Webinar Highlights
The Rise Of MILS

• John Weber
• August 2018
• 

Mortgage-related insurance-linked securities transactions are relatively new to the market.

Like natural catastrophe ILS transactions, they are collateralized securities and used as a substitute for traditional reinsurance. Several of these transactions have been sponsored by private mortgage insurers.

A.M. Best analysts examined how A.M. Best evaluates mortgage-related ILS transactions and the reinsurance credit to private mortgage insurers that sponsor MILS in the webcast, Rating Mortgage Insurance-Linked Securities.

A.M.BestTV’s John Weber hosted a discussion led by Emmanuel Modu, managing director, insurance linked securities, with Dr. Wai Tang, director.

Explaining Mortgage Insurance

Modu: Mortgage insurance is a type of credit enhancement for mortgages with loan-to-value ratios of over 80%. The loan-to-value ratio is a ratio of the loan balance to the market value of the home. MI is most often used to credit enhance mortgages sold to Fannie Mae and Freddie Mac. An important factor in analyzing mortgage insurance is risk in force. This is essentially the maximum exposure faced by mortgage insurers that cover losses in a pool of mortgages.

Risk in force is a product of the MI coverage percentage and the unpaid principal balance of the mortgage.

Let’s go through an example of mortgage insurance. Here you have a borrower who takes out a loan for $360,000 based on their property value of $400,000. A loan-to-value ratio for this mortgage is 90%. The MI promises to pay 25% of the first loss upon a default of the mortgage. The risk in force here is $90,000.

Sometime later, the borrower defaults on the mortgage. There’s a foreclosure process that the mortgage goes through. At the time of default, the unpaid principal balance is $350,000, but there are other expenses to be added to that claim basis for the mortgage insurance. You have unpaid accrued expenses, accrued premium expenses and foreclosure expenses.

All together, the claim basis for this mortgage is $365,000. A quarter of that is $91,250, which is close to the risk in force, but it’s slightly higher because of all of the other expenses. Risk in force is not a great measure for the stress losses an MI company will experience in a pool of mortgages.

The insurance in force of the six current U.S. mortgage insurers is about $960 billion through the end of 2017. Their risk in force is $242 billion through 2017. The ratio of the risk in force to the insurance in force is 25.26%. That’s their MI coverage percentage on the average. Mortgage insurers seek reinsurance through the traditional reinsurance market.

The gross premium written ceded to third-party nonaffiliated insurers has grown. From 2015 to 2017, that ceded amount has grown considerably. In 2017, close to 15% of gross written premiums by the MIs have been ceded to third-party reinsurers.

The other market through which MI companies seek reinsurance is through the capital markets through MILS transactions. These type of transactions began in 2015. You can see that in 2018, there have been two transactions so far of close to $800 million. In total, that’s close to $2 billion of risk in force ceded to the capital markets.

This is really a tiny little fraction of the total risk in force of the six MI companies. It really is only about, I think, 80 basis points of their risk in force. In terms of the companies that are involved in these deals, it’s about 2% of their risk in force. It’s a very tiny amount.

These are the six transactions we’ve seen so far. The first transactions were done by Arch/UGC. Arch and UGC have been involved in four transactions. National Mortgage Insurance has done one transaction, and Essent Guaranty has sponsored another transaction. The maturity of these transactions is generally about 10 years.

Structure of MILS Transactions

Let’s now talk about the general transaction structure of the MILS transactions. We made up a hypothetical pool, which has a feature similar to the features we see in other MILS transactions.

We highlight the current risk in force of about $10.2 billion, the average of mortgage insurance coverage, about 25.5%, which is close to the average MI coverage for the MI industry. We have the weighted average loan-to-value ratio of close to 92%. It’s 91.6% here.

The weighted average credit score is quite high, it’s 748. Ninety-eight percent of the pool is made up of fixed-rate mortgages.

This is the liability structure of the transaction. The Class 3 note is not attached until the ceding insurer absorbs the first $230 million of losses. This means that the Class 3 subordination level is $230 million, which is about 2.25% of the risk in force.

The Class 2 note is not attached until after the ceding insurer absorbs the first $230 million of losses and the Class 3 note is wiped out. This means that the Class 2 subordination level is $255 million or about 2.5% of our risk in force.

The Class 1 note is not attached until after the ceding insurer absorbs the first $230 million of losses and the Class 3 and Class 2 notes are wiped out. You can say that the risk of the subordination level for Class 1 is about $470 million or 4.6% of the risk in force.

Thereafter, after the exhaustion of Class 1 notes at the attachment of the coverage level of the losses, after the Class 1 note is exhausted is absorbed by the ceding insurer. In this case, the ceding insurer does not absorb losses again until the coverage Level B is exhausted, which is the ceding insurer’s first loss position, and Class 3, Class 2, and Class 1 note is exhausted.

The subordination level for the ceding insurer that covers Level A is $740 million or so. That’s a subordination level of 7.25%. This subordination level is important later on in terms of the trigger of our principal allocations we’ll discuss later.

In terms of loss allocation, it’s what we just described. The loss allocation begins from the lowest tranche, which is the coverage Level B absorbed by the insurer. It goes up to Class 3, Class 2, Class 1, and thereafter the loss is absorbed by the mortgage insurer. The hierarchy for the principal allocation is a bit more complicated. Let’s first define a couple of coverage levels.

The first coverage level is coverage Level A. It’s $9.47 billion of the risk in force. The remaining coverage level, which we’ll call RCL, is about $741 million. The principal is allocated on a prorated basis between CLA, coverage Level A, and RCL, but within RCL, principal’s allocated sequentially.

For every $100 that comes in through principal amortization and prepayments initially, $92.75 goes to coverage Level A and the remaining $7.25 goes to pay down Class 1 first, and after Class 1 is gone, the Class 2 and so on. You can see that the liability structure allows for a paydown of the coverage level as well as the notes themselves.

The price line, the paydown in our principal is based on when things are going great. When things are going poorly, all of the principal allocation goes to coverage Level A. The Class 1, 2, and 3 gets nothing until the conditions are great again.

What do we mean by when things are going great or not so great? When things are going great, it means that the 60-plus days delinquency rate is less than 5% and the subordination level for CLA is greater than or equal to 7.25%.

This means that the allocation of principal goes the way we described. That is, the coverage Level A and RCL get a prorated share of our principal. If things are not going great, it means that the 60-plus days delinquency rate is greater than or equal to 5% or the subordination level for coverage Level A is less than 7.25%. In that case, all the principal is allocated to coverage Level A.

Rating MILS

Tang: In this section, we will talk about the A.M. Best approach to rate the mortgage insured linked securities. As MILS are securities, we will assign long-term issue rating, or IR, to the notes.

The IR ratings has a 21-point scale from AAA to C.  AAA is the highest credit, while C is the lowest one. The methodology used to rate the MILS is Best Insured Linked Securities and Structures Methodology, or BILSM, while the criteria used is evaluating mortgage insurance.

The information that A.M. Best will be reviewing for the rating purpose is pretty standard in the liability structure transactions.

In this type of transaction, loan level data’s always available, so A.M. Best would use a third-party credit model, the LoanKinetics from Andrew Davidson, to calculate model mortgage loans related to the reference pool. LoanKinetics takes loan level data as input and output the monthly loss number at various confidence levels.

The key risk factors that are included in the model can be broadly categorized into either micro factors or macro factors. Micro factors are loan level data. Among some of them are the credit score and loan-to-value ratio.

Macro factors are macroeconomic factors. They are home price appreciation and interest rates. These two macro factors drive the default rates’ severity, peak payment rates and transition rates. The output from the models are results corresponding to 20 different scenarios.

The scenarios that we are most interested in are the base scenario and some of the stress scenarios. Each scenario is correlated with a cumulative distribution function or CDF. CDF can be interpreted as the probability that the associated or less severe scenarios will happen.

The CDF of the base scenario is the 50th percentile, the median. Stress scenarios are scenarios with CDF greater than 50th percentile. In the range of 80 or higher percentiles are the ones that we are interested in.

Stresses that are applied to the base scenario are economic stresses and model stresses. Economic stresses are stresses on macro factors. In the interest rate case, it will be the perilous shift of the whole U-curve upward. In terms of home price index, the home price will fall down a certain percentage and then come back gradually, while in the base scenario, the home price index is always going up.

The other stresses are model stresses, which take into account a defect of model error. Therefore, it is stressing the parameters used in the equations to come up with the default rates, loss severity and prepayment rates which give you back both micro and macro factors.

Besides that, there is also a global stress which can be applied to all scenarios. That includes base scenarios. This global stress is different from stresses that we were just talking about here. The stresses that we were talking about here are used to bring us from the base scenario to higher CDF scenarios or stress scenarios.

To determine the rating of the security quantitatively we will run the reference portfolio through eco model LoanKinetics. Global stress, that we just mentioned previously, will be applied to all scenarios. The rating committee may ask for higher or lower global stress.

At the subordination level of each class of securities is the market’s loss level. That the principal of that class of notes start to suffer, we’ll find in the scenario that its cumulative loss in 10 years’ time is equal to the subordination level of that class of note.

After that, we will determine the period in which the losses hit each class. We call this the “time to attachment.” In this presentation, we assume that it is always 10 years, the legal maturity of the notes. In rare cases, the time to attachment can be less than legal maturity, especially for the Class 1 notes. However, it is quite technical. I’m not going to go into details here.

From the CDF associated with the class of notes we can calculate the default probability of that class of notes as one minus CDF. From both the time to attachment and the probability of default, we can correlate that to the Best’s Idealized Issue Default Matrix to obtain the implied rating.

Please be aware that this implied rating is only derived quantitatively. There are other qualitative considerations needed to be taken into account before getting to the final rating. The final rating is determined by a rating committee after taking into account of various risk factors.  As we mentioned, other than quantitative considerations, qualitative considerations are very important in the rating process. Some of the items we will consider are the credit risks of the sponsor paying premium through SPRV. However, in some cases or in some deals the transaction structure may mitigate this risk and that will not be applicable then.

Another item we want to emphasize is the quality control of the origination of mortgage loans. During the financial crisis in 2007, one of the reasons that mortgages did not perform well was the quality of the loan originated during that period was pretty poor.

Reinsurance Credit

Modu: While the rating of the MILS transactions are very important and interesting, we believe that what PMIs most care about is probably the amount of reinsurance credit they’ll get for sponsoring these transactions.

As many of you know, A.M. Best has a capital model called BCAR, which helps evaluate balance sheet strength. The BCAR is made up of two main components, which is available capital and net required capital. An MI that sponsors a MILS transaction will reduce their net required capital.

The formula for net required capital has several components, fixed income, securities risk, equity securities risk, interest rate risk, credit risk, reserves risk, net premiums written risk, business risk and potential catastrophe losses. These items are plugged into the net required capital to calculate the NRC.

The NRC is calculated for five VaR levels according to our methodology. The VaR levels are 95, 99, 99.5, 99.6, and 99.8. Ultimately an MI that sponsors a MILS transaction will have its B5M, which is reserves risk, reduced somewhat.

Let’s take an example.

B5M insurer represents the reserves risks of the MI before any reinsurance. Now some of the risk may be retained by the MI company and some may be ceded to the SPRV through the MILS transactions. Our task is to figure out what B5M will be, which is that the reserves risks are retained by the insurer after ceding to the SPRV.

We have to calculate that. Once we calculate that, we can then plug it into the NRC formula to calculate the BCAR ratio. Generically, how we calculate reserves risk is we begin with the reference pool. We calculate the losses in the reference pool and we determine how fast the losses develop in that reference pool.

Based on that, we can calculate the premiums accruing to the pool based on how fast the premiums are accruing to the pool. Then, we can also calculate the reserves risk as a PV of the losses associated with the reference pool less the credit for premiums. Specifically how we calculate reserves risk for the MILS transaction is that we begin by looking at the gross losses at each broad level.

From that, we can also determine the ceded losses at each broad level. We can then calculate the present value of the ceded losses in the MILS transaction for each broad level. Then, we can calculate the PV of the premiums that are accrued to the ceded losses in the MILS transactions at each broad level. Now, the premiums are spread over an index that’s paid to the noteholders.

To calculate the reserves risk for the MILS transaction, which is B5M MILS, it’s simply the difference of item three and item four. This is a representation of what we just discussed. On the right-hand side, you have the ceded discounted loss, which is effectively the loss absorbed by the noteholders in the MILS transaction at each broad level in the reinsurance transaction.

We have to subtract from that the premium ceded by the MI to the SPRV at each broad level. The difference of those two gives you a B5M MILS, which is the reserves risk associated with the business ceded to the SPRV in the MILS transaction at each broad level.

We’re trying to solve for B5M MILS, obviously.  To refresh your memory, this is a transaction structure. We begin with a risk in force of $10.2 billion. The note balance initially is $511 million. The first loss position is $230 million. The coverage Level A has got $9.47 billion. Let’s take a look at a couple of VaR level analyses.

At the VaR 99 level, we can calculate ceded nominal loss and gross nominal loss based on the CDF from our model. You can see here that we’ve done so. We have the gross nominal, which does not exceed the detachment point of the Class 1 notes and we have the ceded nominal, which is really what we want.

The present value of the ceded nominal loss less the PV of the premiums is the reserves risk at the 99% VaR level. Notice, also, that the coverage Level A has gone down considerably. Initially, it was $9.47 billion. It is now at this VaR level $4 billion or so, because of the principal amortization.

The principal allocations are going mainly to the coverage Level A. Some allocations are going to the Class 1, because it’s gone down from $271 million to $228 million, but not much. The rest of the classes are maintained. The balances are maintained. At the VaR 9.8 level, the same thing applies.

John Weber is a senior associate editor. He can be reached at john.weber@ambest.com.