IMAGIN a New Way
In a roundtable discussion, A.M. Best experts discuss the framework for the new Freddie Mac mortgage credit risk transfer program, IMAGIN.
- John Weber
- August 2018
IMAGIN, or Integrated Mortgage Insurance, is the Freddie Mac pilot program launched in March to attract additional sources of private capital to support low-down-payment mortgage lending.
A newly established U.S. subsidiary of Arch Capital Group, Arch MRT, manages the mortgage credit risk transfer program, insuring Freddie Mac and then auctioning the assumed risk to a panel of (re)insurers. The long-term objective of IMAGIN is to lower the cost of mortgage insurance for borrowers.
A.M. Best analysts addressed questions surrounding the program as part of an A.M.BestTV webinar, Evaluating Mortgage Risk in Reinsurance.
A.M.BestTV’s John Weber hosted the discussion led by Emmanuel Modu, managing director, insurance-linked securities, with Dr. Wai Tang, director, and Steven Chirico, director.
Following is an excerpt from the webinar.
What Is IMAGIN?
Modu: Let’s examine the framework for risk charge and reinsurers that participate in the IMAGIN program, which is a program that is offered by Freddie Mac.
It’s a pilot program and it’s called Integrated Mortgage Insurance, or IMAGIN.
In that program, reinsurers effectively provide mortgage insurance to mortgages held by Freddie Mac.
The program is associated with lender-paid mortgage insurance, and as far as we know, lenders will sell low-down-payment mortgages for Freddie Mac, and at the same time, Freddie Mac gets mortgage insurance coverage from a panel of pre-approved reinsurers managed by Arch MRT, which is an Arch Capital subsidiary.
The reinsurers post some level of collateral, generally based on their credit profiles. There are 12 participating lenders in the pilot program. [At the time of IMAGIN’s launch in March] we have some questions about the program. We haven’t gotten the full details yet, but some of the initial questions we have relate to the underwriting process, for example.
We want to know whether the reinsurers have taken on the same risks as the PMIs are taking. Is there any adverse selection in the mortgages they’re covering? What is the coverage? Are there any risks covered by PMIs that the reinsurers do not cover? Are there any risks covered by the reinsurers that the PMIs do not cover?
Are reinsurers required to participate for more than one year? What is the universe of mortgages covered? Is it similar to the usual PMI universe? And will the data on the reference pools be publicly available so we can do our own analysis on the reference pools?
If the reference pools are standardized, we may be able to use a factor-based approach. If not, we may be able to use a credit risk model like LoanKinetics. If we use the factor-based approach, we need to create a new matrix. We need a new SUL [stressed ultimate loss] matrix from a mortgage insurer’s perspective. And for that, we need the default frequency and contractual coverage amounts for our analysis.
Using the SUL matrix and our distribution of principal balances, we can calculate the ultimate loss associated with the reference pool. And these are the ultimate losses faced by the reinsurers. Remember that our reinsurers are now behaving like mortgage insurers.
Once the ultimate loss is calculated, we can then follow the same general steps we use for calculating a B5m from the CRT programs. An issue that should be discussed at some point is whether we include B5m in the analysis.
To calculate an SUL matrix from the perspective of a mortgage insurer, we need a couple of things. We need a default frequency table bifurcated by a loan-to-value ratio and credit score. We’ll call that Table A. We also need a table of the contractual amounts covered by the mortgage insurers in the event of defaults of the mortgages.
We’ll need that bifurcated by loan-to-value ratios and credit scores, as well. We’ll call that Table B. And a cell-by-cell multiplication of the two tables will give us the matrix we need, really, to assess the risks faced
by reinsurers who are behaving like mortgage insurers.
Here’s an example. We’ve created a hypothetical frequency matrix, and this matrix is based on the 2007 single family loan level data set from the GSEs. We’ve modified it a bit for our purposes, but in this matrix, the frequencies are going up, unlike the SUL matrices you saw earlier, where you had losses going up and then going down. Here, there’s no embedded mortgage insurance, frequencies are increasing from the lowest LTV to the highest LTV.
The next table you need is the assumed contractual coverage by the reinsurers. This is the percentage of losses that the reinsurers were going to hope for if the mortgages in the reference pools that they cover default.
We created a hypothetical table for this. We don’t know the actual numbers, but we assume these are probably not that far off.
When we multiply the new loss matrix we created with the new covers percentage, we get then a new SUL matrix, which is Table C. This is the table we need to assess the losses associated with mortgages covered by the reinsurers who are behaving like mortgage insurers.
How do we use this table? We first have to get a distribution of the underlying principal balance of the reference pool. Here, we assume a reference pool of about $14 billion. It’s actually a real transaction. And then, we’ll distribute the balances by credit score and loan to value ratio.
We’ll take this table and multiply it by the hypothetical loss matrix, and then sum all the balances up to get what would be the stressed ultimate loss associated with the pool covered by the reinsurers. In this case, it happens to be 3.44%.
Given the stressed ultimate loss, we calculate the present value of the loss developments, and then we net the premium credit in order to calculate the risk charge.
The details of the calculation will be the same as with the CRT transactions and as fully described in our new criteria, “Evaluating Mortgage Insurance.” We’ve shown how to calculate B5M for the CRT programs, for the reinsurers of PMIs, and for other reinsurance programs, as well as a framework for calculating B5M for the imagined program. How do we then apply B5M to our BCAR model, which is our capital model?
Applying to a Model
Our BCAR model has got two main components. It’s got the available capital components and net required capital components. We have to adjust these items to account for mortgage risks. Here’s how we do it.
On the left panel, you see the components of available capital for most property and casualty companies. We add back continuous reserves and nonrefundable single premiums to available capital.
On the net required capital side, we correlate our mortgage-related reserves risk with investment risk. The correlation factor is about 50% for both fixed-income securities and equity securities. We also correlate in the B5 row the mortgage-related reserves and the nonmortgage-related reserves. The correlation factor we use is 10%.
This slide shows the interaction of the various underwriting risk elements. You can see that we add B5CM and B5FM to get B5M. Then, we put a 10% correlation between B5M and the nonmortgage-related reserves. B5 and B6 have zero correlation between them per our NRC formula.
This is the new net required capital formula. You see our criteria.
It incorporates the correlation between mortgage-related reserves risk and investment risk which is the first box. It also incorporates the correlation between mortgage-related reserves risk and nonmortgage related reserves risk which is already embedded in the B5 number. If there’s no mortgage risk, this formula reduces back to our normal NRC formula for property and casualty companies.
This new formula is quite critical in the calculation of a BCAR score which is the critical component in our balance sheet strength analysis.
Four Main Verticals
Chirico: Manny and Wai just spent a lot of time talking you through how we calculate what is, in essence, an increase in net required capital to be input into our capital model BCAR and to result in the assessment of balance sheet strength. In a lot of ways, this type of analysis resembles what we do for terrorism stress tests and PML stress tests. In some ways, it doesn’t resemble those things.
It resembles those stress tests, our stress ultimate loss here for mortgage, like a PML and like a terrorism stress test, in that it is making a company hold capital for an event that we’re not sure when it’s going to happen, but we are certain will happen at some point in the future.
It is unlike a PML in that it is part of standard BCAR. If a company writes property catastrophe reinsurance and they write mortgage risk, we assume 100 correlation such that the stress test for mortgage is part of standard BCAR. Then, the property cat stress test is additive to that from a net required capital perspective.
I’m going to talk about—switching gears now—the four main verticals in the interactive rating process from an analyst’s point of view.
Assuming that balance sheet strength vertical is now taken care of by the increase in net required capital maintained in the BCAR model and then qualitatively from a quality of capital perspective analyzed in that vertical. The second vertical that I want to talk about is operating performance. We recognize that mortgage reinsurance can be very lucrative from a combined ratio perspective and a return-on-capital perspective for a reinsurer.
However, we also understand that there’s relatively extreme volatility in the operating results of the mortgage reinsurance business over time. It seems to be unpredictable as to when it goes bad. When it goes bad, it goes bad very quickly.
The usual pillars of how we evaluate operating performance are the same as any other book of business or line of business that a reinsurer writes except for the fact that the volatility is heightened. To illustrate that point somewhat, we put together a table here that you can see takes us through time from 2006 to 2016.
You can see from a loss and LAE ratio perspective very extreme differences in operating performance depending on what year you’re looking at. Like property catastrophe reinsurance that may have years where there’s very low loss ratios, they’re not quite as low in mortgage insurance when there’s no risk.
In the catastrophe situation, the loss and LAE ratios tend to be higher unless the reinsurer absolutely doesn’t know what they’re doing from a property catastrophe perspective.
We incorporate that part of the analysis when we’re looking at operating performance to recognize that we’re looking at things like return on equity and combined ratios over time.
Before a company raises their hand and says, “Well, we wrote mortgage insurance at a 45 combined ratio, and that brought our consolidated combined ratio down by 20 points. Can we have plus one or plus two on our new BCRM?” The answer is no because the way we’re looking at it is that we’re looking at it through time.
We’re taking more of a weighted average view of operating performance similar to what we do with property catastrophe insurance and reinsurance.
Switching over to business profile, which is the next vertical in the interactive rating process, I want to bifurcate my comments between the profile of the actual mortgage book of business being written, as well as talking about how mortgage affects the business profile of the entity consolidated both on the liability side of the balance sheet and the asset side of the balance sheet.
In a mortgage scenario, we’re looking for a distribution of risk enforced by vintage. We’re going to ask companies, “Are you opportunistically writing certain vintages or are you, for lack of better term
s, dollar cost averaging over time your mortgage exposure?”
What we’re looking for here is the diversification to be able to smooth out some of the bumps that we may experience from the results of the mortgage business.
The second bullet, mortgage loan characteristics such as loan age, original loan of value, credit scores, delinquency status, and on and on and on.
What we’re interested in here is, is there a pointed opportunistic play in the mortgage market or is there a diversifying quality in the mortgage reinsurance being written by a reinsurer?
Then lastly but probably most importantly, is the geographic scope of the mortgage guaranty being written by a reinsurer. What we’re looking for here is we understand that there can be hot pockets of downside risk in the mortgage reinsurance products being written, particularly in the PMI space, by reinsurers. We’re looking for diversification so that some of those bumps from a geographic perspective can be ironed out.
We’re also interested in how mortgage affects the business profile of a rated reinsurer. It affects it in two ways. It brings something else to the party from a liability perspective and other lines of business being written.
We are basically assuming that there’s a relatively low correlation between mortgage reinsurance and other types of reinsurance with the possible exception of workers’ comp which tends to be more highly correlated to the economy.
That will diversify the liabilities of the entity. It will provide relief assuming that the mortgage reinsurance business stays in the profitable space that it is currently. It’s accretive from a diversification perspective.
From an asset perspective, we recognize that depending on the type of asset in the investment portfolio of a reinsurer, it can have a material effect on how we view both the quantitative and qualitative risk from a business profile perspective that mortgage reinsurance brings to the entity. Let’s take two examples.
The first example would be a reinsurer that writes a well-diversified book of business from a liability perspective including mortgage and has a relatively safe, low volatility asset portfolio. Let’s say sovereign triple A-rated securities and high rated corporate bonds.
What drives, in essence, some of the qualitative and quantitative measures that we use to assess what mortgage brings to a reinsurer is not necessarily the volume of assets, but the risk and the increase in net required capital those assets attract.
In the example that I gave you, it’d be a relatively low increase in net required capital for those types of assets and would have a resultant lower increase in required capital from the mortgage risk being assumed.
Compare that and contrast that to another entity that’s taking much more risk in the form of a basket of S&P equity securities, alternative assets, hedge fund strategies, private equity, real estate, etc.
Because the net required capital on those more risky assets is higher, it will attract more qualitative and quantitative risk as far as BCAR and the overall analysis of the reinsurer from the introduction of the mortgage risk.
Distilling that down, what we’re saying here is that we’re being concerned about how correlated a reinsurer becomes to the economy in general, what the beta to the economy is, with the introduction of mortgage risk.
As we all know, mortgage risk is highly correlated to the economy, as well as certain risky assets. That is informing and driving our analysis both on a quantitative and on the qualitative side.
Enterprise Risk Management
The last pillar we are going to discuss as far as the rating of a reinsurer is enterprise risk management. In a lot of ways, this is the most important of all the verticals that we’re going to be examining. In a lot of ways, enterprise risk management for mortgage is the same as any other line
of business or book of business written by a reinsurer. We expect written procedures, controls, and safeguards to ensure sound underwriting decisions. That’s across the board. That’s the same.
We expect detailed and up-to date exposure information for an accurate assessment of potential claims associated with the loan portfolio and avoidance of concentration of its risk in force and a quality control program that assesses the effectiveness of the overall insurance business including risk selection, rescission rights, and loss mitigation practices.
What we’re expecting in addition to these enterprise risk management tenets for a reinsurer that writes mortgage risk is the establishment of a well-thought-out early warning system. We recognize that operating results and the capital allocated toward claims will increase precipitously when a catastrophe hits. We saw that in 2008. We saw it in the early ’90s. We saw it in the ’80s and in the 1930s.
The early warning system is going to be key so that the reinsurer can start minimizing the amount of mortgage risk they’re assuming as rates decay, as terms and conditions decay, and as the economy decays vis-à-vis real estate and the mortgage insurance product. What we’re looking for here is again a well-thought-out action plan to respond to these early warning systems that a reinsurer has in place.
That is probably one of the most important things from a rating analyst perspective to make sure their reinsurer has thought about all these things, has incorporated that into their underwriting platform, their underwriting guidelines, and their risk management policies and procedures.
Here’s a laundry list of required information. I’m going to stress and talk about, a little bit, points one and five in particular.
The first question is why are you writing mortgage insurance or reinsurance? We’re very interested how this fits into the greater scope of the reinsurer. We’re looking at [the question,] is this more of an opportunistic play?
If so, we’ll treat it as that. We’ll expect the reinsurer to have controls in place to address that. It’s more part of a larger, longer-term diversification strategy that will also drive some questions and some assessment, particularly in the enterprise risk management regime.
Is what is the organization’s tolerance for mortgage risk? How do you measure it? What correlative assumptions are being considered by the reinsurer?
We expect a highly interactive discussion between the modeling that Manny and Wai spoke to you about, our discussion with companies over enterprise risk management and business profile, and the operating returns.
Then, how the reinsurer thinks about all of those things and how from an organizational perspective they have set risk tolerances, they have established early warning systems, and they have either grown or decreased the amount of mortgage business that they’re writing in any one particular point in time and the reasons thereof.
We expect a highly interactive process. Frankly, the reinsurance organizations where we have been through this process with that have been actively writing mortgage for a period of time now, we find there’s an extreme iterative process, more so than any other line of business.
Other additional requested information, we’re asking for all of the reinsurance agreements associated with U.S. PMI and non-U.S. PMI. The reinsurance agreement on these types of coverages, it’s different than the CRTs where information’s publicly available. We want to see what reinsurers are actually writing, we’ll understand the underlying risk from a loan perspective.
But certainly from a reinsurance contract perspective, we’re going to want to scrub that reinsurance agreement and have discussions with management over how they think about those.
We’re certainly go
ing to talk about the stress levels applied to non-U.S. businesses. Australia’s listed here. But we’re aware companies are writing some European mortgage products and maybe potentially around the world.
Australia, for instance, we understand, is a commodity-driven economy. They have different pressure points and relief valves written into their economy and into the mortgage products that are offered thereof. We’re going to want to get an understanding of that, how the company models that and how they think about that.
We’re going to also have discussions about the precise definition of limits. Early on, maybe about a year and a half ago, we started surveying reinsurers about what types of mortgage reinsurance they were writing, how they were writing it. We asked for limit information. What we found we got back was different definitions of limit.
We are going to ask companies for their definition of limit. We’re then going to right size that so that it’s comparable against companies so that when we do our comparative analysis between reinsurers, we have a very clear definition of what the limits are, what the exposure the reinsurers are assuming is.
Then finally, loan level detail where appropriate. I’ll be frank and tell you that in typical A.M. Best fashion, we’ve been methodical and conservative with how we make assumptions in the rating process. Loan level detail can offer some incisive contemplation of how we’re viewing mortgage in a particular portfolio and can lend some details as how we should look at that.
We understand that we’re being very conservative. We understand this market can be highly volatile, both from a capital perspective and operating performance perspective. A lot of these bits of information that we’re gathering from a company help us identify where a company would fit from a comparative analysis perspective compared to other reinsurers.
John Weber is a senior associate editor. He can be reached at firstname.lastname@example.org.