If there is no interest rate reduction, MPI will cost the insurer more than TMI, but not much more, and the investor will always incur a smaller loss with MPI.

Note that in moving from the prime loan to the Alt-A loan, the TMI premium rose by only .62% while the interest rate premium rose by 3.875%. The increase in risk premium charged by the lender was more than 6 times larger than the increase charged by the insurer, despite the fact that the increase in risk exposure was substantially higher for the insurer because the insurer is in the first loss position.

Extending our model, loss to the insurer occurs if the property value appreciates by less than 30%, whereas loss to the investor doesn’t occur unless the property value declines by more than 2.5%! The property value has to decline by 35% before the investor’s loss equals the insurer’s loss.

On the assumption that the mortgage insurance premium accurately reflected the losses expected over a long time horizon, the interest rate risk premium was grossly excessive.

Interest rate risk premiums are excessive mainly because they are not reserved and depend on investor sentiment that is heavily influenced by current market conditions – as opposed to long-term actuarial loss experience. When losses escalate, as they have during 2007-2008, the prevailing view is that the interest rate risk premiums charged borrowers in prior years must have been too small, which results in marked price adjustments. Interest rate risk premiums are now substantially larger than they were earlier, and eligibility cutoffs, where loans become unavailable at any price, occur at lower values of risk variables.

This reaction by the market is understandable, perhaps unavoidable in the current environment. But relative to what they would be in a reserving environment, interest rate risk premiums are grossly excessive. In a system in which insurers offer MPI and all borrowers pay the prime rate, the interest rate plus MPI insurance premium paid by non-prime borrowers would be substantially smaller than the interest rate plus TMI insurance premium they pay now.

Perhaps the best way to answer this is to place ourselves in the position of a price-setting investor, such as Fannie Mae, and ask the following question: assuming Fannie is willing to price a prime loan at 6% with TMI, should it be willing to take a riskier loan at 6% if it carried MPI and if the insurer’s credit is beyond reproach?

It should. The insurer’s credit is better than that of the prime borrower. Further, in the event of a default, the payments will continue to be made on the riskier loan but not on the prime loan. While the riskier loan will have a higher incidence of default, in the event of default the loss to the investor with MPI is lower. How these balance out is not clear, but if there is any added risk, the coverage can be adjusted to shift it to the insurer, with the borrower paying for it in a higher insurance premium. Any such adjustment would not materially reduce the borrower’s cost saving.

An MPI-based mortgage system would eliminate interest-rate risk-based pricing because MPI would be the lowest-cost way to protect against default risk. All loans would have MPI except those that are prime without it, or not insurable. Loan to loan differences in default risk would be reflected in differences in mortgage insurance premiums.

In this system, financing costs to non-prime borrowers would be substantially lower. Further, the system would be much less vulnerable to default crises such as the one we are in now.

Lower Costs to Borrowers: Mortgage insurers assume almost 100% of the default risk under an MPI policy. A very small amount remains because of the cap on insurer liability. Assuming the cap is adjusted to meet investor requirements, the only material risk remaining to the investor is the risk that the insurer itself will fail, as discussed below. Assuming that risk is nil, interest rate risk premiums disappear. Borrowers would pay different mortgage insurance premiums, but they would all pay the prime interest rate.

MPI would cost insurers little more, and in many cases less than their traditional limited insurance. Hence, the total financing cost to borrowers would drop, with the cost imposed on riskier borrowers dropping the most.

The case in tables 1-3 above illustrate how large the savings can be. Assuming the TMI insurance premium of 1.29% in Table 1 is properly priced to meet losses under that policy, it is more than adequate to meet the lower losses under an MPI policy. Hence, the 3.875% rate premium, which investors require when they are protected only by TMI, is redundant if they have MPI.

Further, with all borrowers eligible for mortgage insurance paying prime rates, the potential for predatory practices would be sharply reduced. Elimination of risk-based pricing would eliminate opportunistic pricing of mortgages at the point of sale, which is one of the most important sources of abuse.

In addition, borrowers would have an important ally in the mortgage insurers, who have a financial interest in seeing that borrowers are not over-charged. Higher rates mean greater risk exposure for the insurer. Insurers would have the clout and information needed to protect borrowers because the basic paradigm of insurer/lender relations would be reversed.

Since the private mortgage insurance industry began, it has been beholden to the lenders because lenders select the insurers to whom they refer borrowers. With MPI, lenders and insurers (and by extension, borrowers) would be on a more equal footing because borrowers could go to insurers first knowing that MPI is a de facto loan approval that will allow them to borrow at the prime rate.

Reduced Systemic Vulnerability: With default risk covered by MPI, rather than by a combination of TMI and rate risk premiums, vulnerability to financial crises would be substantially reduced. Today, only TMI premiums are placed in reserve accounts to protect against future losses. With minor exceptions, interest rate risk premiums not needed to meet current losses become investor income. With MPI replacing rate risk premiums, the process of reserving for contingencies would be extended to cover all default risk, not just part of the collateral risk. As a rough order of magnitude, reserves available to meet losses might be ten times larger.

In addition, risk underwriting would shift into more dependable hands. Mortgage insurance companies already offer underwriting to lenders as a service, but with MPI they will do it for all loans except those that don’t qualify for MPI.

In setting underwriting requirements, lenders and investment banks have a short run orientation that can lead to sharp swings in how liberal or restrictive the requirements are. They become excessively liberal when market sentiment is euphoric, as it was during 2000-20006, and then excessively tight when pessimism reigns, as is the case now. This tendency is encouraged by their ability to pass along most default risk to the next party in the chain. Insurers, in contrast, have a long-term orientation because they remain on the hook for a loan until it is repaid or the insurance is terminated.

In addition, by keeping defaulted mortgages performing until they are paid off, MPI would block the contagious erosion of investor confidence that stems from increasing numbers of non-performing loans. This has been a central feature of the current crisis.

A Gaming Analogy: Imagine a world where all home mortgages are placed in securities, of which there are two types. Both promise to pay the investor the prime rate, but protect them in different ways. On an RRP security, borrowers pay interest rate risk premiums, which are placed in a reserve fund, with each security having its own fund. On an MPI security, protection is provided by MPI policies on all the mortgages.

Make the following assumptions: Every security faces a market environment that is determined by a single twirl of a roulette wheel, which has 15 slots, 14 of them blue and one red. If blue comes up, as it will 93.3% of the time, the environment is one in which house prices increase by 5% a year for the life of the security, and investors lose 1/10 of 1% of loan balances. If red comes up, as it will 6.7% of the time, prices will decline by 5% a year for 4 years before leveling out, and losses will be 6% of loan balances.

In this world, the reserve needed by insurers to cover losses of .10% on 93.3% of the loans they insure, and losses of 6% on 6.7% of them is about .50%. Their premium will be about twice this, or 1%. When an MPI security comes up red, they have the reserves to meet the claims.

In contrast, every RRP security stands on its own. For investors to be fully protected against the possibility that the roulette wheel comes up red, the reserve must be 6% on every security. This is too high to be workable. But RRP securities may nonetheless be created if MPI securities are not available.

If borrowers were charged a risk premium of, say, 3%, the RRP security would be extremely profitable for all concerned 93.3% of the time. If the wheel comes up blue 6 times running, which with our assumed probabilities will happen about 66% of the time, issuers and investors may come to believe the wheel has been fixed so that only blue comes up – a phenomena that Guttentag and Herring referred to as "disaster myopia". [See Upheaval in the Sub-Prime Mortgage Market] If the agencies that investors depend on to tell them whether or not a security is safe share the myopia, basing their analysis on the last 6 turns of the wheel, the stage is set for disaster when the wheel comes up red. This is a pretty good description of what actually happened during 2000-2007.

The gaming analogy illustrates very well why interest rate risk premiums are both two large and too small. They are too large in the sense that most of the time they far exceed what is needed to meet losses, and they are too small in the sense that they are inadequate to meet losses when a default crunch does occur.

MPI could be used right now to reduce the number of foreclosures, and reduce losses on the foreclosures that occur. The appropriate treatment varies with the situation, as follows:

*Loans in default that carry TMI.

*Loans in default that do not have TMI.

* Purchase loans & high-rate existing loans in good standing.

*Existing loans in good standing with negative equity

Loans in Default That Carry TMI: We estimate that there are about 600,000 loans now in default that carry TMI. Case-by-case efforts to modify these loans will make only a tiny dent in the total. There are major impediments to such modifications (see Foreclosures That Shouldn’t Happen But Do), and even when these impediments are removed, the process is costly and time-consuming. MPI makes possible wholesale modifications covering large numbers of loans that are in the interest of all parties - insurers, investors, borrowers and servicers.

If loans in default carry TMI, the insurer is already on the hook for the coming loss, up to the cap on the policy. If the net proceeds from sale of the property do not cover the unpaid balance including accrued interest plus foreclosure expenses, the investor is protected up to the cap amount for any deficiency.

In more normal conditions, deficiencies in coverage seldom arise. In the current market, however, house price depreciation has made deficiencies the rule rather than the exception. Under these circumstances, the deal using MPI shown below will make both the insurer and the investor better off. In the process, a significant number of borrowers will be saved from foreclosure.

The major benefit arises when the payment reduction resulting from the rate reduction allows the borrower to return to good standing. Everyone – borrower, investor, insurer and servicer - are better off. The downside to the investor is that some of these loans may cure themselves without the investor having to reduce the interest rate. After some extensive modeling, we have determined that "losses" to the investor from this source would be much smaller than the gains from loans that return to good standing that would not have cured themselves.

Loan servicers acting for themselves, as opposed to their actions as agent of the investor, have a self-interest in making the package deal described above because the resumption of payments under MPI also means a resumption of servicing fees. Most loans are serviced under contract with investors, with servicing fees paid as a withdrawal from monthly payments. When borrowers stop paying, servicing fees stop. As payments on loans in default resume under MPI, servicing fees resume.

Loans in Default That Do Not Have TMI: When loans in default do not carry TMI, the investors are on the hook for the entire loss. The insurer will not assume the risk except as a new loan that meets its underwriting requirements, and carrying an MPI insurance premium scaled to the risk. These deals must be done one at a time.

To meet its requirements as an insurable loan, the insurer will require the investor to write down the loan balance to 90% or 95% of current property value, and reduce the interest rate to the prime rate. If there is a second mortgage, the investor will have to negotiate a payoff with the second mortgage lender.

The investor would take a loss on the modification, but it would be far smaller than the loss that would have resulted from foreclosure of the original loan. The investor might also receive an equity certificate equal to the write-down of the balance (or balances, if there is a second lien), which would entitle it to a share of the future appreciation in the value of the house.

The borrower gets a new start with 5% or 10% equity, no unpaid interest, and a reduced payment based on the lower rate. Because the payment reduction will be partly offset by a new mortgage insurance premium, this plan will work best with higher-rate mortgages.

The insurer gets an MPI policy which it has underwritten as a piece of profitable business.

This application of MPI to loans that do not now have TMI is very similar to the proposed "FHA Housing Stabilization and Homeownership Retention Act of 2008" (HR5830), which would authorize FHA "to insure refinance loans for substantial numbers of borrowers at risk of foreclosure…" The essentials – writing down the loan balance, providing the existing investor a claim on future appreciation, and re-underwriting the loan – are the same. But there are some major differences:

*HR5830 stipulates some detailed conditions including underwriting requirements designed to protect FHA against excessive losses, such as maximum debt to income ratios. The Act is full of such rigidities, which may result in no or limited response by the private sector. MPI underwriting rules and conditions, in contrast, will be formulated by the private insurers in consultation with the investors who will be acquiring the new loans.

*HR 5830 would create a new Federal administrative infrastructure (an "Oversight Board") with heavy responsibilities related to determination of fees, underwriting rules, insurance premiums, and the like. The Board would be drawn from the Treasury, Federal Reserve Board, and HUD. There is no time limit in HR 5830 for establishing the Board. The MPI plan, in contrast, would use the existing private infrastructure.

*Under HR5830, the Federal Government assumes direct insurance liability for every mortgage written. With the MPI plan, private insurers would assume this liability, with the Federal Government’s role limited to that of re-insurer. This is discussed below.

Purchase Loans and High-Rate Existing Loans in Good Standing: Insurers will offer MPI on new purchase loans as a way to generate income and bolster reserves. For non-prime deals, the terms will be far better than those available in the current market. More precisely, the insurance premium that non-prime borrowers will have to pay will be much smaller than the sum of TMI premiums and interest rate risk premiums they now pay.

Insurers will also offer MPI to existing borrowers who currently pay high rates and have enough equity to meet underwriting requirements.

The major argument of this paper is that MPI, in addition to making a less-vulnerable and more borrower-friendly housing finance system, could help fix the current mortgage mess. MPI would reduce 1) the number of foreclosures, 2) investor losses on loans that do foreclose, and 3) financing costs to non-prime borrowers.

For existing loans with TMI that are in default, converting a TMI policy into an MPI policy should be no problem. Investors, insurers and borrowers will be better off than they are now. But on new loans, there is a potential problem connected to the acceptability of private mortgage insurance.

The core benefit of MPI, the elimination of interest rate risk premiums at little or no cost to the insurer, will materialize only if investors have complete confidence in the insurers. That would not have been a problem 2 years ago, when all the companies were rated AA or AAA. It is a problem today because the insurers have been weakened by their large payouts on foreclosed loans, and they have all been down-graded by the rating agencies.

To make MPI work with maximum effectiveness, insurers need an option to purchase reinsurance from GNMA, a wholly owned Federal agency, which would commit to make the mortgage payments if the private insurer cannot. This is the same type of guarantee that GNMA now provides on securities issued against pools of FHA and VA mortgages.

GNMA has charged 6 basis points for its guarantee, and it has been a consistent source of profit for the Government from the beginning. We would expect a similar guarantee fee on privately insured loans, and a similar experience. GNMA’s reinsurance would come into play only if the private insurer went broke.

There is no reason why a benefit directed to one industry should be permanent, and once the market has been normalized and the insurers have rebuilt their reserves, there should be no further need for it. A good way to minimize the support period is to build automatic premium increases into the program. This would provide a strong incentive for the insurers to rebuild their capital as quickly as possible.

With MPI backstopped by GNMA, investors including the Federal agencies Fannie Mae and Freddie Mac, and investment banks that purchase mortgages not eligible for purchase by the agencies, will pay a price based on the prime rate. Since borrowers will know what the prime rates are, the lenders who sell to the secondary market, and the brokers who feed loans to lenders, will compete on their markups over the prime rate, and on service.

Why should the Federal Government provide special support to the mortgage insurance industry? The major reason is that it will turn the market around at a critical juncture, while initiating a reform process that will make the entire housing finance system less vulnerable and more equitable in the future.

The proposed Government support is not an industry bailout. The probability that the contingent liability assumed by GNMA will ever cost the Government anything is very small.

Indeed, the proposed role for Government could more accurately be viewed as a bailout prevention measure. It will eliminate the need for the more extensive Government support that will become unavoidable if the situation deteriorates further.

To encourage private mortgage insurers to offer and investors to accept MPI on new loans, we recommend that Congress pass Federal legislation enabling the Government National Mortgage Association (GNMA) to re-insure private mortgage insurers that issue MPI. The GNMA guarantee should cover all payments (including the final payment) due the investor pursuant to the MPI policy in the event that the private mortgage insurer fails to pay.

The GNMA guarantee should be made available only to new MPI policies issued by any private mortgage insurer on owner-occupied principal residences.

If an insurer at some point elects to forego a GNMA guaranty on new loans carrying MPI, outstanding loans should retain the guaranty and GNMA should continue to receive the guaranty fee on those loans.

Mortgage insurers eligible for the GNMA guaranty should be licensed to do business in all 50 states, and should be eligible to insure loans acceptable to both Fannie Mae and Freddie Mac. Eligible insurers also should have programs in place acceptable to the Federal Reserve Board for a) the wholesale conversion to MPI of loans in default on which they currently have insurance, and b) new loan programs using MPI covering purchase loans and refinances, including refinances of loans currently in default that do not now have mortgage insurance and that require contract modifications by investors to become eligible for MPI.

Copyright Jack Guttentag and Igor Roitburg 2008