How to Judge Automated Valuation Models

Article excerpt

In the wake of the refi boom, nonprime and alternative-A lending have become popular strategies to sustain and increase volume. That means more risk of fraud or inadequately collateralized loans next year.

Most top lenders have adopted automated valuation models. But when foreclosures occur, valuation is only a reference point; stakeholders in a loan or loan pool mostly care about what their loss is likely to be.

Lenders should understand whether and how collateral risk assessment is built into the automated valuation models they use and should install suitable tools to spot collateral risk and identity fraud. Filtering out risky loans before underwriting can also save a bundle in administrative costs.

The best providers of AVMs (including my company) keep their recipes secret. But market velocity and volatility, foreclosure activity, and historical trends are all vital elements in assessing collateral.

Velocity is a capricious variable. For example, it can skew automated valuation "hits" simply because automated underwriting has accelerated origination volume and caused a temporary surge.

Radically fluctuating housing prices in a given market can also lead to wide swings in the reliability of AVM numbers. …