by Tanta on 3/21/2007 02:59:00 PM

PMI is insurance against mortgage default; it insures the lender, not the borrower. I use the abbreviation “MI” instead of “PMI” because there is a company called PMI that offers insurance with which I do not want to become confused; you may, however, encounter other sources using “PMI” in the generic sense. Fair warning. The “private” part of the term means we are not talking about government-insured loans such as FHA. Such government-sponsored insurance works similarly, but not identically, to private MI.

We often use the term “first loss” position when talking about MI, and I am going to use that term, because that’s how the industry does it. But we should always bear in mind that the true first loss position on any mortgage loan is the borrower’s, whenever there is borrower equity. Here’s how it works.

In exchange for a premium, the mortgage insurer reduces the lender’s exposure to risk. The amount of the premium is determined by the coverage level, loan type, and credit quality of the loan. There are lots of different coverage levels available; most lenders use the ones required by Fannie Mae and Freddie Mac. For instance, the standard coverage level on a 30-year loan with a 95% LTV is 30%. That means that the insurer will cover losses in the event of default up to 30% of the insured loan balance. Another way of putting that is that the lender’s original exposure is reduced to 67% LTV (95% times 70%). Standard coverage for 30-year loans in the 90.01-95% range is 30%; in the 85.01-90% range is 25%; in the 85% and under range is 12%. Less coverage (25%, 12%, and 6%, respectively) is required for loans with a 20-year or less term. Coverage for very high LTV loans (greater than 95%) is not really standard; the GSEs will require a specific coverage level for different loans they might buy in that category. You can generally assume a minimum of 35%. Here is a link to MGIC’s actual rate cards, which show how premiums are priced for different loan types, terms, and features. I expect the terminology will be gobbledygook for a lot of you, but we can get that far into the weeds later if you insist.

Specifically, in the event of default the mortgage insurer pays the lower of 1) the coverage percent (say, 30%) of the unpaid principal balance of the loan plus accrued interest and liquidation expenses or 2) the actual net loss. The MI policy generally gives the insurer the right, but not the obligation, to pay 100% of the unpaid principal balance (and any accrued interest) and take title to the property itself, if it wishes to handle the actual liquidation. It might want to do that if it believes it can recover more than the mortgage servicer can, which might well be the case with a small servicer without expertise in handling REO. What you should bear in mind is that the longer it takes to foreclose and market the REO, the more interest accrues, and the more expenses rack up, resulting in higher net losses and thus higher claims paid by the insurer. The insurer has a vested interest in making sure that the process is both timely and efficient. It controls its exposure by reserving rights to approve and regulate parts of the collections/foreclosure/ REO management processes, by having the option to bid at the foreclosure sale, and by challenging any claims that it considers excessive.

CR UPDATE: Try this link for spreadsheet.

This spreadsheet shows a simple comparison of a lender’s potential losses in the case of a 95% LTV loan with MI, and an 80% LTV loan without MI. The loan term is 30 years in all cases. Three base scenarios are presented: a fully-amortizing loan, an interest-only loan, and a negatively-amortizing loan. For each scenario, it is assumed that the loan defaults 18 months after origination; the columns show the result with no change in the property value, 10% depreciation, and 20% depreciation. The net loss is calculated assuming that it takes 6 months from default for the lender to foreclose, and 6 months from foreclosure for the REO to be sold. Total expenses (foreclosure costs, repair and maintenance of the property, force-placed hazard insurance, RE broker commission, closing costs on the final sale, etc.) are simply assumed to be 15% of the sales price in any scenario. If you’ve been wondering why the coverage level is lower on loans with a shorter term, notice how amortization of the loan balance affects the loss severity. The shorter the loan term, the faster the borrower pays down principal, and so those loans require less MI coverage. You can also see here how interest-only and neg am can get ugly, as well as how the age of a defaulted loan affects severity.

Of course, the actual net loss for any loan is dependent not just on the sales price of the REO, but also on the efficiency of the servicer (how quickly and inexpensively the servicer handles the FC and REO), the property state (which affects the FC time frame, whether FC is judicial or nonjudicial, etc.), the costs of hazard insurance and any property taxes that come due while the lender holds the REO, and so on. You may adjust the spreadsheet with your own assumptions; for many markets and properties, 15% may be a very sunny assumption.

A mortgage insurer is considered in “first loss” position, and so the spreadsheet shows loss to the lender (second-loss position) only after payment of the maximum MI claim amount. The MI’s exposure to loss is limited: it covers only the contractual percent of the loan balance plus accrued interest and expenses. The lender’s potential exposure has no contractual limit: it is solely a matter of how far the property value can depreciate and how expensive it can get to foreclose or liquidate REO. In a real doomsday scenario, the lender’s loss could be 100% net of the MI claim proceeds. You can build your own “doomsday” scenarios on the spreadsheet just by slashing the REO liquidation value to 50% or less of the original sales price, or increasing the default-to-liquidation period of interest accrual, or increasing expenses, or all of it at once. If you wish to play subprime lender, just up the interest rate on the loan (the spreadsheet uses 6.50%).

You will notice, of course, that the 80% LTV loan almost always results in a loss to the lender in foreclosure. Remember, though, that this spreadsheet shows loss severity. It does not address the issue of loss frequency. All other things being equal, the borrower with substantial equity will default less frequently than the borrower with minimal or no equity—the borrower, remember, is in true “first loss” position. The lower default frequency of the 80% LTV loan compensates for its higher loss severity, compared to a high-LTV insured loan. If, of course, the 80% LTV loan is really an 80/15, say, then the default frequency of the 95% MI loan is likely to be comparable to the 95% CLTV loan (which is why sane lenders—both of them—charge a higher interest rate on the first mortgage when the CLTV is that high). In the latter case, the second lien holder is in first loss position. Notice that the spreadsheet shows only one scenario in which there are any proceeds available to a theoretical second lien holder. Obviously, in most cases of foreclosure a second lien is a total charge-off. Therefore, the profitability of second-lien lending is only as good as the underwriting and pricing of the second liens, just as the profitability of insuring first mortgages is only as good as the underwriting and pricing of the premiums. (Note that a second lien holder has accrued interest and expenses, too, at least in the early months of default when it is attempting to collect on the loan. As a rule, second lien holders give up collection efforts and write off loans very quickly, compared to first lien holders, because expenses are rarely recovered.)

Our next installment of Private Mortgage Insurance for UberNerds will discuss the various ways premiums are calculated and remitted, and will explore the differences between primary, bulk, and pool insurance, which bond investors particularly will care about. Go ahead and ask all your questions in the comments anyway; you might as well let me know what you want to know in future installments (a brief explanation of why you care is optional, depending on how much of your inner Nerd you wish to reveal in the intimate setting of the internet). In the meantime, play with the spreadsheet and try not to let the suspense get too much for your blessed little hearts to bear.