Automated underwriting's next frontier: servicing: it's only logical—so why not move automated underwriting tools into servicing, where they can help determine best options for loss mitigation?

Mortgage Banking,  Sept, 2008  by Linda C. Simmons

• E-mail
• Print
• Link

The history of automated decisioning in the mortgage industry is, for the most part, a snapshot of automated underwriting (AU) and, more specifically, a framed photo of the tools provided by the government-sponsored enterprises (GSEs)--Fannie Mae's Desktop Underwriter [R] (DU) and Freddie Mac's Loan Prospector [R] (LP). The adoption of these automated underwriting tools was so swift and so thorough that other investors (e.g., Countrywide Financial Corporation and GMAC-RFC) made the investment in their own tools in no small measure to harvest the automated decisioning momentum. / As inherent as DU and LP have become in the mortgage loan value chain, it's ironic that the automated determination of both loan risk and value remained isolated in the upfront part of the value chain. / While some companies pushed decisioning out in front of underwriting to facilitate high-level product and pricing eligibility for the borrower, the decision was most often positioned as a recommendation rather than an investor-ready determination or a firm commitment. / Along the traditional mortgage value chain, we tend to associate automated decisioning with determining product, pricing, eligibility, credit and collateral underwriting. / To some extent, these decisioning limitations may be a reflection of the somewhat linear, process-oriented approach associated with our traditional "what does it take to get to yes?" lending. Certain steps are taken and boxes checked before another sequence of steps is undertaken. Only with true workflow could certain steps be skipped as not necessary or eliminated as not relevant to a particular lending process.

[ILLUSTRATION OMITTED]

[ILLUSTRATION OMITTED]

The high velocity lending of 2001-2007 produced an interesting nuance to this lending process: If certain steps are skipped and/or eliminated (e.g., no docs, no income verification)--without the benefit of workflow--is my answer still going to be "yes"? While we are living with the outcome of this particular approach, can I improve upon my "yes" answer if I can incorporate the right price and the right product for this borrower's situation?

The next frontier?

Given that hindsight is 20-20, what decisions would be made today if we knew then what we know now? Rather than a purely rhetorical question, we are reliving many of those 2001-2007 decisions today in loan servicing, generally, and in the default-management arena in particular.

While servicing software may drive the process, in many cases humans are making the decisions. We don't want to replicate the nonconforming underwriting process in loan servicing. But with up-to-date data, would the application of automated decisioning capabilities allow us to make better decisions today in loan servicing?

Though hardly a complex concept, this notion of leveraging automated decisioning beyond underwriting is best captured in the term "redecisioning." The concept can be defined as: updating decisions that were made at the point of origination by applying a set of decision solutions with updated data to reevaluate risk and the value of a loan at any point along the value chain.

Redecisioning can be performed as often as necessary to manage risk across the value chain with appropriate data and decisioning rules.

In servicing, any place decisioning occurs--adjustable-rate mortgage (ARM) resets, refinances, cross-sell, escrow analysis, loan modifications, loss mitigation, payment processing--could be improved upon with automated decisioning (see Figure 1).

[FIGURE 1 OMITTED]

Redecisioning takes the management of risk, value, pricing, economic value and even litigation support from coarse to precise. This allows management to focus its efforts on those loans or portions of the portfolio where economic, legal and/or other risk management is needed.

The resulting degree of precision facilitates capital deployment where it is most needed, and begins building the future state for both transparency and consistency across the enterprise.

Prospect to be well-ARMed

The current environment suggests we may need to reconsider how we think about critical determinants of risk and value and--as important--when we think about them. For example, an ARM reset on a subprime loan may have more in common with a product pricing and eligibility and underwriting decision than first considered. An upcoming ARM reset may occur with no ramifications, may trigger a borrower to refinance or seek a loan modification, and/or may push a borrower into default.

By applying product pricing and eligibility rules along with automated decisioning when reviewing pending ARM resets, the loan servicer may ascertain the best outcome for the borrower and retain the loan. This can be accomplished well in advance of a reset. Even if the reset might trigger a default, the loan servicer is well aware of the issue and can begin working with the borrower sooner.

While the emphasis around ARM resets is usually represented as a subprime challenge--and we still have a ways to go through 2008 on that score--market watchers remind us the peak issuance of 5-1 ARMs in the conforming market occurred in late 2004 and 2005. That fact signals another ARM reset "bulge" in 2009 and 2010 (see Figure 2).