September 17, 2008
Mr. Jim B. Rosenberg
Senior Assistant Chief Accountant
Division of Corporation Finance
United States Securities and Exchange Commission
100 F Street, N.E.
Washington, D.C. 20549
| Re: | Radian Group Inc. | |
| Form 10-K for the Fiscal Year Ended December 31, 2007 | ||
| Form 10-Q for the Fiscal Quarter Ended June 30, 2008 | ||
| File Number: 001-11356 |
Dear Mr. Rosenberg:
We have considered carefully each of the comments in your letter dated August 22, 2008 as well as our discussions with the staff over the past several weeks. On behalf of Radian Group Inc., I respectfully provide our responses to these comments below. For your convenience, the text of each comment is reproduced below before the applicable response. References in this letter to “we,” “us,” “our,” or “ours” refer to Radian Group Inc. and its consolidated subsidiaries.
Form 10-Q for the Fiscal Quarter Ended June 30, 2008
Note 2 – Fair Value of Financial Instruments, page 6
| 1. | We note your response to prior comment one of our letter dated July 25, 2008. We are continuing to evaluate your response in the context of the following comments relating to your disclosure. Accordingly, please provide proposed revised disclosure which addresses the comments below. |
Response: Our revised disclosure is set forth in Appendix A to this letter. Below we have provided either specific responses to your comments or a description of where the requested information exists in Appendix A.
Derivative Assets and Liabilities, page 80
| 2. | You disclose that your five year credit-default swap spread widened by 2,493 basis points to 2,530 basis points from January 1, 2007 to June 30, 2008, and that incorporating your own non-performance risk into the valuation of your credit derivatives reduced the fair value of these derivative liabilities by $2.1 billion during the three months ended March 31, 2008, and $2.0 billion for the six months ended June 30, 2008. With the objective of providing enhanced disclosure for investors, please revise your disclosure to include: i) the changes in your credit default swap spread by quarter, and ii) of the initial $2.1 billion effect of the incorporation of your non-performance risk into the valuation of your credit derivatives, how much of this amount related to changes in non-performance risk that occurred prior to the adoption of SFAS 157, and how much of this amount related to changes in non-performance risk that occurred during the initial quarter of adoption. |
Response:We will revise our disclosures to incorporate this information in our future filings beginning with our September 30, 2008 10-Q. In addition, we have included a table at the end of Appendix A that contains this information as of June 30, 2008.
| 3. | As discussed in our recent conference calls, we continue to believe that your disclosures regarding the valuation of each major category of derivative assets and liabilities should be enhanced so that the average investor can obtain a basic understanding of how the value of each material category of derivative asset or liability is determined. While we acknowledge that the valuation of these instruments may be complex, we do not believe such complexities obviate the need to provide readable and understandable disclosure. Accordingly, while more detailed comments are provided below specifically with respect to the category of Corporate CDOs, we believe that your disclosures for each material type of derivative asset or liability should be revised to provide a similar level of readability and understandability. |
Response: Our revised disclosure regarding the valuation of our Corporate and Non-Corporate CDOs is included in Appendix A to this letter.
| 4. | For each category where your valuation model results in the determination of the entry price that another market participant of similar credit quality would pay, please disclose how you ensure that the entry price you derive is within the range of what other market participants would pay. |
Response: As described in our June 30, 2008 10-Q on page 80, there is no principal market we can observe given our business model, the legal form and restrictions inherent in our contracts, and the current market environment. As such, it is not possible to validate definitively our exit price estimates via actual transfers or current market transactions (e.g., the recent sales by monoline market participants of credit protection on structured finance arrangements). In our assumed hypothetical exit market, we have utilized relevant market-based inputs and assumptions that are observable where possible, as we believe that our hypothetical counterparties would use similar inputs to determine exit price. Where market-based inputs or assumptions are not available, we have utilized inputs that are based on our models. Based on our experience in the monoline sector and the relatively small number of sector participants, we believe our assumptions and judgments are consistent, in all material respects, with those of our hypothetical market participants such that our fair value estimates are consistent with the fair value estimates of our hypothetical market participants. When there was an active market, our fair valuation estimates using these methodologies were validated through observed trade prices.
| 5. | We note various references to the valuation of derivative assets and liabilities using dealer quotes or fair value measurements by the transaction counterparties. Please revise your disclosure to clarify the following: |
| • | Whether you adjust the dealer quotes and/or fair value measurements, and if so, how you determine the adjustments; |
| • | The procedures performed to validate the amounts obtained to ensure the fair value determination is consistent with the requirements of SFAS 157. |
Response:We do not utilize dealer or counterparty quotes for valuing our Corporate and Non-Corporate CDO transactions as described in our revised disclosures in Appendix A.
We use fair value measurements or quotes obtained from our counterparties in making our own determination of the estimated fair value of assumed financial guaranty contracts and Mortgage Insurance Domestic and International CDS contracts. Our process for validating and, if necessary, adjusting these measurements or quotes is set forth on pp 83-84 of our June 30, 2008 10-Q.
Corporate CDOs, page 82
| 6. | You disclose that you use market spreads identified through the credit default swap (“CDX”) index, which is an industry standard credit default swap index. Please revise your disclosure to clarify whether observable market prices are obtained from the CDX index. |
Response:See “Defining the Equivalent-Risk Tranche” under “Corporate CDOs” in our revised disclosure in Appendix A.
| 7. | Your disclosure includes various references to collateral which appear to have different meanings. Please clarify your disclosure to distinguish between these references including “individual issuers in our collateral pool,” “standard CDX collateral,” and the requirements for other market participants to post collateral. |
Response:We have replaced references to “collateral” in our revised disclosure with more specific descriptions relating to the referenced entities included in our transactions and the various indices we use for valuing our transactions. See our revised disclosure in Appendix A.
| 8. | Please revise your disclosure to clarify how you use the credit spreads on individual issuers to derive an equivalent-risk tranche on the CDX index. Please address how the Monte Carlo simulation referred to in the August 15, 2008 conference call is used, the significant inputs into this model, whether such inputs are observable or unobservable, and how the probability and severity of loss are considered in your analysis. |
Response: See “Defining the Equivalent-Risk Tranche” under “Corporate CDOs” in our revised disclosure in Appendix A for a description of how we use credit spreads on individual issuers to derive an equivalent-risk tranche on the CDX index. The simulation process we use is described in this section of our revised disclosure. Probability of default and severity of loss are addressed in our model primarily through the use of observed credit spreads, which incorporates these individual assumptions. Accordingly, our disclosures describe the use of credit spreads as a significant input.
| 9. | You disclose on page 16 that the credit protection you provide on corporate CDOs related primarily to senior investment grade debt. Please clarify whether you use the same valuation model for investment grade and non-investment grade CDOs, and if not, revise your disclosure accordingly. |
Response:The same valuation model is used for all of our Corporate CDO transactions.
| 10. | Please clarify your disclosure regarding the nature of adjustments for other factors considered before determining the CDOs equivalent-risk tranche on the CDX Index. Explain the relative significance of these adjustments to the resulting valuation of the CDOs. |
Response:See “Defining the Equivalent-Risk Tranche” under “Corporate CDOs” in our revised disclosure in Appendix A for a description of all significant adjustments utilized in our fair value determination as well as the significance of these adjustments in the resulting valuation.
Non-Corporate CDOs:
| 11. | You disclose non-corporate CDO liabilities of $302 million at June 30, 2008, representing approximately 46% of your total derivative liabilities. You have also stated that each type of non-corporate CDO includes unique features that make the valuation of these instruments complex. Please consider whether disaggregated disclosures for each material category of non-corporate CDOs would enable investors to more clearly understand the material terms, risks and valuations of each of these instruments, and revise your disclosures as appropriate. |
Response:See “Non-Corporate CDOs” in our revised disclosure in Appendix A in which we disaggregate our disclosure for each material category of our Non-Corporate CDO transactions. In addition, we will incorporate a quantitative table into our future filings beginning with our 10-Q for the quarter ended September 30, 2008, that specifies our notional exposure and market values for each material category of Non-Corporate CDO transactions.
| 12. | Please clarify whether your references to dealer quotes in this section represent quotes for the fair premium or quotes for the underlying CDOs or other ABS. If the latter, please expand your disclosure to explain how you internally validate the quotes and determine the fair premium based on this information. |
Response:As addressed in comment 5, we do not utilize dealer quotes for valuing our Non-Corporate CDO transactions. Instead, we use observed trade spreads for valuing certain of these transactions. These observed trade spreads are on transactions that are similar to the underlying ABS deals we insure, and do not represent spreads on CDS contracts that insure those risks. We believe, however, that the observed trade spreads of the underlying transactions represent the most relevant basis for our fair values.
Risk Factors, page 93
| 13. | On pages 98-99 you disclose that your insurance contracts with ceding insurers and credit default swaps contain adjustable features and termination rights triggered by the downgrade of Radian, under which the counterparties may recapture previously ceded business and/or elect to charge additional ceding commissions or terminate the credit default swap agreements. Please disclose the adjustable features and termination rights of these contracts and provide us your analysis determining whether or not the contracts contain embedded derivatives under paragraphs 12-16 of SFAS 133. |
Response:We evaluated the adjustable features and termination rights of these contracts and determined that while they do contain embedded derivatives, they are not required to be bifurcated. All adjustable features and termination rights of these contracts are triggered by a downgrade of the financial strength ratings of our principal financial guaranty subsidiary by S&P and/or Moody’s credit rating
agencies and all adjustable features are limited to those disclosed on pages 98 and 99 of our June 30, 2008 10-Q filing. We compared these features to those discussed in paragraph 61 (c) of SFAS 133 and determined that while these features are embedded derivatives, our creditworthiness is clearly and closely related to the economic characteristics and risks of the insurance contracts. Therefore, these features should not be bifurcated and accounted for separately. The credit default swaps already accounted for as derivatives under SFAS 133 are not required to be evaluated for embedded derivatives.
In connection with the responses to the comments set forth above, we acknowledge that:
| • | We are responsible for the adequacy and accuracy of the disclosure in our filings; |
| • | Staff comments or changes to disclosure in response to Staff comments do not foreclose the Securities and Exchange Commission from taking any action with respect to the filings; and |
| • | We may not assert Staff comments as a defense in any proceeding initiated by the Securities and Exchange Commission or any person under the federal securities laws of the United States. |
If you have any questions about any of the responses to your comments or require further explanation, please do not hesitate to contact me at (215) 231-1444.
| Sincerely, |
/s/ C. Robert Quint |
| C. Robert Quint |
| Chief Financial Officer |
Appendix A – Proposed Revised Disclosure
Corporate CDOs
The fair value of each of our Corporate CDO transactions is estimated based on the difference between (1) the present value of the future contractual premiums we charge and (2) the present value of our estimate of the future premiums that a financial guarantor of similar credit quality to us would charge (we refer to these premiums as the “fair premium”) to provide the same credit protection assuming a transfer of our obligation to such financial guarantor as of the measurement date. The discount rate used to estimate the present value of our contractual premium and fair premium is a risk-free rate plus our own credit default swap spread as of the measurement date.
For each Corporate CDO transaction, we perform three principal steps in determining the fair premium amount:
| • | First, we define a tranche on the CDX index (defined below) that equates to the risk profile of our specific transaction (we refer to this tranche as an “equivalent-risk tranche”); |
| • | Second, we determine the fair premium on the equivalent-risk tranche for those market participants engaged in trading on the CDX index (we refer to each of these participants as a “typical market participant”); and |
| • | Third, we adjust the fair premium for a typical market participant to account for the difference between the non-performance or default risk of a typical market participant and the non-performance or default risk of a financial guarantor of similar credit quality to us (in each case, we refer to the risk of non-performance as “non-performance risk”). |
Defining the Equivalent-Risk Tranche
Direct observations of fair premiums for our transactions are not available since these transactions cannot be traded or transferred pursuant to their terms and there is currently no active market for these transactions. However, credit default swaps on tranches of a standardized index (the “CDX index”) are widely traded and observable, and provide relevant market data for determining the fair premium of our transactions, as described more fully below.
The CDX index is a synthetic CDO that is comprised of a list of corporate obligors and is segmented into multiple tranches of synthetic senior unsecured debt of those obligors ranging from the equity tranche (i.e., the most credit risk or first loss position) to the most senior tranche (i.e., the least credit risk). We refer to each of these tranches as a “standard CDX tranche.” A tranche is defined by an attachment point and detachment point, representing the range of portfolio losses for which the protection seller would be required to make a payment.
Our Corporate CDO transactions possess similar structural features to the standard CDX tranches, but often differ with respect to certain of the referenced corporate entities, term, and attachment and detachment points. Therefore, in order to estimate the equivalent-risk tranche for each of our Corporate CDO transactions, we first determine the attachment and detachment points on the CDX index that have the same estimated probability of loss as the attachment and detachment points in our transactions. We begin by performing a simulation analysis of referenced entity defaults in our transactions to determine the probability of portfolio losses exceeding our attachment and detachment points. The referenced entity defaults are primarily determined based on the following inputs: the market observed credit default swap credit spreads of the referenced corporate entities, the estimated correlations between each of the referenced corporate entities, and the term of the transaction.
For each referenced corporate entity in our CDO transactions, the credit default swap spreads associated with the term of our transactions (“credit curve”) define the estimated expected loss for each entity. The credit curves on individual referenced entities are generally observable. The expected cumulative loss for the portfolio of referenced entities associated with each of our transactions is the sum of the expected losses of these individual referenced entities. With respect to the correlation of losses across the underlying referenced entities, two obligors belonging to the same industry or located in the same geographical region are assumed to have a higher probability of defaulting together (i.e., they are more correlated). An increase in the correlations between the referenced entities generally causes a higher expected loss for the portfolio associated with our transactions. The estimated correlation factors that we use are derived internally based on observable third-party inputs that are based on historical data. These inputs are adjusted to levels implied by observed market pricing on the standard CDX index. The impact of our correlation assumptions currently do not have a material effect on our fair premium estimates in light of the significant impact of our non-performance risk adjustment as described below.
Once we have established the probability of portfolio losses exceeding the attachment and detachment points in our transactions, we then use the same simulation method to locate the attachment and detachment points on the CDX index with the same probabilities. These equivalent attachment and detachment points define the equivalent-risk tranche on the CDX index that we use to determine fair premium.
Determining the Typical Fair Premium
The equivalent-risk tranches for our CDO transactions often are not identical to any standard CDX tranches. As a result, fair premium rates generally are not directly observable from the CDX index for the equivalent-risk tranche and must be separately determined. We make this determination through an interpolation in which we use the observed premium rates on the standard CDX tranches that most closely match our equivalent-risk tranche to derive the premium rate for the equivalent-risk tranche.
Non-Performance Risk Adjustment
The fair premium rate estimated for the equivalent-risk tranche represents the premium rate for a typical market participant – not Radian. Accordingly, the final step in our fair value estimation is to convert this fair premium rate into a fair premium for a financial guarantor of similar credit quality to us. A typical market participant is contractually bound by a requirement that collateral be posted regularly to minimize the impact of this participant’s default or non performance. This makes these transactions less risky to the protection buyer, and therefore, priced differently. Our contracts do not require us to post collateral with our counterparties, which exposes our counterparties fully to our non-performance risk. We make an adjustment to the fair premium amount for a typical market participant to account for this contractual difference, as well as for our perceived default probability which is observable through our credit default swap spread. The amount of the adjustment is computed based on the correlation between the default probability of the transaction and our default probability. Our non-performance risk adjustment currently has a material impact on our fair premium estimates. The impact for each reported period is contained in a table presented on page x. - [Note - this table will be presented at the end of our critical accounting policy discussion on derivatives – see below].
The estimated fair value of our Corporate CDO transactions is reasonably likely to change significantly with changes in the observed standard CDX index spread. As of June 30, 2008, a 25 basis point change in the spread on an equivalent-risk tranche of the CDX index would have resulted in a change of approximately $6 million to our estimated current fair value asset of $65 million, assuming our own credit default swap spread remained constant. As of June 30, 2008, a 1,000 basis point widening or tightening in our credit default swap spread, assuming the CDX spread remained constant, would have resulted in an immaterial change to our fair value. Because our credit default swap spread is at such a wide level, the current fair value amount is relatively insensitive to reasonably likely changes in our credit default swap spread. These credit derivatives are categorized in Level III of the fair value hierarchy.
Non-Corporate CDOs and Other Derivative Transactions
The estimated fair value of our Non-Corporate CDO transactions relates to our guaranty of RMBS CDOs, CMBS CDOs, Trust Preferred securities (“TRUPS”) CDOs, financial guarantees of auto loans and credit card securities, and CDOs backed by other asset classes. Similar to our methodology for our Corporate CDO transactions, the fair value of our Non-Corporate CDO transactions is calculated as the net present value of the difference between our contractual premium and our estimate of the fair premium for these transactions. For our credit card and auto loan transactions, the fair premium is estimated using observed spreads on recent trades of securities that are similar to the securities that we guaranty. In all other instances, we utilize internal models to estimate fair premium as described below.
RMBS CDOs
The fair premium estimates for our two existing RMBS CDO transactions are derived from standard market indices, where market-based credit spreads are observable and can be used to approximate fair premium amounts. The credit quality of the underlying referenced obligations in one of these transactions is similar to that which is included in the AAA-rated ABX.HE index, a standardized list of residential mortgage-backed security reference obligations. Accordingly, the fair premium for this transaction is derived directly from the observed spreads of this index. For our other existing RMBS CDO transaction, the underlying referenced obligations in this transaction are closely equated to the referenced entities included in the TABX.HE index with respect to vintage periods, credit quality, and structure. This index represents a synthetic CDO tranche with underlying referenced entities equivalent to referenced obligations in the BBB-rated ABX.HE index. The fair premium for this transaction is therefore primarily derived from observable prices on the TABX.HE index with an adjustment made to reflect the modestly stronger credit quality of the referenced obligations in our transaction (“A” rated) compared to those included in theTABX.HE index (“BBB” rated). This adjustment for credit quality is determined based on the observed relationship of spreads on the “A” rated and “BBB” rated ABX.HE indices.
CMBS CDOs
The fair premium estimates for our CMBS CDO transactions are derived directly from the observed spreads on the CMBX indices. The CMBX indices represent standardized lists of commercial mortgage-backed security referenced obligations. A different CMBX index exists for different types of underlying referenced obligations based on their various origination periods and credit grades. For each of our CMBS CDO transactions we use the CMBX index that most directly correlates to our transaction with respect to the origination period and credit rating of the referenced obligations included in our transaction. The fair premium rate for each of our transactions is then determined based on the observed spread of the relevant CMBX index.
TRUPS CDOs
Our TRUPS transactions are synthetic CDOs where the underlying referenced obligations are preferred securities of financial institutions consisting primarily of banks and insurance companies whose individual spreads are not observable. In each case, we provide credit protection on a specific tranche of each synthetic CDO. In determining estimated fair values of our TRUPS CDO transactions, we use internally-generated models to calculate the fair premium rate based on the following inputs: our contractual premium (which is estimated to be equal to the fair premium as of the contract date), the estimated change in the spread of the underlying referenced obligations, and the remaining term of the TRUPS CDOs. We start with our contractual premium amounts and then make an adjustment for the estimated change in the spread of the underlying referenced obligations from inception of the transaction to the current measurement date. To determine the spread of the underlying referenced obligations, which are not observable, we assume these spreads to be proportional to the weighted average credit default swap spread of a group of investment grade and high yield institutions whose market risk is determined by us to be similar to the specific financial institutions underlying our TRUPS. The relationship between the spread on these referenced obligations and the spread on the tranche we insure is then determined based on the historical observed relationship between the spread on the referenced entities of the CDX index and the spread on a senior tranche of the CDX index, because the direct relationship for TRUPS CDOs is not observable. A separate adjustment to the premium rate is then calculated for each transaction based on its remaining average life. This adjustment accounts for changes in the remaining average life of our transactions and is based on historically observed prices for corporate obligations with similar remaining maturities.
All Other Non-Corporate CDOs
For all of our other Non-Corporate CDO transactions observed prices and market indices are not available. As a result, we utilize an internal model that estimates fair premium based on a market rate of return that would be required for a monoline insurer to undertake the contract risk. The fair premium amount is calculated as the premium required to achieve a market rate of return (net of expected losses and other expenses) over an estimated internally developed risk-based capital amount. Such market rates of return approximate historical rates of return earned by financial guarantors. Expected losses and our internally developed risk-based capital amounts are projected by our model based on the internal credit rating, term, asset class, and current par outstanding for each transaction.
For each of the Non-Corporate CDO transactions discussed above, we make an adjustment to the fair premium amounts described above to incorporate our own non-performance risk. Our non-performance risk adjustment for our Non-Corporate CDO transactions is derived from the observed relationship of our non-performance risk adjustments to the fair premium on our Corporate CDO transactions.
Changes in the estimated fair values for our Non-Corporate CDO transactions primarily result from changes in the observed indices as discussed above, as well as changes in the market’s perception of our non-performance risk. A 25 basis point change in these observed indices discussed above, assuming our credit default swap spread remained constant, would have resulted in a change of $1 million to our fair value liability of $237 million as of June 30, 2008. As of June 30, 2008, a 1,000 basis point widening or tightening in our credit default swap spread, assuming the spreads for the observed indices remained constant, would have resulted in an immaterial change to our current fair value. Because our credit default swap spread is at such a wide level, the current fair value amount is relatively insensitive to reasonably likely changes. These credit derivatives are categorized in Level III of the fair value hierarchy.
The following table quantifies the impact of our non-performance risk on ourDerivative Assets andDerivative Liabilities (in aggregate by derivative type) presented in our consolidated balance sheets ($ in millions):
| January 1, 2007 | January 1, 2008 | March 31, 2008 | June 30, 2008 | ||||||||
Radian 5 year credit default swap spread | 37 basis points | 628 basis points | 1,095 basis points | 2,530 basis points | |||||||
Derivative type | Impact of initial adoption 1/1/08 Unrealized gain | Cumulative Unrealized gain at 3/31/08 | Cumulative Unrealized gain at 6/30/08 | ||||||||
Corporate CDOs | $ | 638.8 | $ | 1,584.8 | $ | 1,427.5 | |||||
Non-Corporate CDOs | 185.8 | 326.2 | 373.0 | ||||||||
NIMs | 108.2 | 147.2 | 233.9 | ||||||||
Total | $ | 932.8 | $ | 2,058.2 | $ | 2,034.4 | |||||
