Magazine article Mortgage Banking

The State of Loss-Severity Modeling

Magazine article Mortgage Banking

The State of Loss-Severity Modeling

Article excerpt

Whether you're a non-agency trader, a whole-loan portfolio holder or a government-sponsored enterprise (GSE), all have a common risk: Defaulted loans result in losses. Servicers rely upon projected losses to adjust collections or loss-mitigation strategies. Brokers consider potential loss, at least indirectly, in pricing loans. Losses play a key role in every aspect of the mortgage world.

This month's column focuses on loss severity--that fraction of loan balance that is lost on a defaulted loan.

Typical valuation tools, such as automated valuation models (AVMs), which focus on today's values, are not sufficient for loss-severity modeling. The most accurate loss-severity models estimate losses based on inputs available in the past, present and probable future of the loan.

Yesterday and today

Before the housing market collapsed, most mortgage models used a widely available metropolitan statistical area--level (MSA-level) home-price index (HPI). Widespread home-price appreciation made this choice seem reasonable. Why bother with a more-granular HPI to distinguish loans with current loan-to-value (LTV) ratios of 55 percent and bo percent? Both have substantial equity.

The difficulty with loss-severity modeling in the past was that negative-equity loans were rare, according to the HPI. Most borrowers with equity who lost the ability to pay did not enter foreclosure; they simply sold their property at profit. Those borrowers that did enter foreclosure were somehow unable to sell at a profit.

For whatever reason (lack of upkeep, location in an MSA subregion with negative appreciation, etc.), they did not have true equity, and the HPI failed to capture this. This presented a modeling challenge. LPS Applied Analytics' modelers proposed that a loss-severity model should assume a wide probability distribution of current LTVs and consider only the high-LTV range where borrowers can't sell at a profit.

Interestingly, loss-severity modeling has become easier. These days, defaults are not rare events caused by dilapidated properties in the tail region of some LTV distribution. Negative equity is common, and loans that go bad do so because they are truly underwater.

Whether a borrower loses the ability to make payments or makes the choice to default, the end result is the same. Someone takes a loss.

Home-price indexing

A home's property value is the most crucial ingredient in predicting loss severity. The days when coarse MSA-level HPIs had any value as loss-model drivers are gone. Increased granularity is needed to inform home-price predictions. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.