Loss Reserves

As described in Note 3 – “Summary of Significant Accounting Policies – Loss Reserves,” we establish reserves to recognize the estimated liability for losses and LAE related to defaults on insured mortgage loans. Loss reserves are established by estimating the number of loans in our inventory of delinquent loans that will result in a claim payment, which is referred to as the claim rate, and further estimating the amount of the claim payment, which is referred to as claim severity.

Estimation of losses is inherently judgmental. The conditions that affect the claim rate and claim severity include the current and future state of the domestic economy, including unemployment, the current and future strength of local housing markets; exposure on insured loans; the amount of time between default and claim filing, and curtailments and rescissions. The actual amount of the claim payments may be substantially different than our loss reserve estimates. Our estimates could be adversely affected by several factors, including a deterioration of regional or national economic conditions, including unemployment, leading to a reduction in borrowers’ income and thus their ability to make mortgage payments, and a drop in housing values which may affect borrower willingness to continue to make mortgage payments when the value of the home is below the mortgage balance. Changes to our estimates could result in a material impact to our results of operations and capital position, even in a stable economic environment.

The “Losses incurred” section of the table below shows losses incurred on defaults that occurred in the current year and in prior years. The amount of losses incurred relating to defaults that occurred in the current year represents the estimated amount to be ultimately paid on such defaults. The amount of losses incurred relating to defaults that occurred in prior years represents the actual claim rate and severity associated with those defaults resolved in the current year differing from the estimated liability at the prior year-end, as well as a re-estimation of amounts to be ultimately paid on defaults continuing from the end of the prior year. This re-estimation of the claim rate and severity is the result of our review of current trends in the default inventory, such as percentages of defaults that have resulted in a claim, the amount of the claims relative to the average loan exposure, changes in the relative level of defaults by geography and changes in average loan exposure.

Losses incurred on default notices received in the current year decreased in 2016 compared to 2015, and in 2015 compared to 2014, primarily due to a decrease in the number of new defaults, net of cures, as well as a decrease in the estimated claim rate on recently reported defaults.

The “Losses paid” section of the table below shows the breakdown between claims paid on new default notices in the current year, and claims paid on defaults from prior years. Until a few years ago, it took, on average, approximately twelve months for a default that is not cured to develop into a paid claim. Over the past several years, the average time it takes to receive a claim associated with a default has increased. This is, in part, due to new loss mitigation protocols established by servicers and to changes in some state foreclosure laws that may include, for example, a requirement for additional review and/or mediation processes. It is difficult to estimate how long it may take for current and future defaults that do not cure to develop into paid claims.

During 2016, our losses paid included $53 million associated with settlements for claims paying practices and NPL settlements. These settlements reduced our delinquent inventory by 1,273 notices. During 2015, our losses paid included $10 million associated with settlements for claim paying practices. These settlements reduced our delinquent inventory by 1,121 notices. These settlements had no material impact on our losses incurred in either year.

The liability associated with our estimate of premiums to be refunded on expected claim payments is accrued for separately at December 31, 2016 and 2015 and approximated $85 million and $102 million, respectively. This liability was included in "Other liabilities" on our consolidated balance sheets.

In each of 2016, 2015, and 2014, we paid $42 million in connection with a 2012 settlement agreement with Freddie Mac regarding the aggregate loss limit under certain pool insurance policies. The final payment under that settlement agreement was made on December 1, 2016.

The following table provides a reconciliation of beginning and ending loss reserves for each of the past three years:



(In thousands)	2016			2015			2014
Reserve at beginning of year	$	1,893,402		$	2,396,807		$	3,061,401
Less reinsurance recoverable	44,487			57,841			64,085
Net reserve at beginning of year	1,848,915			2,338,966			2,997,316

Losses incurred:
Losses and LAE incurred in respect of default notices received in:
Current year	387,815			453,849			596,436
Prior years ⁽¹⁾	(147,658		)	(110,302		)	(100,359		)
Total losses incurred	240,157			343,547			496,077

Losses paid:
Losses and LAE paid in respect of default notices received in:
Current year	14,823			25,980			32,919
Prior years	689,258			823,058			1,121,508
Reinsurance terminations ⁽²⁾	(3,329		)	(15,440		)	—
Total losses paid	700,752			833,598			1,154,427
Net reserve at end of year	1,388,320			1,848,915			2,338,966
Plus reinsurance recoverables	50,493			44,487			57,841
Reserve at end of year	$	1,438,813		$	1,893,402		$	2,396,807


⁽¹⁾	A negative number for prior year losses incurred indicates a redundancy of prior year loss reserves. See table below for more information about prior year loss development.

⁽²⁾

In a termination, the reinsurance agreement is cancelled, with no future premium ceded and funds for any incurred but unpaid losses transferred to us. The transferred funds result in an increase in our investment portfolio (including cash and cash equivalents) and a decrease in net losses paid (reduction to losses incurred). In addition, there is an offsetting decrease in the reinsurance recoverable (increase in losses incurred), and thus there is no net impact to losses incurred. (See Note 9 – “Reinsurance”)

For the years ended December 31, 2016, 2015 and 2014 we experienced favorable prior year loss reserve development. This development was, in part, due to the resolution of approximately 63%, 60% and 58% for the years ended December 31, 2016, 2015 and 2014, respectively, of the prior year default inventory. In 2016, 2015, and 2014 we experienced improved cure rates on prior year defaults. Additionally, during 2015 the claim rate development was favorably impacted by re-estimations of previously recorded reserves relating to disputes on our claims paying practices and adjustments to IBNR. The favorable development for the years ended 2016 and 2015 was offset, in part, by an increase in the estimated severity on prior year defaults remaining in the delinquent inventory. The decrease in the estimated severity in 2014 was based on the resolution of the prior year default inventory.

The prior year development of the reserves in 2016, 2015 and 2014 is reflected in the table below.



(In millions)	2016			2015			2014
Decrease in estimated claim rate on primary defaults	$	(148	)	$	(141	)	$	(43	)
Increase (decrease) in estimated severity on primary defaults	9			43			(35		)
Change in estimates related to pool reserves, LAE reserves, reinsurance and other	(9		)	(12		)	(22		)
Total prior year loss development ⁽¹⁾	$	(148	)	$	(110	)	$	(100	)


⁽¹⁾	A negative number for prior year loss development indicates a redundancy of prior year loss reserves.

Default Inventory

A rollforward of our primary default inventory for the years ended December 31, 2016, 2015 and 2014 appears in the table below. The information concerning new notices and cures is compiled from monthly reports received from loan servicers. The level of new notice and cure activity reported in a particular month can be influenced by, among other things, the date on which a servicer generates its report, the number of business days in a month and transfers of servicing between loan servicers.



	2016		2015		2014
Default inventory at beginning of year	62,633		79,901		103,328
New Notices	67,434		74,315		88,844
Cures	(65,516	)	(73,610	)	(87,278	)
Paids (including those charged to a deductible or captive)	(12,367	)	(16,004	)	(23,494	)
Rescissions and denials	(629	)	(848	)	(1,306	)
Other items removed from inventory	(1,273	)	(1,121	)	(193	)
Default inventory at end of year	50,282		62,633		79,901

The decrease in the primary default inventory experienced during 2016 and 2015 was generally across all markets and all book years prior to 2013. In 2016 and 2015, the percentage of loans in the inventory that had been in default for 12 or more consecutive months had decreased compared to the respective prior years. Historically as a default ages it becomes more likely to result in a claim. The percentage of loans that have been in default for 12 or more consecutive months and the number of loans in our primary claims received inventory have been affected by our suspended rescissions and the resolution of certain of those rescissions discussed below and in Note 17 - "Litigation and Contingencies".

The number of consecutive months that a borrower has been delinquent is shown in the table below.

Consecutive months in default

	December 31,
	2016			2015			2014
3 months or less	12,194	24	%	13,053	21	%	15,319	19	%
4 - 11 months	13,450	27	%	15,763	25	%	19,710	25	%
12 months or more⁽¹⁾	24,638	49	%	33,817	54	%	44,872	56	%
Total primary default inventory	50,282	100	%	62,633	100	%	79,901	100	%

Primary claims received inventory included in ending default inventory	1,385	3	%	2,769	4	%	4,746	6	%


⁽¹⁾	Approximately 47%, 50% and 53% of the primary default inventory in default for 12 consecutive months or more has been in default for at least 36 consecutive months as of December 31, 2016, 2015 and 2014, respectively.

The length of time a loan is in the default inventory can differ from the number of payments that the borrower has not made or is considered delinquent. These differences typically result from a borrower making monthly payments that do not result in the loan becoming fully current. The number of payments that a borrower is delinquent is shown in the table below.

Number of payments delinquent

	December 31,
	2016			2015			2014
3 payments or less	18,419	36	%	20,360	33	%	23,253	29	%
4 - 11 payments	12,892	26	%	15,092	24	%	19,427	24	%
12 payments or more	18,971	38	%	27,181	43	%	37,221	47	%
Total primary default inventory	50,282	100	%	62,633	100	%	79,901	100	%

Pool insurance default inventory decreased to 1,883 at December 31, 2016 from 2,739 at December 31, 2015 and 3,797 at December 31, 2014.

Claims paying practices

Our loss reserving methodology incorporates our estimates of future rescissions. A variance between ultimate actual rescission rates and our estimates, as a result of the outcome of litigation, settlements or other factors, could materially affect our losses.

The liability associated with our estimate of premiums to be refunded on expected future rescissions is accrued for separately. At December 31, 2016 and 2015 the estimate of this liability totaled $5 million and $7 million, respectively. This liability was included in "Other liabilities" on our consolidated balance sheets.

For information about discussions and legal proceedings with customers with respect to our claims paying practices, including settlements that we believe are probable, as defined in ASC 450-20, see Note 17 – “Litigation and Contingencies.”