and selective, yet willing to underwrite very well-structured transactions at premium levels that are significantly higher than what we saw just six months ago.

Our current quarter's direct RMBS production was also up from the comparable prior period as we continue to receive a high level of inquiries and opportunities. Ambac did not write any direct subprime RMBS during the quarter. We will continue to be selective in the business we write and are obviously hopeful that the pricing improvements noted in RMBS-related classes will spread to other asset classes.

Turning to international finance production, international CEP in the second quarter came in at $94.2 million, down 49% from a very strong second quarter of 2006. The second quarter of 2006 included a strong flow of transactions that were backed up in our pipeline for various reasons, including two large UK transactions that made up almost half of that quarter's production. This quarter's international production is sourced from diverse geographical regions and includes strong writings in whole business and commercial asset-backed securitization.

Now, turning to the GAAP results for premiums, net premiums and other credit enhancement fees earned, excluding refundings, increased to $195.3 million. That's up 4% from the second quarter of 2006. Public finance earned premium, excluding accelerations, grew 2%. Public finance earnings have been impacted by the high level of refundings in the book over the past few years as well as competitive pricing in this business sector. Structured finance earned premiums grew at 10%. Excellent recent production with strong pricing in asset classes such as commercial, ABS and CDOs has offset the negative growth trends caused by high prepayment activity in this sector.

International earned premiums declined 2%. The negative impact of recent runoff and early terminations in the international book has offset the strong production generated over the past year.

Accelerated earnings from refundings and terminations amounted to $43 million pretax in the second quarter of 2007. That is the second quarter in a row of near record level accelerated earnings. Included in the current-quarter amount is $11.4 million from terminated structured finance and international transactions.

Our deferred earnings, representing future earnings on premiums already collected and the future value of installments, stands at $6.3 billion. These deferred earnings will be recognized as earned premium and other credit enhancement fees in the future over the life of the related exposures.

Investment income was $113.2 million, up 8%, substantially due to growth in the portfolio, driven primarily by strong operating cash flows in the financial guaranty business.

Turning to losses and loss adjustment expenses, loss provisioning amounted to $17.1 million in the quarter, compared to $12.8 million in the second quarter of 2006.

I will now provide some details on that activity. Total net loss reserves at June 30, 2007 amounted to $251 million, up from $226.5 million at March 31, 2007. Total loss reserves include case bases to reserves which are recorded for bonds in our portfolio that have defaulted and active credit reserves or ACR for bonds that have not defaulted.

Case reserves of $47.3 million at June 30 are up $9.6 million from March 31. The increase is primarily due to recoveries received during the period related to the Eurotunnel transaction. Ambac had anticipated those recoveries and had recorded an offsetting receivable within its case reserve. The recoveries were received during the second quarter and the receivable balance has been extinguished, causing the reserve balance to increase. As a result, total net claim recoveries received during the quarter amounted to $7.5 million. ACR of $203.7 million at June 30 are up $14.9 million from March 31, driven by additional reserves set up for certain transactions, mostly in the public finance sector and to a lesser extent in subprime RMBS. Those additional reserves were partially offset by ACR reserve releases on transactions showing favorable credit trends throughout our public and structured finance book.

Our below-investment-grade exposures declined by $326 million during the quarter to about $4.7 billion, or less than 1% of our total portfolio. Our below-investment-grade mortgage-backed and home equity asset class balance increased $350 million to $1.1 billion. Pool debt obligations remain relatively unchanged at $69 million of below-investment-grade exposure. To date, we have seen no significant deterioration in our subprime RMBS-related exposure.

Internal rating actions relating to the mortgage-backed and home equity asset classes involve three non subprime second lien transactions.

It is also noteworthy to mention that none of our direct subprime or CDO tranches were included in the recent rating actions by the rating agencies. Additionally, we have recently internally downgraded a limited number of pre-2004 vintage subprime transactions, in all cases from a rating of A to a rating of BBB. However, there have not been any internal rating actions within the CDO portfolio or subprime mortgage asset classes relating to the 2005 or 2006 vintages. Ambac continues to actively monitors these transactions, closely analyzing (inaudible) performance and structural productions available to us.

Our gross financial guaranty underwriting and operating expenses for the second quarter of 2007 amounted to $48.9 million. That's up 7% from $45.5 million in the second quarter of 2006. The increase is primarily due to increased compensation expenses.

Our Financial Services segment is comprised of the investment agreement business and derivative products business. Financial services net revenues, excluding realized and unrealized gains and losses, were $9.2 million. That's down 15% from the comparable prior period, primarily due to lower revenues from each of our financial services products in the second quarter of 2007.

Our return on equity on a GAAP reported basis was 11.5% for the quarter, reflecting the negative mark-to-market adjustment that we took. On an operating basis, our return on equity was [14.1]%.

Ambac announced today that we are increasing our quarterly dividend by 17% from $0.18 to $0.21 per share. Ambac has raised its dividend every year since going public in 1991.

Just a quick comment on our share buyback program -- as mentioned in our earnings release, during the second quarter, the Company completed its $400 million accelerated share buyback funded primarily by the February issuance of hybrid debt. Under the accelerated share buyback, total number of shares purchased amounted to 4.46 million shares, including 204,000 shares received in the second quarter. Additionally, during the second quarter, Ambac purchased 194,000 common shares outside of the accelerated share buyback program at a cost of $16.9 million. Ambac remains committed to managing our capital structure to a level that is consistent with a AAA-rated Company while maximizing shareholder value.

I just wanted to give a quick update as well on the Financial Accounting Standards Board project for financial guaranty. As many of you know, the FASB released its proposed statement of Financial Accounting Standards titled "Accounting for Financial Guaranty Insurance Contracts" on April 18 of this year. The comments on that exposure draft were due back to the FASB by mid-June, June 18. Those comments, in conjunction with a roundtable discussion with users of financial guaranty financial statements, including representatives from the industry who will be considered by the FASB staff before a final statement is drafted. That roundtable is expected sometime in early September.

FASB recently disclosed, on its Web site, that it expects to issue a final standard in the first quarter of 2008. For reasons highlighted in Ambac's comment letter to the FASB, which can be found on the FASB's Web site at www.FASB.org, and until a final standard is released and its provisions are well understood, it would be premature to estimate the impact on Ambac's financial position and results of operations. Our past financial filings with the SEC discuss the FASB's project and the potential for future changes to loss reserving, premium revenue recognition and expense accounting. We will be updating that disclosure with our next quarterly filing.

So, in summary, Ambac had a good quarter with solid new business production. Ambac remains diligent in the structuring of transactions, particularly in those asset classes where demand for our product is improving. We hope to see an improving business environment across more asset classes in the near future as the investment community begins to demand increased risk pricing across the various markets.

That concludes my prepared remarks on the financial results. I would now like to turn the conference over to Tom Gandolfo to discuss CDOs, including our underwriting approach and our current exposure. After Tom's remarks, we will then open it up for general questions. Tom?

Thanks, Sean. What I thought I would do is give just a brief overview of the types of CDOs that have been getting most of the press, how we anticipate and where we participate in the market, a brief overview of our underwriting process and then comments on our existing book of business.

Ambac has participated in both the high-grade CDO of ABS and mezzanine CDO of ABS markets. Both high grade and mezzanine transactions are primarily comprised of mortgage-backed securitizations and other ABS CDO securitizations. High-grade transactions are generally comprised of securitizations rated at least A- and having significant AA and AAA holdings. The underlying mortgages are a mix of subprime, mid-prime and prime.

Mezzanine transactions are comprised of securitizations rated primarily BBB- through BBB+ and are primarily comprised of subprime collateral. As a result, mezzanine transactions require a greater level of subordination to achieve a AAA rating than a high-grade transaction would. Throughout 2006, Ambac only participated in the high-grade space, as we didn't feel the mezzanine market adequately compensated us in terms of subordination and premium.

In the first quarter of 2007, we began to see significant price increases in much-improved subordination levels. As a result, Ambac participated in two mezzanine transactions in the first quarter of 2007 and another two transactions in the second quarter as pricing structures improved further. As you can see in our Web page disclosure, we received significant subordination in excess of the level that would've been required to achieve the AAA ratings from at least one of the major agencies.

Ambac does not underwrite based solely on the deal's public rating. While a AAA rating from a major agency is a minimum requirement, we put each deal to a rigorous review process and then sign an Ambac internal rating. This is why you'll see, on our Web site disclosure, that all our CDO of ABS exposures were executed at subordination levels well in excess of a rating agency AAA attachment point. We believe our credit-risk analysis goes far beyond that which a typical CDO investor would perform.

(inaudible) highlights of our underwriting process for ABS CDOs include the following -- a rigorous review of the CDO manager. This includes on-site visits by our senior credit people to review the managers, key personnel, their systems and their oversight controls. We do a detailed assessment of the triggers and control rights embedded in the CDO. These triggers and rights include rights that would suspend the reinvestment period, force the deal into a sequential pay mode, give us manager removal rights and give us liquidation rights -- because remember, we anticipate always at the most senior point in the capital structure.

All risk-return modeling is performed by a risk-management group that is independent of the underwriting group. So they don't report into my area; they report separately to our head of risk, and they are responsible for controlling and running all models.

We do a detailed review and re-rating of all the underlying RMBS collateral in the deal. We do a base case and a stress case model. Our base case model assumes a minimum 30% default time correlation. The decision that goes into what we actually use obviously depends on the compensation and quality of the pool, the vintage, the servicer, but at a minimum, we use 30%. Our base case model then assumes average recovery rates in the underlying RMBS of 40%, so we assume a 60% severity.

In addition to the base case, we run a stress case model that is presented to credit committee when they get these deals approved, so we present both base and stress to the credit committee. The stress case modeling doubles the base case correlation, so at a minimum in their stress case, we would have a correlation of 60%. The stress case modeling cuts the base case recovery rates in half, so from 40 to 20, and our stress case modeling actually cuts the ratings of the underlying RMBS collateral. One example would be we do it by three notches and then we do another example by six notches, which of course increases the probability of default in our model significantly. So as you can see, we really put these transactions through quite a rigorous test before we would approve a deal.

Market liquidity for ABS securities, particularly the mezzanine deals, has been significantly reduced as the subprime sector has displayed disappointing performance. We agree with the market consensus that the 2006 vintage subprime is of significantly poorer quality than previous vintages, and that would include the first quarter of 2007 as well. Investors are struggling to understand better the ultimate risk of loss in this asset class. As a result, credit spreads have widened significantly. We don't see price stability coming back to this market in the near term.

Fortunately, Ambac is in a very favorable liquidity position compared to many other holders of CDS exposure on this asset class. The reason for that is our contracts, our (inaudible) agreements do not require us to post collateral based on market value declines, regardless of the size of the mark-to-market and regardless of our rating. So, this is an extremely valuable option that really is only afforded to the monoline players. So when you hear about forced selling in the market, that would never be a problem for us, so we guard that very carefully.

We are very comfortable with the quality of our book. None of our CDO tranches that Ambac has directly written protection on have been downgraded or even put on watch by either Moody's, Fitch or S&P. We will keep you apprised of this, any changes, because we disclose to Moody's, S&P and Ambac internal ratings of our CDO of ABS exposures on our Web site. That disclosure will be updated quarterly.

Sean?

Okay, now we would like to open the conference up for questions.

Thank you. Ladies and gentlemen, at this time, we will be conducting a question-and-answer session. (OPERATOR INSTRUCTIONS). Ken Zerbe, Morgan Stanley.

Obviously, you have been more active in writing the mezzanine CDOs and CDO [squareds] during 2007. Could you just tell us? I mean, some of the cumulative loss numbers that we are hearing [against] the subprime are pretty high, and I know they mostly apply to '06, obviously some to first quarter '07. How do we get comfortable that, even today, that the mezzanine CDO [squareds] are still going to be money-good even with 30%, 40%, 50% subordination?

Sure. It's Tom. I would kind of lump -- whether it's a CDO squared or a mezzanine deal, the underlying collateral is the same, which is primarily BBB tranches, so it's really irrelevant whether it's a CDO squared or a mezzanine deal. I think the way we look at it is we do not believe -- obviously or we wouldn't have underwritten the deal -- that all 100% of the universe of 2006 collateral at the BBB level is going to be written down. I believe think our belief has been confirmed. If you look at -- and there's no shortage of published data. Almost daily, the major banks are putting out data on estimated losses in these tranches. I think that the market consensus would agree with us.

I think another area where -- perhaps could give you some comfort is whether you agree with the rating agencies or not on the initial rating, certainly the rating agencies in the last couple of months have this asset class under a microscope, and they have looked at it very intensely. If you believe (technical difficulty) they believe the losses were going to be at excessive levels that would obliviate (sic) these BBB tranches, there certainly wouldn't be any BBB tranches today because they would be written -- they would be downgraded CC.

So I think that's our view. I think it's supported by the market. I think, in terms of why subordination is so important in these deals, it's because will there be some BBB tranches that will experience excessive (inaudible)? Absolutely. There will be tranches where the originator was of poor quality or the servicing isn't done right, and there will be losses. So, that's why subordination is so important, because the subordination will absorb those tranches that will experience those losses.

Okay. All right. The second question I had, just again looking at your enhanced CDO disclosures, it looks like you are at 1 mezzanine CDO squared, $1.4 billion or 30% subordination below you. But then you'll also did a high-grade CDO, which was not necessarily a squared deal, with 35% subordination. I mean, naturally, I would just assume that the mezzanine deals, whether again squared or not squared, they shouldn't necessitate greater subordination. Why does high-grade have more than the mezz deal?

The mezz deal that you're looking at is actually comprised of 100% AA tranches, so there's no A, -- there's nothing below AA in that mezzanine deal which -- there's a lot of nuances here but I will try to just touch on the high-level ones. Those AA tranches of a CDO benefit from all the overcollateralization triggers and other triggers and rights that are embedded in that CDO. So in and of itself, there's a lot of subordination in the inner CDO, if you will. So that's -- what you see in the 30% is the subordination in the outer CDO. There's also a fair amount of subordination in the inner CDO there. So that's what you see.

Now, candidly, that high-grade deal that you see with 35% subordination is -- although we haven't opened the floodgates here, we are still selectively writing business. The way the market has turned here in the left four or five months is we are able to get incredible terms. This deal with the 35% subordination is a very good deal that, if we were to have written it six months ago, this deal would have been offered to us probably with 15% subordination and a lot less premium than we got on it. So this is just an opportunity to come in and write pretty good business at really good subordination levels. That's what you're seeing with that 35% deal.

Great. My final question -- with your stock in the high 70s, is there a certain point where you would just say "Let's get much more aggressive with share buybacks because we're going to get a better return on buybacks in the new business at this point"?

Yes, Ken, that's certainly goes into the decision. You know, as we've stated before, the primary part of the decision is our view on our excess capital levels, considering business opportunities, potentials for downgrades and the like that would potentially eat into that excess capital position. But with the stock where it is, it does kind of help go into -- if you had a range, if you will, it might go towards the more aggressive part of the range of a particular buyback decision.

Great. Thank you very much.

Tamara Kravec, Banc of America Securities.

Can you hear me? Sorry. I can't get off speaker.

The first question I have is, just so I'm understanding this clearly, that in your high-grade CDOs, the CDO, CDO collateral that you disclose, which averages about just under 20% of all the deals across the vintages, that is comprised mostly of BBB underlying collateral that is variable in nature. Is that correct?

That's correct, Tamara. If you look -- what those are, those are A, AA and AAA-rated tranches of other CDOs that are comprised of BBB-rated tranches and mortgages. So when you look at our Web site and you look at our disclosure -- I didn't bring it with me -- but the subprime component there would be disclosed in the numbers.

Okay. The RMBS component, you have -- in your breakdown, you have the RMBS and you have subprime, and then you have the CDO squared component. The RMBS component, have you -- you said you were stress-testing it for a three-notch downgrade, a six-notch downgrade. Does that include the RMBS component of that? In other words, how worried are you that the rating agencies aren't done yet and that some of the prime RMBS is actually not really prime, it's going to move down into the subprime and worse collateral?

It absolutely includes the RMBS class. When we do our stress runs, our down-three or down-six, we don't limit it to just subprime. We do that across the board, so all collateral. Candidly, particularly with the 2006 vintage, when the downgrades came out, we were surprised at how drastic the market reaction was because we thought the market kind of expected a round of downgrades. But no, absolutely, we apply that to all the mortgages.

Okay. Can some of the issues that potentially could be downgraded may already be on your internal watchlist? Would that be fair to say?

Are you talking in the CDOs, or in our direct mortgage?

Either/or, like, if any of the insured amounts that you have in terms of the collateral -- could you have something on watch right now that hasn't yet been downgraded that could be, but that maybe you have reserves up against already?

Yes, I can address that question, address it both from the CDO side. The rating agencies (inaudible) downgrades and what they have done with the negative watch not include any of our specific tranches, so that includes CDOs and RMBS transactions. Obviously, with the CDOs, there will be collateral in those CDOs that either has been downgraded or is on negative watch.

In relation to our direct business, we have had -- we've seen some very limited number of transactions where a subordinated tranche would either be either on a watchlist or has experienced maybe a one to a two-notch downgrade of the subordinated tranches.

We don't necessarily consider it an action in and of itself because we're looking at the underlying data tapes and the performance of the underlying pools and doing our own modeling off of that based upon (inaudible) data.

We did take some internal actions that obviously were not taken by the agencies and we did downgrade some transactions; we downgraded three transactions, which you can also see on our Web site, our new Web site disclosure on our consumer asset-backed, but we downgraded three transactions to just under below-investment (technical difficulty) classification due to some performance statistics that we saw. Those transactions were not subprime transactions, and they were related to second-mortgage transactions.

Okay. Then if you could give us an idea of how many of your -- or what percentage of your transactions have (technical difficulty) triggers on them.

Virtually -- in the CDO, all of the high-grade deals have OC triggers, and all but one of the mezzanine deals have OC triggers.

All right, great. Thank you.

Darin Arita, Deutsche Bank.

Good morning. I guess the question is for Tom. It sounds like (inaudible) CDO manager selection processes is very rigorous. I guess one of the reasons why Ambac has not participated very much in the our MBS market or subprime market over the past few years is other market participants such as CDOs were willing to buy the RMBS with insufficient protection according to Ambac (inaudible). So would it make sense at all for Ambac to turn around and ensure the AAA tranches of those CDO managers?

Yes. I think what we can do and I guess the advantage we have in the CDO market is when we look at these deals -- and you raise a good point. To give you an example of a recent deal that we did, we're doing this all in the primary market. We don't do this in the secondary market. So we have the ability, before the transaction is sold, to go in and say, you know, we can look at the originators and we classify the originators; we have a system here where it's Tier 1, Tier 2 and Tier 3, obviously Tier 3 being the weaker seller/servicers. We can decide if there's too much Tier 3 in there, we can say either, let's say the deal has 10% Tier 3. We could say to the banker, okay, we want another 10% subordination because we're going to just write that entire bucket off or pull those names out of the deal.

So, we're not trapped just because we do the [CO] that we have to except what's presented to us. We are in a month, three weeks before the deal hits the market, so we have the ability to have a say as to what goes into the deal. If we can't get our way, we can always either make it up with subordination or just don't do the deal.

Okay. All right, but you definitely have a view on the quality of the various CDO managers and that goes into (multiple speakers)?

Absolutely, I can tell you what we will do. In some cases, we will look at names and we will just assume 0, 0 value to the name. That's part of our underwriting process. We certainly have a view on every originator and seller/servicer.

Okay. Thinking more about the collateral within the CDOs, and if we just look at -- stepping back from CDOs and looking at the subprime residential mortgage-backed securities that Ambac has insured, it seems like there has been a shift in appetite for having adjustable-rate mortgages to having completely fixed-rate, first-lien mortgages in those deals. Would it be fair to assume that Ambac, on its CDO exposures, would have had a bias towards first-line, fixed-rate mortgages?

Absolutely first lien but not necessarily fixed-rate. In the CDO market, generally speaking, kind of 80% because the debt that's issued against these CDOs is generally floating-rate debt. It's very difficult to buy fixed-rate mortgages and issue floating-rate debt because for obvious asset liability management reasons. So in CDOs, you have a heavier weighting towards floating rate collateral.

But as far as second liens go, we, Ambac, has a bias against second liens in the CDOs.

As you think about the potential for interest payment resets, how do you flow that through your [mottos] and stress test?

What we do is, we look at -- two of the components in our -- if you look at our models and stress tests, obviously CPRs are affected by interest rates, so we stress the CPR, the constant prepayment rate. We look at -- everything is modeled on a current interest rate curve, so we assume that, if there's collateral that was issued at a teaser rate, we assume that's going to get reset into the cash flows. It's factored into our constant default rate assumption. When we downgrade three and six notches, that factor is due to our probability of default. So I think we capture -- I don't think anything has fallen through the cracks.

Darin, just some comments on the direct subprime, and that is on the book. Obviously, we did ratchet back the origination, particularly to 2005, 2006, and the first half of 2007 here. The one deal we did do in 2007 was a first-quarter deal. That had 100% fixed-rate collateral. The deals that we did in 2006, 88% of the underlying pools were fixed-rate collateral, as well as 2005. So we've had a very heavy bias towards that type of collateral that will not be subject to reset.

Okay, great. Thank you.

Mark Lane, William Blair.

Good morning. A few questions -- first of all, Sean, can you just give us an idea? Theoretically, if you have $1 billion CDO of high-grade ABS and it's downgraded one notch or three notches, what would be the commensurate increase in capital requirement?

If it were the to be downgraded one notch from AAA, in this particular case, to AA+, I think we did internally run some scenarios. Those numbers, the entire book now of CDO of ABS, I believe those numbers were in the neighborhood of a couple of hundred million for the entire book.

To go from AAA to AA +, one notch? Okay.

Yes.

Okay. So is it safe to assume that it kind of accelerates from there, or is that the biggest hurdle, just the initial downgrade or --? How does it work if it goes down?

Actually, the way the rating agency models work is they are based on a probability of default and a loss given default for individual transactions. Then those transactions are put through a modeling exercise, and the agencies are interested in the tail event, so if you were to run a scenario 1 million times, they are looking at the very, very tail-end of what potentially bad could happen. That would be the impact on our capital. So when I quote those numbers, that's the exercise -- is at the very tail end event and how that impacts their AAA modeling.

Mark, if these deals were to deteriorate significantly to the point where the super senior tranche was downgraded, what would happen is the deals would become much shorter because, as deterioration occurs, triggers kick in, which would suspend reinvestment, would take the manager's ability away to reinvest, and then the deal would go sequential, which means cash flows would be redirected up to the senior class in the CDO and would start paying down our layer on an accelerated basis. So you know, your seven-year deal would become shorter.

So Tom, your stress test of the CDOs -- is there anything that, in the last three to six months, you've changed within your modeling process to include other factors that may be predictive that weren't included before?

Not a whole lot. Fortunately, the reason we didn't underwrite mezzanine deals in 2006 is we just didn't think the risk/reward profile was right. So we took a harsh view a year ago; we took a harsh view two years ago. So we didn't have to change much currently. I probably would say the only thing today that I would look at differently, you know, we would try to always limit second-lien buckets to, you know, kind of 5% to 10%. Today, we -- and we can do this because we can pretty much structure -- we are holding a lot more cards now in the structuring. I would probably keep that bucket to 0 or maybe 5%. Or I just would ask for more subordination to cover that bucket. But other than that, probably nothing different.

Okay. That includes all of the documentation issues and all that sort of stuff as well, right?

Correct. The documentation, we've always taken that very seriously. The most important -- well, everything is important but the collateral support (inaudible) that carves us out from the requirement to post collateral, you know, that's been important to us since 2000.

Have you taken -- the question is have you taken a broader look or a deeper look now just at your MBS book overall? What are you seeing, if anything?

In the context of underwriting, Mark?

No, in the context of your prime book basically, your home equity loan book and (multiple speakers).

Yes. What we're seeing there, (inaudible) broad state we are seeing some select deterioration of pools. Particularly obviously the transactions that we downgraded to below-investment grade, we are seeing some increase in default levels. The severity on those transactions are already quite high, just due to the nature of the collateral being second-lien collateral, so taking that out, out to using those types of elevated levels and projecting out over the term of the transaction, which we are able to do on a monthly basis and track it very closely. Some select transactions looked a little bit more stressed. But as a general point, over the entire portfolio, we are not seeing that. I think that's largely due to selectivity a little bit. Tom talked about the CDO managers but obviously the servicers of underlying pools make a big difference. The nature of the collateral and the ability -- as Tom mentioned the (inaudible) pools on primary deals, when you're in before you can sculpt your terms, has helped us, and that's what we are seeing.

(technical difficulty)

Andrew Wessel, JPMorgan.

I just had a couple of follow-up questions on the CDOs. Could you highlight for us, on your CDO breakout that you give, what if any of the high-grade or mezz deals have primarily or majority second-lien exposure?

None of them have primarily second-lien exposure.

Okay, good. Then based on, you know, it's hard to use as an index but based on ABX pricing which technicals abound there and I'm sure it's wider than what really fundamentals would say, but the BBB tranche is kind of pricing at 45% to 55% losses currently. Can you make some comments towards how you look at that index versus the actual performance of bonds you're insuring? And then kind of as a tack-on to that, to the deals that you actually have the CDS exposure to as opposed to a RAP, when you're marking those deals, how does that come into play?

Yes, let me take the first part and then I will let Sean talk to the mark-to-market, because that's done by our independent finance group.

I look at the same indexes that you look at. It is really hard, right now, to know how much of that spread widening is fundamental and how much is technical. I'm sure it's a little fundamental and a lot technical, because I think there's just such an overwhelming demand to short or buy protection on subprime mortgages that there's a lot more buyers of protection than there are sellers. So I think that has impacted that index from a technical perspective. So, I don't know what the fundamental value is there.

What we do is, when we look at our deals, we don't feel we underwrite the market. As I said earlier, I think when we come in and we look at BBB- or BBB names that are in the pool that we underwrite where we don't like -- I mean, candidly, if we saw New Century or Freemont, those happen to the Tier 3 on our list -- we will either ask for more subordination, we will write the credits off in our modeling, or we won't do the deal if we can't get our way.

So I think you can't do that in the index. You know, the index is what it is and you have got to take it for what it is. So I think that's kind of the difference.

Sure.

Then, to translate that over to the mark-to-market, we are not pricing off an index. We're going out and actually getting underlying bond quotes from the dealers who arrange those particular transactions, so it's transaction-specific. As Tom mentioned, there's unique elements to transactions of these underlying transactions. So, we are actually getting pricing from underlying dealers and taking that pricing, turning it into spreads and looking at how the markets moved from the origination of that particular transactions. That will translate into a mark-to-market gain or loss because, effectively, we're trying to mark our financial guaranty contract, or the credit default swap contract, to a value.

Now, we have seen situations in the quarter. Obviously with spreads going out there doesn't seem to be some technical factors getting into the pricing as well due to the lack of liquidity in the marketplace. So, we are seeing that in the underlying bond quotes, which is obviously working its way through our mark-to-market.

Okay. Another follow-up into the mortgage exposure in general. I mean following Countrywide's commentary about their, you know, the big marker in (inaudible) they took on prime home equity loans and how they are performing much worse. What's your review on kind of higher-quality, the higher-quality paper out there, and what the deterioration effect could be for the general market as we move up into all-day and prime quality? I guess mainly you're looking at first liens but to the extent that you're looking at home equity or second liens as well?

I will comment on that. Again, you know it's hard to make general market comments. A company like Countrywide perhaps can do that. Let's kind of talk about our select transactions that we are seeing. We're not seeing, on the first lien side, nothing that's causing any internal downgrades, so we're looking at monthly data tapes and looking at the data there.

We are seeing, in the cases where we had some internal rating actions, we are seeing some poorer performance in limited numbers of home equity. In one case, a HELOC is (inaudible) 2005 deal, and in the other case, a 2006 closed and second deal. The closed and second, this particular deal has all fixed-rate collateral. They are not subject to any types of interest rate shocks but nonetheless, we are seeing a higher level of underlying defaults there. When we model that out, we assume 0 recoveries there. So the default rate drives cumulative losses from our perspective in the pool.

So in that particular case, even with those elevated levels of defaults, we wouldn't be looking towards any potential claim payments, and we clearly are not there yet until around the year 2010, so still quite a way far out, so there's some projections going on here and some estimates to come up with those underlying results. But nonetheless, we thought it appropriate to take an internal action on those deals. (multiple speakers)

I will point out, you know, our Web site does have disclosures that we've enhanced and that you can sort them any which way you want, so you can see all of those particular detailed disclosures and you can see our current rating. You see the specific deals where we took actions against.

Okay, thanks. My last question, just in general, in pricing in the structured finance market, are you seeing moves or rumbles towards better pricing outside of residential mortgage, or is it still kind of holding tight or in a holding pattern?

I think, in CDO land, outside of -- ABS pricing is incredibly strong. But outside of ABS, we are starting to see some contagion into, let's say, CLOs, for example, nowhere near as dramatic as ABS, but you are seeing pricing improve in that market because I think there's a little bit of a flight to quality going on out there, which is widening spreads in some of the asset classes.

Great, thank you very much.

Geoff Dunn, KBW.

I just had a quick question for Tom. Some of the pushback we get on any exposure to CDO squared is you guys can't really see past the first layer. Can you just give a little bit more color on how deep can your underwriting go when you're looking at CDOs that have CDO collateral in it? Are you just looking at the tranches, or can you get a feel for the actual underlying CDO and how good of a quality that tranche actually is for your exposure?

Sure. No, what we do is we of course look at the inner CDO when we underwrite this. I think we've -- pure CDO squares, I think we've done three deals. But what we do is we use a system; we use a lot of systems, but one of the primary ones is a system called [Intex], which is widely used in the market. What Intex does it allows us to run cash flow scenarios on the inner CDOs that are consistent with our view, Ambac's view, of the nature of the underlying subprime collateral. So, if we think we want to apply a certain kind of CPR or a certain default rate, that's our view, which may be more conservative than the markets and normally is -- we can do so in this modeling. Then we take those cash flows, and we use them to estimate what type of write-down, based on our assumptions, what type of write-down could we expect in that inner CDO. That's how we size what we think we need for subordination in the outer CDO.

Okay, so you are actually looking at the granular assets of the underlying CDO?

We don't go in and actually model the RMBS in the inner CDO. What we look at is the components of the inner CDO. Let's say it's 90% BBB subprime, as an example. What we would do is we would say, okay, here's what we think the proper default rates should be used, here's what we think the CPR is. Then we can simulate the cash flows on that inner CDO then run it through the waterfall, determine what we think the write-down should be, and that's how we size them. So, we're doing it at the CDO level and the broad asset mix, not the -- we don't do it for each underlying RMBS in the inner CDO.

I got you. And another question -- back at your investor day, there was talk that the global infrastructure market in continental Europe was returning. Are you still seeing that in your pipeline, or has there been any change in that trend?

Yes, we are seeing certain transactions. You know, we closed obviously a large toll road transaction, big infrastructure transaction earlier in the year. We are seeing different geographies, not necessarily continental Europe but we are seeing transactions in Canada, Mexico and other places. So I would say characterization is we are seeing select activity in that particular area. We are seeing it more robust, however, in the international sector on the structured finance side, so that continues to be robust. That's how I would characterize what we are seeing right now.

Still a net improvement over the kind of pipeline we might have seen a year or in the past two years?

Yes, I mean we cleaned out the pipeline obviously in the second quarter last year, finished some deals that were in there for a while. So I would characterize it as being steady, continue to be lumpy-type transactions. I know we are looking at some infrastructure deals that would probably be in the pipeline towards the end of the year, decent-sized transaction with decent CEP. I would characterize it as being more steady, not necessarily dramatically up or down.

Heather Hunt, Citigroup.

Thank you. Good morning. I just wondered if you could describe the geographic scope of the RMBS that were downgraded and sort of compare that to the (inaudible) scope of your portfolio and of the market. Is there anything in particular sort of -- is it pretty well diversified or concentrated?

Yes, Heather, you're talking about our internal-downgraded transactions?

Yes.

Yes, I think those particular transactions, one of them did have a concentration towards California. The HELOC transaction had a concentration towards California collateral, heavily owner-occupied, single-family homes but concentrated in California. I'm not sure of the other transaction that was downgraded. I would have to get back to you on that, kind of where exactly that collateral was based.

Okay, then on the CDOs, I think one of the advantages of the CDOs is diversification of the sources of collateral. I wonder if you could describe the scope of diversification in terms of the number of different loans, geographic dispersion and the types of loans, meaning (inaudible).

One of the things that -- with the high-grade CDOs and in the mezzanine CDOs, what we look for, you know there's concentration limits. So generally speaking, in RMBS, we look for a granular pool, so in RMBS, any single issue generally wouldn't be more than X%, 2%, 3% of the total pool. So what you won't have is you won't have a single kind of RMBS securitization representing a disproportionate size of the total CDO. So that's how we get some diversification. We certainly look for diversification in originator, but at the end of the day, they are heavily, you know, 90% residential mortgage, so the CDO is highly correlated to the mortgage industry.

How many different mortgages and originators might be comprised in a given (multiple speakers)?

Oh, gees, if it's a $1 billion high-grade deal, it can be thousands of mortgages.

How about mortgage-backed securities?

Figure if it's 2% concentration limits, so there could be 50, yes.

Okay, and then originators?

Almost all the originations -- the CDOs are generally going to represent the market, which is why what we do is we can go in with the portion of the market we don't like and we can size the subordination accordingly. But in general, the CDO will probably represent the volume in the market.

Thanks. Then as a follow-on, are you starting to see any distress on European RMBS? It is still early there but it sounds like there's a little bit of distress there. Is that creating opportunity at all?

Heather, we haven't seen any distress in that particular area or any information coming out or data that's indicating that there's distress.

Okay. I guess I've been hearing that from folks at Citigroup.

Then finally, you know, you were saying earlier in March and April that the ROE on your new business has gone up. I wondered if you could compare the ROE of your current business to what you were doing maybe six months ago.

Okay. I think, you know, I will compare it in general terms and I will kind of go by sector. I think what we're seeing in public finance, we are seeing, as we mentioned on the last couple of calls, is kind of a change in the mix of business we are writing to more higher-grade, less structured transactions. Generally, even though the capital requirements are lower, we've seen both that impact, as well as competition, even though spreads generically in (technical difficulty) finance have held steady, we've seen some erosion there.

On the structured finance, as Tom mentioned, we wrote some transactions. We had a particularly good quarter in structured credit, so we were able to enjoy increased pricing in those particular transactions, significantly higher and with better structured than we've seen, so I think in that particular market, we've seen some good activity.

From structured finance on the commercial, on the commercial ABS side, those transactions typically are highly structured transactions and will carry a higher return, if you will. We continue to see a lot of activity there, working on a lot of deals internally that hopefully will (multiple speakers).

Is it like -- I mean, is it like 10% -- I mean, is the increase, you know, did you go from like 15% to 30%? I know that there were some deals that you were getting pretty good, really high ROEs. Are you willing to even comment on that?

I think, Heather, it's all over the board. It depends on which sector you're looking at. In CDO land, the ROEs have gone up significantly, but I hesitate because I can't promise you it's going to stay that way. So I wouldn't factor that into your projections. Today, the ROEs are extremely attractive, but the market -- part of this is liquidity-driven. If liquidity comes back into the market, then that could be impacted.

Al Copersino, Madoff.

Thanks very much. As far as how Ambac sets reserves, it's obviously pretty clear how the case reserves are set. For the ACR, though, did you just give us an update on what the thought process is there, if that has -- if there's any change to that process at all?

No, there hasn't been any change. We consistently apply that process. The ACR process is driven off of our internal -- originates from our surveillance group, their underlying view on transactions and underlying rating of transactions. That in turn will drives a classified list, as well as an internal rating for a particular transaction.

Our ACR is originally driven by that and the classification, particularly when a transaction gets to below investment-grade category. When that happens, what we do is, in order to measure a particular loss, a loss reserve for a quarter, we're looking in all the deals on that list and every quarter reanalyzing what an appropriate expected loss would be on that transaction.

So, in certain situations, we would be utilizing a view of the probability of default times the severity of a default if it were to occur. So, it's kind of an expected loss approach. In certain cases where we have active monitoring of a transaction, in certain cases as it continues to go down our classified list, then that could be on a daily, weekly, monthly basis. We will form a more concrete view, a better view of the transaction and then we will adjust our reserves accordingly.

So, while the particular bond kinds are probability and default and loss given defaults are driven by the particular bond kind that we're looking at, certain attributes of transactions would dictate the reserve setting. For example, you might have a situation where the underlying bond kind might be involved with transportation but there's some type of government guarantee involved. So we would consider, obviously, the unique attributes of a particular transaction in setting the number.

On the mortgage side, we would be looking at -- surveillance in our normal course is looking at information and modeling out transaction performance, all the way out to the end, making certain assumptions regarding prepayment speeds, default rates, recovery rates, and obviously looking at that to determine an appropriate rating. But that would also help us determine what an appropriate level of loss would be if we were to establish a loss.

You guys, in the past, have spoken about, since the '91 IPO, what your losses have been as a percentage of par. I think I can calculate from my own model what the losses have been as a percentage of premium. I don't know if you have this but could you break that out separately for the structured finance, or in particular the CDO piece? Do you know, historically, what the loss performance of Ambac has been on that piece in particular?

I don't have that handy. I mean, that's something that we could do. You know, I have broad numbers which I can -- I have something in front of me that has net losses paid since going public net of recoveries, that is, of $275 million. Net premium earned since going public is $6.2 billion. So, it gives you a sense. You know, I know, in those numbers, there is some losses taken for healthcare transactions. We took a loss for an EETC transaction, so those are the bigger ones in that number.

But CDOs -- I wouldn't want to guess on the call. You know, we can get back to you with a specific number.

No, that's very helpful. I appreciate it. Thank you.

Gary Ransom, Fox-Pitt Kelton.

Yes, thank you. Most of my questions have been answered but I thought I would ask on the other side of the credit portfolio. You make a comment in the press release that there was favorable movement or stable, or some offsetting things. Can you elaborate on what is going well or favorably in the credit portfolio?

Sure. We've taken some actions in the credit portfolio which have had a direct impact, in certain cases, on the level of reserves we are taking. Some of the credits down in -- that are the "Katrina Portfolio" have performed very well, particularly those that were initially downgraded for transactions that were outside of the New Orleans area. So, there were some transactions we had in Mississippi Gulf Coast towns as well as an Alabama transaction.

There continues to be quite a bit of, one, insurance money as well as state and other local funding available to these particular communities, so those transactions were upgraded and that had an impact during the quarter.

We also had the favorable resolution of the Eurotunnel transaction. In that particular transaction, we had reserves for -- additional reserves for loss-adjustment expenses that we thought perhaps that could go on longer than what had taken place. So that created some favorable numbers in our reserves.

The last thing is we upgraded to EETC, so the airplane trust certificate transactions. So those were the positive impacts during the quarter.

Could you tell us what your Katrina reserves are at the end of June?

The Katrina reserve at the end of June is $38 million.

Okay. All right, thank you very much.

Jonathan Adams, Oppenheimer Capital.

Could you elaborate on the comment you made with regard to additional capital to support a transaction if it were downgraded from AAA? My question is whether that is the rated tranche of the CDO or whether that is the so-called super senior that you are insuring.

Yes, that would be the super senior.

So you would have to, say, with your internal rating or the rating agencies would indicate that that exposure that you have, which again is higher than the AAA tranche, would be downgraded?

Yes, that's a very drastic scenario, but yes, you would have to go through all of the subordination presumably to get down to the natural AAA, and then it would be a one-notch downgrade, so that would take pretty -- very dramatic downgrades of the underlying collateral in those particular transactions.

[David Hinton], Boston Partners.

I had an immediate follow-up to what was just stated. I had missed the actual numbers that you had cited for the response or your answer to Mark Lane's portion of the question, so could you just repeat the numbers please?

I'd don't -- I'm looking for the exact numbers (inaudible) get back with you. Those numbers, if we were to take the CDO of ABS portfolio downgrade -- they are all super AAA -- and downgrade it from that level one notch to a AA+, I believe the number was a couple hundred million, so in the low end of that, so 200 to 300. But I don't have it in front of me, but that is a sense for what would happen. So that would presumably -- again that's a tail loss that would be calculated (inaudible) a model that would be absorbed by our current excess capital position.

Thanks. That was all I had.

Jerry Solomon, Bear Stearns.

There has been a lot of data released on the follow-ups. I just want to make sure I'm not missing anything. In terms of your -- you have '06 subprime vintage exposure. I'm just wondering. What rating categories do those fall in?

Those transactions are all BBB transactions. Just to give you a sense for the underlying there, we have two Countrywide transactions and one home loan. You can obviously see those transactions on our Web site. But 88% of that is fixed-rate collateral, 100% first lien.

But all of the AAA (inaudible) (multiple speakers)?

All BBB.

We've heard, there's been some -- I think I may disagree with the question I was going to ask before, but we've heard some stories that a lot of the RMBS in the CDO market is sort of in disarray and may be shut down a little bit for the summer. But I guess from a current business flow perspective -- and granted these deals don't get done overnight, but I guess near-term flow, has it slowed down a little bit, Tom, or is it still just -- I mean, you seem pretty active in the first quarter, the second quarter at least with the CDO exposures. How is it going in the third quarter I guess?

Sure. Well, in the mezzanine ABS space, if you're willing to do 2006/early 2007 vintage, you could write all the business you want, which obviously we're not going to do. You know, we have internal risk limits that limit how much of particular ratings or a particular vintage and we can take, so even regardless of subordination. So we've kind of slowed down significantly in that space. I don't think you're going to see too much more of that vintage written.

You will see us dabble in the high-grade of that vintage, depending if we can get -- if we can engineer the deal working with the bankers such that we get AA, kind of AAA underlying collateral and subordination levels and prices we want. You can still get those deals done.

I think where it's going to slowdown is the brand-new mortgage securitizations, because there's been such a decrease in volume. I think you're going to see a slowdown in that for a while, in the brand-new vintage.

Just out of curiosity, the RMBS in the CDO exposure, is that something -- do you reinsure that or do you keep most of it on your books?

In the CDOs, we keep most of it on the books because of any type of risk that we wanted to transfer there, we would capture in the subordination levels. We would just ask for more subordination.

[Tony Dalapenia], John Hancock.

Yes, thank you. A follow-up to a previous question where you talk about your U.S. subprime exposure by vintage, and you said I think the $1.1 billion of net (inaudible) outstanding direct, which is subprime, you're saying all of that is underlying rated BBB?

Yes.

Could you give more color on the '05 and '07 vintages as well by that rating -- by rating category?

Yes. The '07 transaction is one transaction. That's in A-rated transaction, again, fixed-rate, typical. The FICO average is around 611. The loan-to-value is 77%. 70% of that pool is full documentation. So that's what that particular pool is. 2006 is BBB. 2005, yes, it's a mix. It kind of goes across the spectrum and includes (inaudible) BBB and some AAA as well.

Okay. Is the '06 mainly fixed-rate as well?

80D%. Two of the deals are 100% fixed-rate, and one of the deals, the smaller deal, has a 13% fixed rate, 87% hybrid arms, the 327s.

Okay. Obviously, I don't follow the asset-backed sector closely, but you read these generalizations, from listening to the rating agencies, that the '06 vintage and the '07, at least the delinquencies (inaudible) say 10% to 15%, and they expect cumulative losses to be anywhere from 6% to 10% or even more, maybe. Could you give us some generic performance of your pool of '06/'07, exactly what that has been doing, so we can sort of compare versus what the experts are reporting in the press?

Yes. Most recently, I think Moody's released a subprime report just the other day, and as one would expect, there is a distinguishment between good-performing pools and poorly performing pools. Obviously, you're going to run around the range of numbers that I've seen are running around out there. You know, what they are quoting is the best-performing pools may see losses of the 5% to 7% range, whereas the worst-performing pools might be 10% and North of 10%.

Our particular transactions, we typically have loss coverage, even in the BBB deals, north of 8%. Those particular transactions, we haven't had to take any rating actions against those. I think they are performing kind of as expected in our underlying modeling. We do use a base case, as Tom indicated; he stressed the CDOs. There's also stress scenarios for the base case for the RMBS transactions. So, we are not seeing the level of deterioration that would cause any particular movement in the rating category for any of the subprime deals.

We did internally downgrade -- and again, you will see this on our Web site -- some older vintage deals that are coming to end of their life, so relatively smaller balances. We did take some downgrade action, so we're looking at all the pools and the performance of the pools, but these in the '06 and the '07 vintage are performing.

Yes, and that was my next question, that 6% and (inaudible) bucket of the subprime, what vintages are those? I guess my -- obviously, you're not following the sector would have a bad assumption, I would have thought anything over the last four or five years would have had home price appreciation. They allowed a lot of this stuff. I guess could you comment what vintages those are and why my assumption is wrong there?

Most of that dates back to the 2000-2002 years, but you go back and if you see our -- look into our operating supplements on that particular category, you'll see those numbers have been there for quite some time.

This quarter, we did downgrade 2006 vintage transaction, and that was a little over $300 million, and then we have a 2005 of $88 million in those numbers. So that, coupled with some paydowns of the underlying mortgages (inaudible) equates to the increase of the below-investment-grade of approximately $350 million.

Okay. The last question, looking at your list of the high-grade and mezzanine ABS, that nice chart you put together, I guess, just to be clear here, most of those are in synthetic I think you said, synthetic-type transactions?

We execute -- everything you see on that chart we executed in the credit default swap market.

Okay. Does that apply as an actual deal that, say, may have that, as you say here, AAA subordination of 8% or whatever? I don't know if it's a junior AAA or a mezzanine AAA or the senior AAA and someone then comes to you, or are these just hodgepodges of reference securities that you guys put together?

No, no, no, these are actual deals. In fact, everyone -- you don't see the names on the Web site but there's a name behind every one. For example, there are names you would probably recognize, that they issued a CDO. A bank or a funding vehicle will buy the AAA tranche (multiple speakers) super AAA tranche. Then they will come in and ask us to provide a CDS which hedges it. So let's say the AAA tranche is trading at LIBOR plus 30, they may say, okay, we are willing to pay Ambac -- we're willing to take less if Ambac comes in and wraps this AAA tranche for us. And these are actual deals.

I guess the question -- in those actual deals, the tranche that you're attaching, I know you have extra, extra subordinate enhancement where you attach to, but are these the senior-most tranches of those deals, or looking at the enhancement level would suggest to me probably it's the junior tranche.

No, no, no, these are the senior-most. What happens is let's say the AAA attachment point is 10%, and we want to attach at 20%. The banker has to go out and find somebody to take that; we call it a AAA mezzanine which is kind of a 10% to 20% attachment point. They have to go out and find somebody to buy that paper. So we would be senior to that.

I guess that's why I'm confused. I see these AAA subordination numbers that your list.

(multiple speakers)

That's the junior class, so you're not attaching this. So maybe the question is, in these high-grade [EBS], typically how many AAA-type classes are there?

There could be two to three, but in that, if you look at the very first one, you see where it's at 8%. That's what it takes, 8% subordination to get a AAA. We attach at 22%, which means there's 14% AAA below us that has to be placed out in the market.

Right. But typically that's senior class in that deal, which would be rated AAA. What would their subordination be? Obviously, the other AAAs would be enhancement for them. That's what I'm getting at.

Yes, there is -- in a typical deal, there could be two or three classes of AAA notes. We would attach at the most senior level.

In the fact that you had these very high numbers versus the junior AAA, does that mean, in your mind, the enhancement levels that are shown there, that you don't believe the rating agencies in saying that the junior AAA tranche -- I don't think -- as you say, the first one, 8% is really a AAA?

It's not that we don't believe the rating agencies; it's just that we have our capital at risk. So we want to push the market to as far as we can, and we run stress scenarios that we use to size our subordination level. So, we just want to make sure that we get enough subordination for what we are getting paid here to make the risk/return profile work for us.

Okay. Last question -- typically, the rating agencies look at AAA coverage anywhere from four to five times expected losses, and it varies by asset type. Would it be fair to say that your attachment points on these deals would be representative that you guys are looking at the similar four to five times your expected charge-offs, or are these numbers maybe indicating six, seven or even higher multiple times of what you think expected charge-offs are going to be?

It would be a higher multiple. On a base-case scenario, it would be higher than what you just put out. But it's all over the board, depending on asset class.

Okay, very good. Thank you.

[Helen Stewart], Russell Investment Group.

Thank you. Should there be losses arising out of CDS on CDOs, say, under a stress scenario? What would the claims development look like? In other words, would there be an acceleration of claims because there's CDS over, say, a few months or would it be spread out over a matter of years?

In our CDS contracts, what we do is restructure them to get as close to an insurance policy as we can. So we pay -- it's called "pay-as-you-go". There's two templates for pay-as-you-go but to keep it simple, we would pay scheduled principal and interest on the underlying CDO transaction. So we're not writing put options, in other words.

Okay, good. So for instance, and I suppose that stress scenario is somewhat equivalent on that whole 60 billion of, say, that 200 million. Again, I know that's tail risk but -- so what would be the time over which, though, that 200 million would be paid if that tail event happened?

When you say 200 million, are you talking about the capital implications of a downgrade?

I was just using that as a proxy, yes.

Well, that would just be -- that particular number was a number that's more of a rating risk, of a downgrade risk, not a (inaudible) type risk or any types of likelihood of paying a claim. You know, that particular number was an effort we undertook just to get the sensitivity of that book to the rating agencies' models for financial guarantors and the level of capital required for a AAA financial guarantor. (multiple speakers)

I see. But if there were claims, what would that term structure look like, that pay-as-you-go?

Yes, it would depend on if was a six or seven-year deal, or a ten-year deal. Generally, you would pay interest for seven or ten years, and then you would pay principal at the end. It would depend on the magnitude of defaults of the underlying. So you know, if 10%, if you are into your layer and then 10% more of the collateral [fails], that would kind of give you the relative size. So, it would depend on the ultimate size of the underlying defaults, and on the original nature of the deal, again if it was a seven-year deal or a ten-year deal or a fifteen-year deal.

Okay. Anyway, the bottom line is even though they are CDS, they aren't accelerated?

Correct. Two things we do is when we modify the (inaudible) is we try to take out the acceleration risk, and we take out the collateral posting risk.

Nandu Narayanan, Trident Investment Management.

My question really relates more to just the correlations across the various risks that you in sure. Because I'm looking at the CDO portfolio, which is quite large. At the same time, you also have other guarantees in force as well. Obviously, given what's going on in the housing market, yesterday I think Countrywide commented that this is about the worst conditions they've seen since the great Depression pretty much, and the possibility also that things may get dramatically worse in the credit market, given the fact that foreclosures in housing are already at very high levels despite the fact that we're not in a recession.

How exactly do you capture the possibility of all of your credit insurance being correlated going forward? Specifically because you also said simultaneously that you weren't making many changes in your models over the last three to six months. How do you capture the possibility that things may deteriorate? I would like to get a sense of what your assumptions are in terms of what might be going on in the credit market and what you expect.

Sure, let me -- I will touch on CDOs and then we can talk about the general capital modeling as well.

In the CDOs, what we think we've done, relative to the market, and (inaudible) interested to hear this, but we think that our base case is already somewhat stress correlation, so when we underwrite a transaction, we're not underwriting based upon correlations that, let's say, a trader would use to buy a particular securitization, a CDO. We are underwriting based on correlations that are already higher than what we think the market is doing. But whether you agree with us or not there, I think, if you were on the call earlier, you may have heard me say our base case correlation assumption on an ABS CDO is about 30%. When we underwrite on a stress scenario, which is also presented to our credit committee, we use a correlation scenario of 60%, so we double the base case scenario.

So, we think that we underwrite to a fairly sensible correlation level, so then when the credits do deteriorate, which we know correlation increases when that happens, we think we are in pretty good shape. We can talk a little bit -- Sean -- about our overall capital (inaudible).

Yes, from a macro perspective, just looking at the portfolio -- two ways I guess the mortgages in general. One is, from a credit risk standpoint and the nature of risk, and obviously you're looking towards a diversity of the underlying pools and sculpting the pools as you think necessary to diversify out your exposures to this particular asset class.

Also, what we're doing is we are doing internal capital modeling that's a dynamic process. We run all the agency models and the agency models have some pretty harsh statistics in those as well for the financial guarantors. But we also run an internal model where we come up with our own correlation levels and we're able to do stress scenarios based upon that from the standpoint of what potentially could be our worst-case loss, and obviously still support our AAA rating. So the correlation -- Tom talks about the correlation in a particular deal. We also step back and look at it from a macro basis as well as the agencies looking at it as well from a macro basis.

Just given the current conditions in housing, because obviously if most of the housing reports that we are reading are suggesting that we're sort in a 50-year downturn in the market or something like that, or actually more than that, about a 70-year event in housing market, how exactly are you capturing that in your thinking right now? Because obviously the issue that we're struggling with is that the subprime adjustment that we're seeing right now in all of these ABS markets may not necessarily be technical; it may actually be real. Because it looks like, under a variety of scenarios with home price declines, that you might actually see most of these AAA-, BBB or even A tranches completely wiped out with a possibility of credit losses going although it up into AA or AAA, fairly significant losses at that, because we see also that, in foreclosures now, some of the builders are even projecting that 50% losses on foreclosures or on the inventory write-downs they're taking are very large.

Yes, well, (inaudible) said, we kind of agree that, in particular with 2006, and that's how we underwrote the deals. I said earlier we use actually a 60% severity, not a 50% that you just said. We use a 60% severity assumption in our models for CDOs, and then in our stress case, we actually use an 80% severity in there. So you know, we can't underwrite to how the market was originating these deals because our capital is at risk. So we have to underwrite to a stress scenario and we used a similar scenario for our direct RMBS.

I will say, too, that the portfolio is, as we've been talking about, the recent vintage portfolio, particularly in subprime, we've pulled back on that. Just broadly, though, the ratings in the mortgage space has declined from higher levels, so you have a bit of a seasoning in the portfolio. It's not just concentrated in recent vintages.

To give you a sense for some of the underlying RMBS, we typically -- we've looked at, in our base case modeling, in order to come up with a base-case model loss on the transaction, typically looking at a price decline of 33%, which dates back to some levels that was experienced in the oil patch crisis and the like. So we take that to a model loss and then we are attaching, even at the BBB scenario, we are attaching it at greater than eight points. So two times coverage, over two times coverage. So there's quite a bit of stress already built into that, not to say that some transactions might perform more poorly, but that gives you a sense for how drastic our modeling is.

At this time, I'm showing no further questions in queue. I'd like to turn the call back over to management.

Well, thank you very much. Thank you for participating in our second-quarter conference call. We look forward to answering any questions that you may have. As we mentioned a couple of times during the call, we believe we've enhanced our [red plate] disclosures and we would certainly encourage people to look to our Web site for any additional information. Thank you very much.

This concludes today's teleconference. Thank you for your participation.