Mortgage Insurance and Loss Severity This blog post is the first in a two-part series about Mortgage Insurance and Loss Severity. During the implementation of RiskSpan’s Credit Model, which enables users to estimate loan-level default, prepayment, and loss severity based on loan-level credit characteristics and macroeconomic forecasts, our team explored the many variables that affect loss severity. This series will highlight what our team discovered about Mortgage Insurance and loss severity, enabling banks to use this GSE data to benchmark their own MI recovery rates and help estimate their credit risk from MI shortfalls. RiskSpan reviewed the historical performance of Mortgage Insurers providing loan loss benefits between 1999 and 2015. Our analysis centered on Borrower and Lender-Paid Mortgage Insurance (referred to collectively as MI in this post) in Freddie Mac’s Single Family Loan-Level Dataset. Similar data is available from Fannie Mae, however, we’ve limited our initial analysis to Freddie Mac as its data more clearly reports the recovery amounts coming from Mortgage Insurers. Mortgage Insurance Benefit Options Exhibit 1: Mortgage Insurance Percentage Option Benefit Calculation Mortgage Insurance Benefit = Calculated Losses x MI Percent Coverage Calculated Losses include: UPB at time of default Unpaid Interest Other costs, such as attorney and statutory fees, taxes, insurance, and property maintenance. Mortgage insurance protects investors against the event a borrower defaults. Mortgage Insurers have many options in resolving MI claims and determining the expected benefit, the amount the insurer pays in the event of a defaulted loan. The primary claim option is the Percentage Option, where the loan loss is multiplied by the MI percentage, as shown in Exhibit 1. Freddie Mac’s dataset includes the MI percentage and several loss fields, as well as other loan characteristics necessary to calculate the loss amount for each loan. The Mortgage Insurer will elect to use other claim options if they result in a lower claim than the Percentage Option. For example, if the Calculated Losses less the Net Proceeds from the liquidation of the property (i.e., net losses) are less than the Mortgage Insurance Benefit via the Percentage Option, the Mortgage Insurer can select to reimburse the net losses. Mortgage insurers can choose to acquire the property, known as the Acquisition Option. The mortgage insurer acquires the property after paying the full amount of the Calculated Losses on the loan to the investor. There were no instances in the data of Mortgage Insurers exercising the Acquisition Option after 2006. Causes of Mortgage Insurance Shortfalls Freddie Mac’s loan- level dataset allows us to examine loans with MI, which experienced default and sustained losses. We find that there are cases in which mortgages with MI coverage are not receiving their expected MI benefits after liquidation. These occurrences can be explained by servicing and business factors not provided in the data, for example: Cancellation: Mortgage Insurance may be cancelled either by non-payment of MI premium to the insurer or loan reaching a certain CLTV threshold. Per the Homeowners Protection Act of 1998, servicers automatically terminate private mortgage insurance (PMI) once the principal balance of the mortgage reaches 78% of the original value or if the borrower asks for MI to be cancelled once mark-to-market loan-to-value is below 80%. Denial: Mortgage Insurers may deny a claim for multiple factors, such as not filing a Notice of Default with the Mortgage Insurer within MI policy guideline’s time frame, not submitting the claim within a timely period after the liquidation event, inability to transfer title, or not providing the necessary claim documentation, usually from underwriting, from loan origination to Mortgage Insurers at time of claim. Rescission: Mortgage Insurers will rescind an MI claim but will refund the MI premiums to the servicer. Rescission of claims are usually linked to the original underwriting of the loan and might be caused by multiple factors, such as underwriting Negligence by the lender, third-party fraud, or misrepresentation of the Borrower. Curtailment: Mortgage Insurers will partially reimburse the filed claim if the expenses are outside of their MI policy scope. Examples of curtailment to MI claims include excess interest, taxes, and insurance expenses beyond coverage provision of the Master Policy. Most current MI policies do not have these restrictions, non-covered expenses such as costs associated with physical damage to the property, tax penalties, etc., and delays in reaching foreclosure in a timely manner. Receivership: During the mortgage crisis, several of the Mortgage Insurers (for instance Triad, PMI, and RMIC) became insolvent and the state insurance regulators placed them into receivership. For the loans that were insured by Mortgage Insurers in receivership, claims are currently being partially paid (at around 50% of the expected benefit) with the unpaid benefit being deferred. This unpaid benefit runs the risk of not being paid. These factors are evident in the data and our analysis as follows: Cancellations: The Freddie Mac dataset does not provide the MI in force at the time of default, so we cannot identify cases of cancellation. These cases would show up as an instance of no MI payment. Denials & Rescissions: Our analysis excludes any loans that were repurchased by the lender, which would likely exclude most instances of MI rescission and denial. In instances where the Mortgage Insurer found sufficient case to rescind or deny, Freddie Mac would most likely find sufficient evidence for a lender repurchase as well. Curtailments: The analysis includes the impact of MI curtailment. Receivership: The analysis includes the impact of Mortgage Insurers going into receivership. Shortfalls of Expected Mortgage Insurance Recoveries In the exhibits below, we provide the calculated MI Haircut Rate by Vintage Year and by Disposition Year for the loans in our analysis. We define the MI Haircut Rate as the shortfall between our calculated expected MI proceeds and the actual MI proceeds reported in the dataset. The shortfall in MI recoveries is separated into two categories: MI Gap and No MI Payment. MI Gap represents instances where some actual MI proceeds exist, but they are less than our calculated expected amount. The shortfall in actual MI benefit could be due to either Curtailment or partial payment due to Receivership. No MI Payment represents instances where there was no MI recovery associated with loans that experienced losses and had MI at origination. No payment could be due to Rescission, Cancellation, Denial, or Receivership. For purposes of this analysis, the Severity Rate represented below does not include the portion of the loss outside of the MI scope. For example, in 2001, average severity rate was 30%, but only 19% was eligible to be offset by MI. This was done in order to give a better understanding of the MI haircut’s effect on the Severity Rate. Exhibit 1: Mortgage Insurance Haircut Rate by Vintage Years We can observe an MI Haircut Rate averaging at 19.50% for vintages 1999 to 2011 with higher haircuts for the distressed vintages 2003 to 2008 at 23.50%. Exhibit 2: Mortgage Insurance Haircut Rate by Disposition Year Our analysis shows the MI Haircut Rate prior to 2008 on average was 6.5% and steadily increased to an average of 25% from 2009 thru 2014. We will explain below. Exhibit 3: Mortgage Insurance Haircut Rate and Expense to Delinquent UPB Percentage by Months Non-Performing In this analysis, we observe the MI Haircut Rate steadily increased by the number of months between when a loan was first classified as non-performing and when a loan liquidated. This increase can be explained by increased curtailments tied to expenses that increase over time, such as expenses associated with physical damage of the property, tax penalties, delinquent interest, insurance and taxes outside the coverage period, and excessive maintenance or attorney fees. Interest, taxes, and insurance typically constitute 85% of all loss expenses. This analysis of mortgage insurance is an exploratory post into what causes the shortfall in MI claims and how those shortfalls can affect loss severity. RiskSpan will be addressing a series of topics related to Mortgage Insurance and loss severity. In our next post we will address how banks can use this GSE data to benchmark their own MI recovery rates and help estimate their credit risk from MI shortfalls.