December 3, 2010
Mr. Jim B. Rosenberg
Senior Assistant Chief Accountant
United States Securities and Exchange Commission
Division of Corporation Finance
100 F Street, N.E.
Washington, D.C. 20549
Re: | Ambac Financial Group, Inc. | |
Form 10-K for the fiscal year ended December 31, 2008 | ||
Filed on March 16, 2009 | ||
Form 10-Q for the quarter ended June 30, 2009 | ||
Filed on August 10, 2009 | ||
File Number: 001-10777 |
Dear Mr. Rosenberg:
This letter is submitted on behalf of Ambac Financial Group, Inc. (“Ambac”) in response to comments received from the Securities and Exchange Commission staff (“Staff”), in a letter dated October 13, 2010 (the “Letter”), with respect to Ambac’s Form 10-K for the fiscal year ended December 31, 2008 filed on March 16, 2009 (“2008 Form 10-K”) and Ambac’s Form 10-Q for the quarter ended June 30, 2009 filed on August 10, 2009(“June 30, 2009 Form 10-Q”). The Letter is a follow up to our response letter dated July 1, 2010, with respect to Ambac’s 2008 Form 10-K and June 30, 2009 Form 10-Q.
Ambac appreciates the efforts of the Staff in this review process. Enhancement of the overall disclosures in our filings is an objective that we share with the Staff and one that we continuously consider in our filings. In connection with responding to your comments, we acknowledge that Ambac is responsible for the adequacy and accuracy of the disclosures in our filings; that SEC staff comments or changes to disclosure in response to SEC staff comments do not foreclose the SEC from taking any action with respect to the filing; and that Ambac may not assert SEC staff comments as a defense in any proceeding initiated by the SEC or any person under the federal securities laws of the United States.
Securities and Exchange Commission Staff Comments:
Form 10-Q for the Fiscal Quarter ended June 30, 2009
Item 2. Management’s Discussion and Analysis of Financial Condition and Results of Operations
Critical Accounting Policies and Estimates
Borrower Default Burnout, page 65
1. | On page three of your February 8 response, you state that “We believe using this single realization factor is equivalent to the results which would be obtained using the methodology shown in the example in ASC 944-40-30-33.” Please address the following: |
• | Please confirm our understanding that you use the same realization factor for every contract. |
• | If you use the same realization factor for all contracts, will the outcome of your methodology equal the outcome of the probability weighted outcome approach in ASC 944-40-30-32 for all contracts if the possible cash flow scenarios for each contract donot bear the same proportional relationship to the cash flow scenario (i.e. all breached loans |
December 3, 2010
Page 2
will be repurchased) for each contract from which the realization factor is calculated. If yes, please provide a quantified example that contains multiple contracts demonstrating your conclusion. Please also describe the circumstances when this mathematical outcome would be likely to occur and those when it would be unlikely to occur. |
Ambac response:
Regarding the first bullet point, we confirm that we use the same realization factor for every contract for which we record a subrogation recovery.
Regarding the second bullet point, subsequent to receiving the SEC comment letter of October 13, 2010 the Staff provided to Ambac a separate page of numerical examples entitledExamples of Probability Weighting/Realization Factors to help clarify the above question for us. There were four securitization examples on that page, with each securitization having 3 different cash flow scenarios. As we understand the Staff’s example, the “recovery as a percentage of maximum recovery” amount shown in each securitization example represents each cash flow scenario’s proportional relationship to cash flow scenario #3 which, as we understand it, represents all breached loans being repurchased. We believe the Staff’s example already demonstrates it is theoretically possible that securitizations with different proportional cash flow scenarios may result in the same realization factor, as illustrated in Securitizations #1 and #4. More specifically, the “recovery as a percentage of maximum recovery” is 27.0%, 70.0% and 100.0% for Securitization #1 and 35.5%, 60.5% and 100.0% for Securitization #4 under Scenarios 1, 2 and 3 respectively, yet both securitizations result in the same realization factor. The Staff’s example also demonstrates it is theoretically possible that securitizations with different proportional cash flows may result in different realization factors, as illustrated in Securitizations #1 and #3.
However, in developing our subrogation recovery estimate, we do not go through the process of determining whether the proportional relationship among possible recovery cash flow scenarios for each transaction are, in fact, different, and whether such a difference might warrant using different realization factors for various transactions. Rather, as previously discussed in our February 8, 2010 response, we develop a range of realization factors based on various assumptions about potential outcomes and qualitatively evaluate that range in determining a single realization factor which is then applied to all of our transactions for which a subrogation recovery is recorded. We believe using the same realization factor for each transaction is appropriate at this stage because the settlement negotiation and/or litigation with the various sponsors related to all of our transactions is not sufficiently advanced enough to make different assumptions about the range of realization factors and associated probabilities for individual transactions. We do believe as the settlement negotiations and/or litigation progresses, differences in facts and circumstances may arise among transactions, in which case we may need to modify our methodology to calculate a probability-weighted transaction-specific realization factor.
2. | For many of your insurance contracts your subrogation recovery equals the sum of “ever-to-date paid losses plus the present value of projected future paid losses” for the contract, which also appears to be the maximum amount you are seeking in your lawsuits. For these contracts, the Company’s methodology does not apply a realization factor to this maximum. Accordingly, for these contracts it does not appear that the Company has considered the uncertainties surrounding the settlement negotiation and litigation processes. Please advise. |
December 3, 2010
Page 3
Ambac response:
Ambac is not exclusively pursuing recoveries for the sum of our ever-to-date paid losses plus the present value of projected future paid losses (collectively referred to as “ever-to-date incurred losses”). Rather, consistent with its right to pursue multiple legal theories and bases of recovery that maximize the relief to which it is legally entitled, Ambac is pursuing several alternative bases for recovery. This includes, in all cases, the enforcement of the sponsors’ contractual obligation to repurchase all loans which have breached representations and warranties and are required to be repurchased pursuant to the terms of the transaction documents. As we have disclosed in previous SEC financial statement filings, loans repurchased by the sponsor result in cash inflows into the securitization trust. In accordance with the transaction’s contractual waterfall, the cash from loan repurchases is aggregated with other cash collections from the underlying mortgages and is first used to pay down outstanding bondholders, pay various expenses of the trust and then to reimburse Ambac for all prior claim payments made to the extent bonds have been paid off. To the extent there is cash remaining in the securitization trust after all bonds have been repaid and Ambac has been fully reimbursed, that cash will be paid to junior stakeholders, according to the terms of the waterfall. Thus, the financial benefit to Ambac from sponsor repurchases will never exceed Ambac’s ever-to-date paid losses plus actual future losses plus interest and expenses of enforcement. As a result, the actual subrogation recovery amount that is used in the probability-weighted expected net cashflow scenarios to compute the net loss reserve is the lesser of i) the estimated subrogation recovery (which is the sponsor’s estimated repurchase obligation after applying the realization factor) or ii) our ever-to-date incurred loss for each transaction.
For the computation of second-lien residential mortgage-backed securitizations (“RMBS”) loss reserves, we utilize a roll-rate methodology as further described in theCritical Accounting Policies and Estimates - RMBS section of our SEC filings. This approach assigns a probability weight to various claim outflow scenarios, which have different input assumptions related to individual transactions’ cumulative prepayment rates, cumulative default rates and loss severities. We then incorporate the estimated subrogation recovery inflow into each claim outflow scenario to compute our net loss reserve. Below is a simplified illustration of our process, which ignores the effects of discounting:
Scenario 1 | Scenario 2 | Scenario 3 | Probability- weighted net loss reserve | |||||||||||||||||
Actual ever-to-date (“ETD”) claim payments (historical) | (a | ) | $ | (100 | ) | $ | (100 | ) | $ | (100 | ) | |||||||||
Estimated gross loss reserves (future expected losses) | (b | ) | $ | (50 | ) | $ | (70 | ) | $ | (30 | ) | |||||||||
ETD incurred losses (“cap”) | (c | ) | $ | (150 | ) | $ | (170 | ) | $ | (130 | ) | |||||||||
Estimated subrogation recovery | (d | ) | $ | 150 | $ | 150 | $ | 150 | ||||||||||||
Net estimated future cash flow | (e | ) | $ | 100 | $ | 80 | $ | 100 | ||||||||||||
Probability | (f | ) | 60 | % | 20 | % | 20 | % | ||||||||||||
Probability-weighted cash flow | (g | ) | $ | 60 | $ | 16 | $ | 20 | $ | 96 |
(a) | – Represents actual ETD claim payments made and therefore does not vary by scenario. |
(b) | – Estimated gross loss reserves, i.e. future expected cash outflows prior to subrogation recoveries. |
(c) | – Sum of actual ETD claim payments (a) plus estimated gross loss reserves (b). This number represents the maximum amount (or “cap”) we will include as a recovery in our loss reserve calculation. Refer to item (e) below for further discussion. |
(d) | –This is a single number we calculate as we have previously described in our disclosures and previous comment letter responses and does not vary by scenario. |
(e) | – Represents the estimated gross loss reserve (b) plus the lesser of i) the absolute value of ETD incurred losses (c) or ii) estimated subrogation recovery (d). |
(f) | – Probability weight as determined by the surveillance analyst and approved by management. |
(g) | –Net estimated future cash flow (e) multiplied by Probability (f). |
December 3, 2010
Page 4
In the above illustration the probability-weighted net loss reserve is a $96 net cashinflow and thus would be recorded as a subrogation recoverable asset on our balance sheet. This situation exists in the above example because a large portion of estimated subrogation recoveries relates to actual ETD claims previously paid. If the probability-weighted net loss reserve was a net cashoutflow, it would be recorded as a loss and loss expense liability on our balance sheet.
For the computation of first-lien RMBS loss reserves, we utilize a third-party Monte-Carlo cash flow model which generates 300 claim outflow scenarios, as described in the Critical Accounting Policies and Estimates - RMBS section of our SEC filings. We assign the same probability to each scenario (as is typical of a Monte Carlo process), and therefore the average of the 300 scenarios represents the estimated gross loss reserve (prior to incorporating subrogation recoveries). We then incorporate the subrogation recovery to compute net loss reserves for first-lien transactions by performing the same calculation as shown above for second-lien transactions. However, this calculation is only performed once for theaverage of the 300 scenarios, rather than performing separate calculations for all 300 scenarios due to cost-benefit constraints. We do not believe the computed net loss reserve amount would be materially different had we performed separate calculations for all 300 scenarios.
We believe the above described approaches are consistent with the guidance in ASC 944, with respect to recording claim liabilities based on probability-weighted net cash flow scenarios.
Furthermore, it would be inappropriate to apply the realization factor to our ever-to-date incurred losses, since the ever-to-date incurred loss amount is not the starting point from which we are pursuing recoveries. We believe our current approach of applying the realization factor to the contractual amount the sponsors are required to repurchase, which is one of the outcomes we are pursuing in all settlement negotiations and/or litigation, is the appropriate amount to which the realization factor should be applied, provided we limit (or cap) the subrogation recovery we record to the ever-to-date incurred loss.
As the Staff points out in the above comment, we are effectively recording a subrogation recovery amount equal to our ever-to-date incurred losses for some of our transactions. This circumstance is the result of the extremely high breach rates coupled with the significant credit enhancement that absorbs losses prior to Ambac in these transactions. However, we would also note that there are a significant number and dollar amount of transactions where we are recording a subrogation recovery that is actually less than our ever-to-date incurred losses. At the Staff’s request (made verbally on November 29, 2010) we are providing the table below as of September 30, 2010, which breaks out the subrogation recoveries by dollar amount and number of transactions (“count”) where the estimated subrogation recovery is: i) greater than ever-to-date incurred losses for all cash flow scenarios (and therefore we cap the recovery to the ever-to-date incurred loss amount), ii) less than ever-to-date incurred losses for all cash flow scenarios and iii) a mixture of greater than and less than ever-to-date incurred losses for the various scenarios within the same transaction.
December 3, 2010
Page 5
9/30/10
($ in millions) | Subrogation greater than incurred – all scenarios | Subrogation less than incurred – all scenarios | Subrogation greater or less than incurred- mixed scenarios | Total | ||||||||||||||||||||||||||||
Dollars | Count | Dollars | Count | Dollars | Count | Dollars | Count | |||||||||||||||||||||||||
Random sample: | ||||||||||||||||||||||||||||||||
2nd lien | $ | 228.2 | 2 | $ | 1,048.0 | 4 | $ | 91.0 | 1 | $ | 1,367.2 | 7 | ||||||||||||||||||||
1st lien(a) | 219.2 | 4 | 126.4 | 1 | — | — | 345.6 | 5 | ||||||||||||||||||||||||
Total | 447.4 | 6 | 1,174.4 | 5 | 91.0 | 1 | 1,712.8 | 12 | ||||||||||||||||||||||||
Adverse sample: | ||||||||||||||||||||||||||||||||
2nd lien | — | — | 561.0 | 11 | $ | 11.1 | 1 | 572.1 | 12 | |||||||||||||||||||||||
1st lien(a) | 16.5 | 1 | 119.9 | 2 | — | — | 136.4 | 3 | ||||||||||||||||||||||||
Total | 16.5 | 1 | 680.9 | 13 | 11.1 | 1 | 708.5 | 15 | ||||||||||||||||||||||||
Total | $ | 463.9 | 7 | $ | 1,855.3 | 18 | $ | 102.1 | 2 | $ | 2,421.3 | 27 | ||||||||||||||||||||
(a) | As discussed above, for first-lien transactions the above analysis represents a comparison of the estimated subrogation recovery to the incurred loss average of the 300 scenarios (i.e. a single calculation), rather than comparing estimated subrogation recovery to each separately calculated incurred loss for all 300 scenarios. |
Follow-up point raised by the Staff:
We would also like to address a separate point raised by the Staff in our verbal discussions about how we calculate our estimated subrogation recovery. The point, as we understand it, is that because we are applying the breach rate to the sum of ever-to-date collateral losses plus the current unpaid loan pool balance (“CULPB”) in the calculation of our subrogation recovery, and given the theoretical possibility the CULPB may include performing loans, we potentially could be overstating our subrogation recovery. In other words, since a performing loan should have no impact on future Ambac claim payments, then a sponsor repurchase of a performing loan should have no impact on future Ambac recoveries and therefore why should we record such a recovery. We acknowledge this point. We believe our current methodology is appropriate because the breach rate itself was derived from a sample ofall the original loans in the collateral pool and because the ever-to-date losses plus the CULPB is representative of theentire collateral pool, and therefore it would seem this is the appropriate population to which we should apply the breach rate. In other words, it is an “apples to apples” approach.
Furthermore, it has not been possible to accurately derive a breach rate with respect to future non-performing loans in a collateral pool because it is not possible to identify a statistically valid sample of future non-performing loans. Although we do project our future losses on the current collateral pool in order to develop our loss reserves, this estimation process does not, and cannot, identify whichspecific loans in the collateral pool will be non-performing in the future.
Nonetheless we attempted to alleviate the Staff’s concern by providing an “Alternative View” with the Ambac example we discussed with the Staff on November 18, 2010. Ambac’s example was based on actual data for a transaction where we currently record subrogation recoveries. In this “Alternative View” calculation we applied the breach rate (calculated fromall the original loans in the pool) to the sum of the ever-to-date collateral losses plus the estimated future collateral pool losses (representing the aggregate non-performing loans in the pool), in order to calculate a hypothetical subrogation recovery for this transaction. As we would expect the breach rate for non-performing loans (if such a calculation were possible) to be higher than the breach rate for all the original loans in the pool, we believe the “Alternative View” calculation is a conservative way to analyze our subrogation recovery. That said, we also believe this “Alternative View” is not theoretically correct because the breach rate was derived from a sample ofall the original loans in the collateral pool and is being applied to only the non-performing loans in the collateral pool, i.e. this is an “apples to oranges” approach. Nonetheless, this “Alternative View” calculation demonstrates that our current approach for estimating subrogation recoveries for the transaction example provided produces the same result as a more conservative calculation. We also believe that, with respect to all of our
December 3, 2010
Page 6
remaining transactions for which we have recorded subrogation recoveries, applying our current approach would not produce a materially different subrogation recovery result than an alternative approach (if one were possible) that derives and applies a breach rate from a population of only non-performing loans.
Please feel free to contact me at (212) 208-3393 or Richard Alger at (212) 208-3196 should you require further information or have any questions.
Sincerely, |
/s/ Robert Eisman |
Robert Eisman |
Senior Managing Director and |
Chief Accounting Officer |
Copy to: | David Wallis | |
President and Chief Executive Officer | ||
Ambac Financial Group, Inc. | ||
Kevin J. Doyle, Esq. | ||
Senior Vice President and General Counsel | ||
Ambac Financial Group, Inc. | ||
David Trick | ||
Senior Managing Director and Chief Financial Officer | ||
Ambac Financial Group, Inc. | ||
Michael Groll, Esq. | ||
Dewey & LeBeouf LLP | ||
Richard Spitzer, Esq. | ||
Dewey & LeBeouf LLP | ||
Joel Parker | ||
Accounting Branch Chief | ||
Securities and Exchange Commission | ||
Paul Laurenzano | ||
Partner | ||
KPMG |