Expected Loss to be Paid

The insured portfolio includes policies accounted for under three separate accounting models depending on the characteristics of the contract and the Company's control rights. The Company has paid and expects to pay future losses on policies which fall under each of the three accounting models. The following provides a summarized description of the three accounting models prescribed by GAAP with a reference to the notes that describe the accounting policies and required disclosures throughout this report. The three models are: (1) insurance, (2) derivative and (3) VIE consolidation.

In order to effectively evaluate and manage the economics and liquidity of the entire insured portfolio, management compiles and analyzes loss information for all policies on a consistent basis. The Company monitors and assigns ratings and calculates expected losses in the same manner for all its exposures regardless of form or differing accounting models.

This note provides information regarding expected claim payments to be made under all contracts in the insured portfolio, regardless of the accounting model. Net expected loss to be paid in the tables below consists of the present value of future: expected claim and LAE payments, expected recoveries in the transaction structures, cessions to reinsurers, and expected recoveries for breaches of representations and warranties (R&W) and other loss mitigation strategies. Expected loss to be paid is important from a liquidity perspective in that it represents the present value of amounts that the Company expects to pay or recover in future periods, regardless of the accounting model. Expected loss to be paid is an important measure used by management to analyze the net economic loss on all contracts.

Accounting Policy

Insurance Accounting

For contracts accounted for as financial guaranty insurance, loss and LAE reserve is recorded only to the extent and for the amount that expected losses to be paid, exceed unearned premium reserve. As a result, the Company has expected loss to be paid that have not yet been expensed. Such amounts will be recognized in future periods as deferred premium revenue amortizes into income. Expected loss to be expensed is important because it represents the Company's projection of incurred losses that will be recognized in future periods (excluding accretion of discount). See "Financial Guaranty Insurance Losses" in Note 6, Contracts Accounted for as Insurance.

Derivative Accounting, at Fair Value

For contracts that do not meet the financial guaranty scope exception in the derivative accounting guidance (primarily due to the fact that the insured is not required to be exposed to the insured risk throughout the life of the contract), the Company records such credit derivative contracts at fair value on the consolidated balance sheet with changes in fair value recorded in the consolidated statement of operations. The fair value recorded on the balance sheet represents an exit price in a hypothetical market because the Company does not trade its credit derivative contracts. The fair value is determined using significant Level 3 inputs in an internally developed model while the expected loss to be paid (which represents the net present value of expected cash outflows) uses methodologies and assumptions consistent with financial guaranty insurance expected losses to be paid. See Note 7, Fair Value Measurement and Note 8, Contracts Accounted for as Credit Derivatives.

VIE Consolidation, at Fair Value

For financial guaranty (FG) insurance contracts issued on the debt of variable interest entities over which the Company is deemed to be the primary beneficiary due to its control rights, as defined in GAAP, the Company consolidates the FG VIE. The Company carries the assets and liabilities of the FG VIEs at fair value under the fair value option. Management assesses the expected losses on the insured debt of the consolidated FG VIEs in the same manner as other financial guaranty insurance and credit derivative contracts. See Note 9, Consolidated Variable Interest Entities.

Expected Loss to be Paid

The expected loss to be paid is equal to the present value of expected future cash outflows for claim and LAE payments, net of inflows for expected salvage and subrogation (e.g., excess spread on the underlying collateral, and expected and contractual recoveries for breaches of R&W or other expected recoveries), using current risk-free rates. When the Company becomes entitled to the cash flow from the underlying collateral of an insured credit under salvage and subrogation rights as a result of a claim payment or estimated future claim payment, it reduces the expected loss to be paid on the contract. Net expected loss to be paid is defined as expected loss to be paid, net of amounts ceded to reinsurers.

The Company updates the discount rate each quarter and reflects the effect of such changes in economic loss development. Expected cash outflows and inflows are probability weighted cash flows that reflect the likelihood of all possible outcomes. The Company estimates the expected cash outflows and inflows using management's assumptions about the likelihood of all possible outcomes based on all information available to it. Those assumptions consider the relevant facts and circumstances and are consistent with the information tracked and monitored through the Company's risk-management activities.

Economic Loss Development

Economic loss development represents the change in net expected loss to be paid attributable to the effects of changes in assumptions based on observed market trends, changes in discount rates, accretion of discount and the economic effects of loss mitigation efforts.

Expected loss to be paid and economic loss development include the effects of loss mitigation strategies such as negotiated and estimated recoveries for breaches of R&W, and purchases of insured debt obligations. Additionally, in certain cases, issuers of insured obligations elected, or the Company and an issuer mutually agreed as part of a negotiation, to deliver the underlying collateral or insured obligation to the Company.

In circumstances where the Company has purchased its own insured obligations that have expected losses, expected loss to be paid is reduced by the proportionate share of the insured obligation that is held in the investment portfolio. The difference between the purchase price of the obligation and the fair value excluding the value of the Company's insurance is treated as a paid loss. Assets that are purchased by the Company are recorded in the investment portfolio, at fair value, excluding the value of the Company's insurance. See Note 10, Investments and Cash and Note 7, Fair Value Measurement.

Loss Estimation Process

The Company’s loss reserve committees estimate expected loss to be paid for all contracts by reviewing analyses that consider various scenarios with corresponding probabilities assigned to them. Depending upon the nature of the risk, the Company’s view of the potential size of any loss and the information available to the Company, that analysis may be based upon individually developed cash flow models, internal credit rating assessments and sector-driven loss severity assumptions or judgmental assessments. In the case of its assumed business, the Company may conduct its own analysis as just described or, depending on the Company’s view of the potential size of any loss and the information available to the Company, the Company may use loss estimates provided by ceding insurers. The Company monitors the performance of its transactions with expected losses and each quarter the Company’s loss reserve committees review and refresh their loss projection assumptions and scenarios and the probabilities they assign to those scenarios based on actual developments during the quarter and their view of future performance.

The financial guaranties issued by the Company insure the credit performance of the guaranteed obligations over an extended period of time, in some cases over 30 years, and in most circumstances the Company has no right to cancel such financial guaranties. As a result, the Company's estimate of ultimate losses on a policy is subject to significant uncertainty over the life of the insured transaction. Credit performance can be adversely affected by economic, fiscal and financial market variability over the long life of most contracts.

The determination of expected loss to be paid is an inherently subjective process involving numerous estimates, assumptions and judgments by management, using both internal and external data sources with regard to frequency, severity of loss, economic projections, governmental actions, negotiations and other factors that affect credit performance. These estimates, assumptions and judgments, and the factors on which they are based, may change materially over a reporting period, and as a result the Company’s loss estimates may change materially over that same period.

Changes in the Company’s loss estimates for structured finance transactions generally will be influenced by factors impacting the performance of the assets supporting those transactions. For example, changes over a reporting period in the Company’s loss estimates for its RMBS transactions may be influenced by such factors as the level and timing of loan defaults experienced; changes in housing prices; results from the Company's loss mitigation activities; and other variables.

Similarly, changes over a reporting period in the Company’s loss estimates for municipal obligations supported by specified revenue streams, such as revenue bonds issued by toll road authorities, municipal utilities or airport authorities, generally will be influenced by factors impacting their revenue levels, such as changes in demand; changing demographics; and other economic factors, especially if the obligations do not benefit from financial support from other tax revenues or governmental authorities. Changes over a reporting period in the Company’s loss estimates for its tax-supported public finance transactions generally will be influenced by factors impacting the public issuer’s ability and willingness to pay, such as changes in the economy and population of the relevant area; changes in the issuer’s ability or willingness to raise taxes, decrease spending or receive federal assistance; new legislation; rating agency downgrades that reduce the issuer’s ability to refinance maturing obligations or issue new debt at a reasonable cost; changes in the priority or amount of pensions and other obligations owed to workers; developments in restructuring or settlement negotiations; and other political and economic factors.

The Company does not use traditional actuarial approaches to determine its estimates of expected losses. Actual losses will ultimately depend on future events or transaction performance and may be influenced by many interrelated factors that are difficult to predict. As a result, the Company's current projections of probable and estimable losses may be subject to considerable volatility and may not reflect the Company's ultimate claims paid.

In some instances, the terms of the Company's policy gives it the option to pay principal losses that have been recognized in the transaction but which it is not yet required to pay, thereby reducing the amount of guaranteed interest due in the future. The Company has sometimes exercised this option, which uses cash but reduces projected future losses.

The following tables present a roll forward of the present value of net expected loss to be paid for all contracts, whether accounted for as insurance, credit derivatives or FG VIEs, by sector, after the benefit for expected recoveries for breaches of R&W and other expected recoveries. The Company used risk-free rates for U.S. dollar denominated obligations, that ranged from 0.0% to 3.23% with a weighted average of 2.73% as of December 31, 2016 and 0.0% to 3.25% with a weighted average of 2.36% as of December 31, 2015.

Net Expected Loss to be Paid

Roll Forward



	Year Ended December 31,
	2016			2015
	(in millions)
Net expected loss to be paid, beginning of period	$	1,391		$	1,169
Net expected loss to be paid on the CIFGH portfolio as of July 1, 2016	22			—
Net expected loss to be paid on Radian Asset portfolio as of April 1, 2015	—			190
Economic loss development due to:
Accretion of discount	26			32
Changes in discount rates	(15		)	(23		)
Changes in timing and assumptions	128			310
Total economic loss development	139			319
Paid losses	(354		)	(287		)
Net expected loss to be paid, end of period	$	1,198		$	1,391

Net Expected Loss to be Paid

Roll Forward by Sector

Year Ended December 31, 2016



	Net Expected Loss to be Paid (Recovered) as of December 31, 2015(2)		Net Expected Loss to be Paid (Recovered) on CIFG as of July 1, 2016			Economic Loss Development			(Paid) Recovered Losses (1)			Net Expected Loss to be Paid (Recovered) as of December 31, 2016 (2)
	(in millions)
Public finance:
U.S. public finance	$	771	$	40		$	276		$	(216	)	$	871
Non-U.S. public finance	38		2			(7		)	—			33
Public finance	809		42			269			(216		)	904
Structured finance:
U.S. RMBS	409		(22		)	(91		)	(90		)	206
Triple-X life insurance transactions	99		—			(22		)	(23		)	54
Other structured finance	74		2			(17		)	(25		)	34
Structured finance	582		(20		)	(130		)	(138		)	294
Total	$	1,391	$	22		$	139		$	(354	)	$	1,198

Net Expected Loss to be Paid

Roll Forward by Sector

Year Ended December 31, 2015



	Net Expected Loss to be Paid (Recovered) as of December 31, 2014		Net Expected Loss to be Paid (Recovered) on Radian Asset portfolio as of April 1, 2015		Economic Loss Development			(Paid) Recovered Losses (1)			Net Expected Loss to be Paid (Recovered) as of December 31, 2015 (2)
	(in millions)
Public finance:
U.S. public finance	$	303	$	81	$	416		$	(29	)	$	771
Non-U.S. public finance	45		4		(11		)	—			38
Public finance	348		85		405			(29		)	809
Structured finance:
U.S. RMBS	584		4		(82		)	(97		)	409
Triple-X life insurance transactions	161		—		11			(73		)	99
Other structured finance	76		101		(15		)	(88		)	74
Structured finance	821		105		(86		)	(258		)	582
Total	$	1,169	$	190	$	319		$	(287	)	$	1,391

____________________


(1)	Net of ceded paid losses, whether or not such amounts have been settled with reinsurers. Ceded paid losses are typically settled 45 days after the end of the reporting period. Such amounts are recorded in reinsurance recoverable on paid losses included in other assets. The Company paid $16 million and $25 million in LAE for the years ended December 31, 2016 and 2015, respectively.


(2)	Includes expected LAE to be paid of $12 million as of December 31, 2016 and $12 million as of December 31, 2015.

Future Net R&W Recoverable (Payable)(1)



	Future Net R&W Benefit as of December 31, 2016			Future Net R&W Benefit as of December 31, 2015		Future Net R&W Benefit as of December 31, 2014
	(in millions)
U.S. RMBS:
First lien	$	(53	)	$	0	$	232
Second lien	47			79		85
Total	$	(6	)	$	79	$	317

____________________

(1)

The Company’s agreements with R&W providers generally provide that, as the Company makes claim payments, the R&W providers reimburse it for those claims; if the Company later receives reimbursement through the transaction (for example, from excess spread), the Company repays the R&W providers. See the section “Breaches of Representations and Warranties” for information about the R&W agreements. When the Company projects receiving more reimbursements in the future than it projects paying in claims on transactions covered by R&W settlement agreements, the Company will have a net R&W payable.

The following table presents the present value of net expected loss to be paid for all contracts by accounting model, by sector and after the benefit for expected recoveries for breaches of R&W.

Net Expected Loss to be Paid (Recovered)

By Accounting Model



	As of December 31, 2016						As of December 31, 2015
	Public Finance		Structured Finance		Total		Public Finance		Structured Finance		Total
	(in millions)
Financial guaranty insurance	$	904	$	179	$	1,083	$	809	$	430	$	1,239
FG VIEs (1) and other	—		105		105		—		136		136
Credit derivatives (2)	0		10		10		—		16		16
Total	$	904	$	294	$	1,198	$	809	$	582	$	1,391

___________

(1) Refer to Note 9, Consolidated Variable Interest Entities.

(2) Refer to Note 8, Contracts Accounted for as Credit Derivatives.

The following table presents the net economic loss development for all contracts by accounting model, by sector and after the benefit for expected recoveries for breaches of R&W.

Net Economic Loss Development (Benefit)

By Accounting Model



	Year Ended December 31, 2016								Year Ended December 31, 2015
	Public Finance		Structured Finance			Total			Public Finance			Structured Finance			Total
	(in millions)
Financial guaranty insurance	$	269	$	(105	)	$	164		$	410		$	(25	)	$	385
FG VIEs (1) and other	—		(8		)	(8		)	—			16			16
Credit derivatives (2)	—		(17		)	(17		)	(5		)	(77		)	(82		)
Total	$	269	$	(130	)	$	139		$	405		$	(86	)	$	319

__________

(1) Refer to Note 9, Consolidated Variable Interest Entities.

(2) Refer to Note 8, Contracts Accounted for as Credit Derivatives.

Selected U.S. Public Finance Transactions

The Company insures general obligation bonds of the Commonwealth of Puerto Rico and various obligations of its related authorities and public corporations aggregating $4.8 billion net par as of December 31, 2016, all of which are BIG. For additional information regarding the Company's exposure to general obligations of Commonwealth of Puerto Rico and various obligations of its related authorities and public corporations, please refer to "Exposure to Puerto Rico" in Note 4, Outstanding Exposure.

On February 25, 2015, a plan of adjustment resolving the bankruptcy filing of the City of Stockton, California under chapter 9 of the U.S. Bankruptcy Code became effective. As of December 31, 2016, the Company’s net par subject to the plan consists of $113 million of pension obligation bonds. As part of the plan settlement, the City will repay the pension obligation bonds from certain fixed payments and certain variable payments contingent on the City's revenue growth.

The Company projects that its total net expected loss across its troubled U.S. public finance credits as of December 31, 2016, including those mentioned above, which incorporated the likelihood of the various outcomes, will be $871 million, compared with a net expected loss of $771 million as of December 31, 2015. On July 1, 2016, the CIFG Acquisition added $40 million in net economic losses to be paid for U.S. public finance credits. Economic loss development in 2016 was $276 million, which was primarily attributable to Puerto Rico exposures.

Certain Selected European Country Sub-Sovereign Transactions

The Company insures and reinsures credits with sub-sovereign exposure to various Spanish and Portuguese issuers where a Spanish and Portuguese sovereign default may cause the sub-sovereigns also to default. The Company's exposure net of reinsurance to these Spanish and Portuguese credits is $342 million and $76 million, respectively. The Company rates most of these issuers BIG due to the financial condition of Spain and Portugal and their dependence on the sovereign. The Company's Hungary exposure is to infrastructure bonds dependent on payments from Hungarian governmental entities. The Company's exposure net of reinsurance to these Hungarian credits is $236 million, all of which is rated BIG. The Company estimated net expected losses of $29 million related to these Spanish, Portuguese and Hungarian credits. The economic benefit of approximately $7 million during 2016 was primarily related to changes in the exchange rate between the euro and U.S. Dollar.

Approach to Projecting Losses in U.S. RMBS

The Company projects losses on its insured U.S. RMBS on a transaction-by-transaction basis by projecting the performance of the underlying pool of mortgages over time and then applying the structural features (i.e., payment priorities and tranching) of the RMBS and any R&W agreements to the projected performance of the collateral over time. The resulting projected claim payments or reimbursements are then discounted using risk-free rates.

The further behind a mortgage borrower falls in making payments, the more likely it is that he or she will default. The rate at which borrowers from a particular delinquency category (number of monthly payments behind) eventually default is referred to as the “liquidation rate.” The Company derives its liquidation rate assumptions from observed roll rates, which are the rates at which loans progress from one delinquency category to the next and eventually to default and liquidation. The Company applies liquidation rates to the mortgage loan collateral in each delinquency category and makes certain timing assumptions to project near-term mortgage collateral defaults from loans that are currently delinquent.

Mortgage borrowers that are not more than one payment behind (generally considered performing borrowers) have demonstrated an ability and willingness to pay throughout the recession and mortgage crisis, and as a result are viewed as less likely to default than delinquent borrowers. Performing borrowers that eventually default will also need to progress through delinquency categories before any defaults occur. The Company projects how many of the currently performing loans will default and when they will default, by first converting the projected near term defaults of delinquent borrowers derived from liquidation rates into a vector of conditional default rates (CDR), then projecting how the CDR will develop over time. Loans that are defaulted pursuant to the CDR after the near-term liquidation of currently delinquent loans represent defaults of currently performing loans and projected re-performing loans. A CDR is the outstanding principal amount of defaulted loans liquidated in the current month divided by the remaining outstanding amount of the whole pool of loans (or “collateral pool balance”). The collateral pool balance decreases over time as a result of scheduled principal payments, partial and whole principal prepayments, and defaults.

In order to derive collateral pool losses from the collateral pool defaults it has projected, the Company applies a loss severity. The loss severity is the amount of loss the transaction experiences on a defaulted loan after the application of net proceeds from the disposal of the underlying property. The Company projects loss severities by sector based on its experience to date. The Company continues to update its evaluation of these loss severities as new information becomes available.

The Company has been enforcing claims for breaches of R&W regarding the characteristics of the loans included in the collateral pools. The Company calculates a credit for R&W recoveries to include in its cash flow projections. Where the Company has an agreement with an R&W provider (such as its agreements with Bank of America and UBS, which are described in more detail under "Breaches of Representations and Warranties" below), that credit is based on the agreement. Where the Company does not have an agreement with the R&W provider but the Company believes the R&W provider to be economically viable, the Company estimates what portion of its past and projected future claims it believes will be reimbursed by that provider.

The Company projects the overall future cash flow from a collateral pool by adjusting the payment stream from the principal and interest contractually due on the underlying mortgages for the collateral losses it projects as described above; assumed voluntary prepayments; and servicer advances. The Company then applies an individual model of the structure of the transaction to the projected future cash flow from that transaction’s collateral pool to project the Company’s future claims and claim reimbursements for that individual transaction. Finally, the projected claims and reimbursements are discounted using risk-free rates. The Company runs several sets of assumptions regarding mortgage collateral performance, or scenarios, and probability weights them.

The Company's RMBS loss projection methodology assumes that the housing and mortgage markets will continue improving. Each period the Company makes a judgment as to whether to change the assumptions it uses to make RMBS loss projections based on its observation during the period of the performance of its insured transactions (including early stage delinquencies, late stage delinquencies and loss severity) as well as the residential property market and economy in general, and, to the extent it observes changes, it makes a judgment as whether those changes are normal fluctuations or part of a trend.

Year-End 2016 Compared to Year-End 2015 U.S. RMBS Loss Projections

Year-End 2015 Compared to Year-End 2014 U.S. RMBS Loss Projections

Based on its observation during the period of the performance of its insured transactions (including early stage delinquencies, late stage delinquencies and loss severity) as well as the residential property market and economy in general, the Company chose to use the same general assumptions to project RMBS losses as of December 31, 2015 as it used as of December 31, 2014, except that, for its first lien RMBS loss projections for 2015, it shortened by twelve months the period it is projecting it will take in the base case to reach the final CDR as compared with December 31, 2014. The methodology and revised assumptions the Company used to project first lien RMBS losses and the scenarios it employed are described in more detail below under " - U.S. First Lien RMBS Loss Projections: Alt A First Lien, Option ARM, Subprime and Prime", and the methodology and assumptions the Company uses to project second lien RMBS losses and the scenarios it employs are described in more detail below under " - U.S. Second Lien RMBS Loss Projections." In 2015 the economic benefit was $124 million for first lien U.S. RMBS and loss development was $42 million for second lien U.S. RMBS.

U.S. First Lien RMBS Loss Projections: Alt-A First Lien, Option ARM, Subprime and Prime

The majority of projected losses in first lien RMBS transactions are expected to come from non-performing mortgage loans (those that are or in the past twelve months have been two or more payments behind, have been modified, are in foreclosure, or have been foreclosed upon). Changes in the amount of non-performing loans from the amount projected in the previous period are one of the primary drivers of loss development in this portfolio. In order to determine the number of defaults resulting from these delinquent and foreclosed loans, the Company applies a liquidation rate assumption to loans in each of various non-performing categories. The Company arrived at its liquidation rates based on data purchased from a third party provider and assumptions about how delays in the foreclosure process and loan modifications may ultimately affect the rate at which loans are liquidated. Each quarter the Company reviews the most recent twelve months of this data and (if necessary) adjusts its liquidation rates based on its observations. The following table shows liquidation assumptions for various non-performing categories.

First Lien Liquidation Rates



	December 31, 2016	December 31, 2015	December 31, 2014
Current Loans Modified in the Previous 12 Months
Alt-A and Prime	25%	25%	25%
Option ARM	25	25	25
Subprime	25	25	25
Current Loans Delinquent in the Previous 12 Months
Alt-A and Prime	25	25	25
Option ARM	25	25	25
Subprime	25	25	25
30 – 59 Days Delinquent
Alt-A and Prime	35	35	35
Option ARM	35	40	40
Subprime	40	45	35
60 – 89 Days Delinquent
Alt-A and Prime	45	45	50
Option ARM	50	50	55
Subprime	50	55	40
90+ Days Delinquent
Alt-A and Prime	55	55	60
Option ARM	55	60	65
Subprime	55	60	55
Bankruptcy
Alt-A and Prime	45	45	45
Option ARM	50	50	50
Subprime	40	40	40
Foreclosure
Alt-A and Prime	65	65	75
Option ARM	65	70	80
Subprime	65	70	70
Real Estate Owned
All	100	100	100

While the Company uses liquidation rates as described above to project defaults of non-performing loans (including current loans modified or delinquent within the last 12 months), it projects defaults on presently current loans by applying a CDR trend. The start of that CDR trend is based on the defaults the Company projects will emerge from currently nonperforming, recently nonperforming and modified loans. The total amount of expected defaults from the non-performing loans is translated into a constant CDR (i.e., the CDR plateau), which, if applied for each of the next 36 months, would be sufficient to produce approximately the amount of defaults that were calculated to emerge from the various delinquency categories. The CDR thus calculated individually on the delinquent collateral pool for each RMBS is then used as the starting point for the CDR curve used to project defaults of the presently performing loans.

In the base case, after the initial 36-month CDR plateau period, each transaction’s CDR is projected to improve over 12 months to an intermediate CDR (calculated as 20% of its CDR plateau); that intermediate CDR is held constant for 36 months and then trails off in steps to a final CDR of 5% of the CDR plateau. In the base case, the Company assumes the final CDR will be reached 6.5 years after the initial 36-month CDR plateau period. Under the Company’s methodology, defaults projected to occur in the first 36 months represent defaults that can be attributed to loans that were modified or delinquent in the last 12 months or that are currently delinquent or in foreclosure, while the defaults projected to occur using the projected CDR trend after the first 36 month period represent defaults attributable to borrowers that are currently performing or are projected to reperform.

Another important driver of loss projections is loss severity, which is the amount of loss the transaction incurs on a loan after the application of net proceeds from the disposal of the underlying property. Loss severities experienced in first lien transactions have reached historically high levels, and the Company is assuming in the base case that these high levels generally will continue for another 18 months. The Company determines its initial loss severity based on actual recent experience. As a result, the Company updated severities for specific asset classes and vintages based on observed data, as shown in the tables below. The Company then assumes that loss severities begin returning to levels consistent with underwriting assumptions beginning after the initial 18 month period, declining to 40% in the base case over 2.5 years.

The following table shows the range as well as the average, weighted by outstanding net insured par, for key assumptions used in the calculation of expected loss to be paid for individual transactions for direct vintage 2004 - 2008 first lien U.S. RMBS.

Key Assumptions in Base Case Expected Loss Estimates

First Lien RMBS(1)



	As of December 31, 2016					As of December 31, 2015					As of December 31, 2014
	Range				Weighted Average	Range				Weighted Average	Range				Weighted Average
Alt-A First Lien
Plateau CDR	1.0	%	–	13.5%	5.7%	1.7	%	–	26.4%	6.4%	2.0	%	–	13.4%	7.3%
Final CDR	0.0	%	–	0.7%	0.3%	0.1	%	–	1.3%	0.3%	0.1	%	–	0.7%	0.3%
Initial loss severity:
2005 and prior	60.0%					60.0%					60.0%
2006	80.0%					70.0%					70.0%
2007	70.0%					65.0%					65.0%
Option ARM
Plateau CDR	3.2	%	–	7.0%	5.6%	3.5	%	–	10.3%	7.8%	4.3	%	–	14.2%	10.6%
Final CDR	0.2	%	–	0.3%	0.3%	0.2	%	–	0.5%	0.4%	0.2	%	–	0.7%	0.5%
Initial loss severity:
2005 and prior	60.0%					60.0%					60.0%
2006	70.0%					70.0%					70.0%
2007	75.0%					65.0%					65.0%
Subprime
Plateau CDR	2.8	%	–	14.1%	8.1%	4.7	%	–	13.2%	9.5%	4.9	%	–	15.0%	10.6%
Final CDR	0.1	%	–	0.7%	0.4%	0.2	%	–	0.7%	0.4%	0.2	%	–	0.7%	0.4%
Initial loss severity:
2005 and prior	80.0%					75.0%					75.0%
2006	90.0%					90.0%					90.0%
2007	90.0%					90.0%					90.0%

____________________

(1) Represents variables for most heavily weighted scenario (the “base case”).

The rate at which the principal amount of loans is voluntarily prepaid may impact both the amount of losses projected (since that amount is a function of the CDR, the loss severity and the loan balance over time) as well as the amount of excess spread (the amount by which the interest paid by the borrowers on the underlying loan exceeds the amount of interest owed on the insured obligations). The assumption for the voluntary conditional prepayment rate (CPR) follows a similar pattern to that of the CDR. The current level of voluntary prepayments is assumed to continue for the plateau period before gradually increasing over 12 months to the final CPR, which is assumed to be 15% in the base case. For transactions where the initial CPR is higher than the final CPR, the initial CPR is held constant and the final CPR is not used. These CPR assumptions are the same as those the Company used for December 31, 2015.

In estimating expected losses, the Company modeled and probability weighted sensitivities for first lien transactions by varying its assumptions of how fast a recovery is expected to occur. One of the variables used to model sensitivities was how quickly the CDR returned to its modeled equilibrium, which was defined as 5% of the initial CDR. The Company also stressed CPR and the speed of recovery of loss severity rates. The Company probability weighted a total of five scenarios as of December 31, 2016. The Company used a similar approach to establish its pessimistic and optimistic scenarios as of December 31, 2016 as it used as of December 31, 2015, increasing and decreasing the periods of stress from those used in the base case.

In the Company's most stressful scenario where loss severities were assumed to rise and then recover over nine years and the initial ramp-down of the CDR was assumed to occur over 15 months and other assumptions were the same as the other stress scenario, expected loss to be paid would increase from current projections by approximately $27 million for Alt-A first liens, $8 million for Option ARM, $46 million for subprime and $1 million for prime transactions.

In the Company's least stressful scenario where the CDR plateau was six months shorter (30 months, effectively assuming that liquidation rates would improve) and the CDR recovery was more pronounced, (including an initial ramp-down of the CDR over nine months), expected loss to be paid would decrease from current projections by approximately $13 million for Alt-A first liens, $22 million for Option ARM, $25 million for subprime and $0.1 million for prime transactions.

U.S. Second Lien RMBS Loss Projections

Second lien RMBS transactions include both home equity lines of credit (HELOC) and closed end second lien. The Company believes the primary variable affecting its expected losses in second lien RMBS transactions is the amount and timing of future losses in the collateral pool supporting the transactions. Expected losses are also a function of the structure of the transaction; the voluntary prepayment rate (typically also referred to as CPR of the collateral); the interest rate environment; and assumptions about the draw rate and loss severity.

In second lien transactions the projection of near-term defaults from currently delinquent loans is relatively straightforward because loans in second lien transactions are generally “charged off” (treated as defaulted) by the securitization’s servicer once the loan is 180 days past due. The Company estimates the amount of loans that will default over the next six months by calculating current representative liquidation rates. A liquidation rate is the percent of loans in a given cohort (in this instance, delinquency category) that ultimately default. Similar to first liens, the Company then calculates a CDR for six months, which is the period over which the currently delinquent collateral is expected to be liquidated. That CDR is then used as the basis for the plateau CDR period that follows the embedded five months of losses.

For the base case scenario, the CDR (the plateau CDR) was held constant for six months. Once the plateau period has ended, the CDR is assumed to gradually trend down in uniform increments to its final long-term steady state CDR. (The long-term steady state CDR is calculated as the constant CDR that would have yielded the amount of losses originally expected at underwriting.) In the base case scenario, the time over which the CDR trends down to its final CDR is 28 months. Therefore, the total stress period for second lien transactions is 34 months, comprising six months of delinquent data and 28 months of decrease to the steady state CDR, the same as of December 31, 2015.

HELOC loans generally permit the borrower to pay only interest for an initial period (often ten years) and, after that period, require the borrower to make both the monthly interest payment and a monthly principal payment, and so increase the borrower's aggregate monthly payment. Some of the HELOC loans underlying the Company's insured HELOC transactions have reached their principal amortization period. The Company has observed that the increase in monthly payments occurring when a loan reaches its principal amortization period, even if mitigated by borrower relief offered by the servicer, is associated with increased borrower defaults. Thus, most of the Company's HELOC projections incorporate an assumption that a percentage of loans reaching their amortization periods will default around the time of the payment increase. These projected defaults are in addition to those generated using the CDR curve as described above. This assumption is similar to the one used as of December 31, 2015.

When a second lien loan defaults, there is generally a very low recovery. The Company assumed as of December 31, 2016 that it will generally recover only 2% of the collateral defaulting in the future and declining additional amounts of post-default receipts on previously defaulted collateral. This is the same assumption used as of December 31, 2015.

The rate at which the principal amount of loans is prepaid may impact both the amount of losses projected as well as the amount of excess spread. In the base case, an average CPR (based on experience of the past year) is assumed to continue until the end of the plateau before gradually increasing to the final CPR over the same period the CDR decreases. The final CPR is assumed to be 15% for second lien transactions, which is lower than the historical average but reflects the Company’s continued uncertainty about the projected performance of the borrowers in these transactions. For transactions where the initial CPR is higher than the final CPR, the initial CPR is held constant and the final CPR is not used. This pattern is generally consistent with how the Company modeled the CPR as of December 31, 2015. To the extent that prepayments differ from projected levels it could materially change the Company’s projected excess spread and losses.

The Company uses a number of other variables in its second lien loss projections, including the spread between relevant interest rate indices. These variables have been relatively stable and in the relevant ranges have less impact on the projection results than the variables discussed above. However, in a number of HELOC transactions the servicers have been modifying poorly performing loans from floating to fixed rates, and, as a result, rising interest rates would negatively impact the excess spread available from these modified loans to support the transactions. The Company incorporated these modifications in its assumptions.

In estimating expected losses, the Company modeled and probability weighted five possible CDR curves applicable to the period preceding the return to the long-term steady state CDR. The Company used five scenarios at December 31, 2016 and December 31, 2015. The Company believes that the level of the elevated CDR and the length of time it will persist, the ultimate prepayment rate, and the amount of additional defaults because of the expiry of the interest only period, are the primary drivers behind the likely amount of losses the collateral will suffer. The Company continues to evaluate the assumptions affecting its modeling results.

The Company believes the most important driver of its projected second lien RMBS losses is the performance of its HELOC transactions. The following table shows the range as well as the average, weighted by outstanding net insured par, for key assumptions for the calculation of expected loss to be paid for individual transactions for direct vintage 2004 - 2008 HELOCs.

Key Assumptions in Base Case Expected Loss Estimates

HELOCs(1)



	As of December 31, 2016					As of December 31, 2015					As of December 31, 2014
	Range				Weighted Average	Range				Weighted Average	Range				Weighted Average
Plateau CDR	3.5	%	–	24.8%	13.6%	4.9	%	–	23.5%	10.3%	2.8	%	–	6.8%	4.1%
Final CDR trended down to	0.5	%	–	3.2%	1.3%	0.5	%	–	3.2%	1.2%	0.5	%	–	3.2%	1.2%
Liquidation rates:
Current Loans Modified in the Previous 12 Months	25%					25%					25%
Current Loans Delinquent in the Previous 12 Months	25					25					25
30 – 59 Days Delinquent	50					50					55
60 – 89 Days Delinquent	65					65					70
90+ Days Delinquent	80					75					80
Bankruptcy	55					55					55
Foreclosure	75					75					75
Real Estate Owned	100					100					100
Loss severity	98%					98%					90	%	–	98%	90.4%

____________________


(1)	Represents variables for most heavily weighted scenario (the base case).

The Company’s base case assumed a six month CDR plateau and a 28 month ramp-down (for a total stress period of 34 months). The Company also modeled a scenario with a longer period of elevated defaults and another with a shorter period of elevated defaults. Increasing the CDR plateau to eight months and increasing the ramp-down by three months to 31 months (for a total stress period of 39 months), and doubling the defaults relating to the end of the interest only period would increase the expected loss by approximately $39 million for HELOC transactions. On the other hand, reducing the CDR plateau to four months and decreasing the length of the CDR ramp-down to 25 months (for a total stress period of 29 months), and lowering the ultimate prepayment rate to 10% would decrease the expected loss by approximately $23 million for HELOC transactions.

Breaches of Representations and Warranties

The Company entered into agreements with R&W providers under which those providers made payments to the Company, agreed to make payments to the Company in the future, and / or repurchased loans from the transactions, all in return for releases of related liability by the Company. As of December 31, 2016, the Company had two such agreements remaining. Under the Company's agreement with Bank of America Corporation and certain of its subsidiaries (Bank of America), Bank of America agreed to reimburse the Company for 80% of claims on the first lien transactions covered by the agreement that the Company pays in the future, subject to a cap the Company currently projects it will not reach. Under the Company’s agreement with UBS Real Estate Securities Inc. and affiliates (UBS), UBS agreed to reimburse the Company for 85% of future losses on three first lien RMBS transactions. Bank of America and UBS have posted collateral to secure their obligations under these agreements. The Company also had an R&W reimbursement agreement with Deutsche Bank AG and certain of its affiliates (collectively, Deutsche Bank), but Deutsche Bank's reimbursement obligations under that agreement were terminated in May 2016 in return for a cash payment to the Company. The Company uses the same RMBS projection scenarios and weightings to project its future R&W benefit or payable as it uses to project RMBS losses on its portfolio.

As of December 31, 2016, the Company had a net R&W payable of $6 million to R&W counterparties, compared to an R&W recoverable of $79 million as of December 31, 2015. The decrease represents improvements in underlying collateral performance and the termination of the Deutsche Bank agreement described above, partially offset by the addition of R&W recoverable related to a RMBS insured by CIFGNA and still being pursued by the Company. The Company’s agreements with providers of R&W generally provide for reimbursement to the Company as claim payments are made and, to the extent the Company later receives reimbursements of such claims from excess spread or other sources, for the Company to provide reimbursement to the R&W providers. When the Company projects receiving more reimbursements in the future than it projects to pay in claims on transactions covered by R&W settlement agreements, the Company will have a net R&W payable.

Triple-X Life Insurance Transactions

The Company had $2.1 billion of net par exposure to financial guaranty Triple-X life insurance transactions as of December 31, 2016. Two of these transactions, with $126 million of net par outstanding, are rated BIG. The Triple-X life insurance transactions are based on discrete blocks of individual life insurance business. In older vintage Triple-X life insurance transactions, which include the two BIG-rated transactions, the amounts raised by the sale of the notes insured by the Company were used to capitalize a special purpose vehicle that provides reinsurance to a life insurer or reinsurer. The amounts are invested at inception in accounts managed by third-party investment managers. In the case of the two BIG-rated transactions, material amounts of their assets were invested in U.S. RMBS. Based on its analysis of the information currently available, including estimates of future investment performance, and projected credit impairments on the invested assets and performance of the blocks of life insurance business at December 31, 2016, the Company’s projected net expected loss to be paid is $54 million. The economic benefit during 2016 was approximately $22 million, which was due primarily to a benefit resulting from a purchase of a portion of an insured obligation to mitigate loss.

Student Loan Transactions

The Company has insured or reinsured $1.4 billion net par of student loan securitizations issued by private issuers and that it classifies as structured finance. Of this amount, $109 million is rated BIG. The Company is projecting approximately $32 million of net expected loss to be paid on these transactions. In general, the losses are due to: (i) the poor credit performance of private student loan collateral and high loss severities, or (ii) high interest rates on auction rate securities with respect to which the auctions have failed. The economic benefit during 2016 was approximately $14 million, which was driven primarily by the commutation of certain assumed student loan exposures earlier in the year.

Other structured finance

The Company's other structured finance sector has BIG net par of $966 million, comprising primarily transactions backed by TruPS, perpetual preferred securities, commercial receivables and manufactured housing loans. The economic benefit during 2016 was $3 million, which was attributable primarily to improved performance of various credits.

Recovery Litigation

Public Finance Transactions

On January 7, 2016, AGM, AGC and Ambac Assurance Corporation (Ambac) commenced an action for declaratory judgment and injunctive relief in the U.S. District Court for the District of Puerto Rico to invalidate the executive orders issued by the Governor on November 30, 2015 and December 8, 2015 directing that the Secretary of the Treasury of the Commonwealth of Puerto Rico and the Puerto Rico Tourism Company retain or transfer (in other words, claw back) certain taxes and revenues pledged to secure the payment of bonds issued by the PRHTA, the PRCCDA and the PRIFA. The Commonwealth defendants filed a motion to dismiss the action for lack of subject matter jurisdiction, which the Court denied on October 4, 2016. On October 14, 2016, the Commonwealth defendants filed a notice of PROMESA automatic stay.

On July 21, 2016, AGC and AGM filed a motion and form of complaint in the U.S. District Court for the District of Puerto Rico seeking relief from the stay provided by PROMESA. Upon a grant of relief from the PROMESA stay, the lawsuit further seeks a declaration that the Moratorium Act is preempted by Federal bankruptcy law and that certain gubernatorial executive orders diverting PRHTA pledged toll revenues (which are not subject to the Clawback) are preempted by PROMESA and violate the U.S. Constitution. Additionally, it seeks damages for the value of the PRHTA toll revenues diverted and injunctive relief prohibiting the defendants from taking any further action under these executive orders. On October 28, 2016, the Oversight Board filed a motion seeking leave to intervene in the action, which motion was denied on November 1, 2016, without prejudice, on procedural grounds. On November 2, 2016, the Court denied AGC’s and AGM’s motion for relief from the PROMESA stay on procedural grounds. The PROMESA stay expires on May 1, 2017.

For a discussion of the Company's exposure to Puerto Rico related to the litigation described above, please see Note 4, Outstanding Exposure.

On November 1, 2013, Radian Asset commenced a declaratory judgment action in the U.S. District Court for the Southern District of Mississippi against Madison County, Mississippi and the Parkway East Public Improvement District to establish its rights under a contribution agreement from the County supporting certain special assessment bonds issued by the District and insured by Radian Asset (now AGC). As of December 31, 2016, $20 million of such bonds were outstanding. The County maintained that its payment obligation is limited to two years of annual debt service, while AGC contended the County’s obligations under the contribution agreement continue so long as the bonds remain outstanding. On April 27, 2016, the Court granted AGC's motion for summary judgment, agreeing with AGC's interpretation of the County's obligations. On May 11, 2016, the County filed a notice of appeal of that ruling to the United States Court of Appeals for the Fifth Circuit.

Triple-X Life Insurance Transactions

In December 2008 AGUK filed an action in the Supreme Court of the State of New York against J.P. Morgan Investment Management Inc. (JPMIM), the investment manager for a triple-X life insurance transaction, Orkney Re II plc (Orkney), involving securities guaranteed by AGUK. As of December 31, 2016, the Company insures $423 million net par of Orkney securities. The action alleges that JPMIM engaged in breaches of fiduciary duty, gross negligence and breaches of contract based upon its handling of the Orkney investments. After AGUK’s claims were dismissed with prejudice in January 2010, AGUK was successful in its subsequent motions and appeals and, as of December 2011, all of AGUK’s claims for breaches of fiduciary duty, gross negligence and contract were reinstated in full. On January 22, 2016, AGUK filed a motion for partial summary judgment with respect to one of its claims for breach of contract relating to a failure to invest in compliance with the Delaware Insurance Code. On February 21, 2017, the court issued a decision on the motion. While the court denied the motion on the ground that the gross negligence of JPMIM in breaching the contract was a fact issue to be decided at trial, the court did find as a matter of law that JPMIM breached the contract relating to a failure to invest in compliance with the Delaware Insurance Code. A trial date has been set for mid-March 2017.

RMBS Transactions

On February 5, 2009, U.S. Bank National Association, as indenture trustee (U.S. Bank), CIFGNA, as insurer of the Class Ac Notes, and Syncora Guarantee Inc. (Syncora), as insurer of the Class Ax Notes, filed a complaint in the Supreme Court of the State of New York against GreenPoint Mortgage Funding, Inc. (GreenPoint) alleging GreenPoint breached its R&W with respect to the underlying mortgage loans in the GreenPoint Mortgage Funding Trust 2006-HE1 transaction. On March 3, 2010, the court dismissed CIFGNA's and Syncora’s causes of action on standing grounds. On December 16, 2013, GreenPoint moved to dismiss the remaining claims of U.S. Bank on the grounds that it too lacked standing. U.S. Bank cross-moved for partial summary judgment striking GreenPoint’s defense that U.S. Bank lacked standing to directly pursue claims against GreenPoint. On January 28, 2016, the court denied GreenPoint’s motion for summary judgment and granted U.S. Bank’s cross-motion for partial summary judgment, finding that as a matter of law U.S. Bank has standing to directly assert claims against GreenPoint. On November 28, 2016, GreenPoint filed an appeal. CIFGNA originally had $500 million insured net par exposure to this transaction; $23 million insured net par remains outstanding at December 31, 2016.

On November 26, 2012, CIFGNA filed a complaint in the Supreme Court of the State of New York against JP Morgan Securities LLC (JP Morgan) for material misrepresentation in the inducement of insurance and common law fraud, alleging that JP Morgan fraudulently induced CIFGNA to insure $400 million of securities issued by ACA ABS CDO 2006-2 Ltd. and $325 million of securities issued by Libertas Preferred Funding II, Ltd. On June 26, 2015, the Court dismissed with prejudice CIFGNA’s material misrepresentation in the inducement of insurance claim and dismissed without prejudice CIFGNA’s common law fraud claim. On September 24, 2015, the Court denied CIFGNA’s motion to amend but allowed CIFGNA to re-plead a cause of action for common law fraud. On November 20, 2015, CIFGNA filed a motion for leave to amend its complaint to re-plead common law fraud. On April 29, 2016, CIFGNA filed an appeal to reverse the Court’s decision dismissing CIFGNA’s material misrepresentation in the inducement of insurance claim. On November 29, 2016, the Appellate Division of the Supreme Court of the State of New York ruled that the Court’s decision dismissing with prejudice CIFGNA’s material misrepresentation in the inducement of insurance claim should be modified to grant CIFGNA leave to replead such claim.

On January 15, 2013, CIFGNA filed a complaint in the Supreme Court of the State of New York against Goldman, Sachs & Co. (Goldman) for material misrepresentation in the inducement of insurance and common law fraud, alleging that Goldman fraudulently induced CIFGNA to insure $325 million of Class A-1 Notes (the Class A-1 Notes) and to purchase $10 million of Class A-2 Notes (the Class A-2 Notes) issued by Fortius II Funding, Ltd. CDO. CIFGNA and Goldman agreed to separately arbitrate the issue of liability with respect to CIFG’s purchase of the Class A-2 Notes, and on February 4, 2015, an arbitration panel awarded CIFGNA $2.5 million in damages. On September 11, 2015, CIFGNA filed an amended complaint to allege that the arbitration award collaterally estopped Goldman from disputing its liability for fraudulent inducement in respect of the Class A-1 Notes. On October 20, 2016, AGC (as successor to CIFGNA) and Goldman reached a settlement of the action.