Q/A with Chuck Rumfola of Veros

MBA NewsLink recently posed questions to Chuck Rumfola, senior vice president of strategic initiatives with Veros Real Estate Solutions, Santa Ana, Calif.  

Rumfola is responsible for building strategic relationships within the mortgage industry to create and implement programs that reduce lender and borrower risk. Prior to Veros, he worked for Fannie Mae for nearly 20 years, most recently as vice president of single-family strategic initiatives; he also held executive positions in capital markets and mortgage operations.  

MBA NEWSLINK: Veros was chosen by Freddie Mac and Fannie Mae in 2010 to build, support and maintain their Uniform Collateral Data Portal. Five years later, how would you assess your efforts in creating a more transparent process?   

CHUCK RUMFOLA: The GSEs had a very specific desired outcome when embarking on the UAD and UCDP initiatives–they wanted to drive better appraisal quality. By doing that, the GSEs could address the severity of credit losses, one of their biggest problems that led to conservatorship.  

Appraisal quality had deteriorated and inflated appraisal values were very common in the industry. The GSEs had already changed many of their credit policies to reduce the frequency of loans going into default (e.g., eliminated low documentation loans, raised the credit score minimums and increased down payments), but now needed to modify their appraisal policies, which involved infrastructure changes, to protect themselves going forward.    

When the work started on the Federal Housing Finance Agency’s Uniform Mortgage Data Program mandate, the GSEs agreed to a guiding principle that they were not going to compete on the data. Instead, the teams would standardize and collect the data, with the real competition coming from who could best use the data for superior modeling and analytics that could lead to better risk management and new products.  

Five years later, look at what has evolved:  

–UCDP has over 6,000 registered companies and 33,000 users. Additionally, over 28 million standardized appraisals have been submitted through UCDP. This has created an appraisal database that both entities use for modeling and analytics.

–Fannie Mae (Collateral Underwriter) and Freddie Mac (Loan Collateral Advisor) have introduced new tools to help their lenders better manage collateral risk.

–FHA, VA and USDA have all adopted the UAD standard.

–The GSEs have modified their Representation and Warranty business model by introducing a sunset period. And, recently, the GSEs have announced further policy changes clarifying different loan defects and remedies for those defects.  

–Both GSEs have relaxed down payment requirements and will now go up to 97 percent LTVs.

–FHA has introduced its own appraisal portal (Electronic Appraisal Delivery), which is mandated for all lenders in June.

–Overall, improved appraisal quality.  

Obviously, UAD and UCDP were not the only factors enabling what has evolved over the past five years. I would argue that having standardized data and a database for risk modeling and analytics were certainly major factors.  

Finally, UMDP has been so successful that the GSEs have broadened the initiative to include the Uniform Closing Data and Uniform Loan Application Dataset.    

NEWSLINK: Looking ahead, what are key issues you anticipate addressing for the GSEs and your clients?   

RUMFOLA: I see two major issues. One is the continued development of the Rep & Warrant framework. Lenders want more certainty from the GSEs when they sell or securitize their loans. There was a tremendous amount of friction between lenders and the GSEs during and after the mortgage meltdown. The GSEs accused the lenders of rep and warrant violations and issued billions of dollars of “repurchase requests.” Conversely, the lenders accused the GSEs of issuing “repurchase requests” on trivial loan defects that didn’t have an impact on the loan performance. These arguments took years to resolve.    

The second issue relates to the first. To manage the risk of repurchase requests, lenders added credit overlays and began to originate loans well within the GSEs’ credit box. This was a defensive reaction by lenders. There was less of a chance of a repurchase request from the GSEs if the loan performed. Given that, the lenders became much more conservative with their credit policies, essentially only lending to the most credit-worthy borrowers. The result was very tight credit and lower origination volumes.  

The challenge is to get lenders to lend to the outer limits of the GSEs’ credit box. You do that by providing more rule clarity and certainty for lenders. The work being done to clarify the rep and warrant framework, along with the tools each GSE is developing to help lenders understand potential issues with loans, are providing more certainty for lenders and should result in an increased level of comfort in eliminating credit overlays and originating within the entire GSE credit box.   

NEWSLINK: What are the biggest challenges in making predictive models more accurate?  

RUMFOLA: The challenges always come back to the data. If you had one complete, perfectly updated data set, creating predictive models would be easy, but that doesn’t exist. In its absence, modelers need to be able to test and validate the available data to put a finer point on its predictive usefulness for end users.  

While industry data are very good, it is not consistent in frequency of updates, delivery methods, geographic parameters, available variables or how and which variables are updated. The key to successful predictive modeling is properly cleansing the data to provide the best quality and consistency, and then working those models around all the anomalies and continuing to do so in a timely manner.  

Automated valuation models have become extremely accurate over the years, to the point we are at today in which the margin of error on AVM test files is extremely low. Veros and other fellow AVM providers will continue to strive to narrow that gap, and at the same time it will be important to maintain the proper controls around various new or evolving data sources.  

NEWSLINK: How would you rate lenders’ use of data, particularly on the front end of the transaction? How can they use their data more effectively?   

RUMFOLA: Lenders are getting savvier, no doubt. They are checking and using data earlier in the loan lifecycle to detect risk and ensure investor compliance. In response to GSE loan data initiatives, lenders are gathering and checking data using third-party tools such as income validation tools, AVMs, collateral valuation management and appraisal risk scoring tools, and systems that review data compliance prior to portal submissions.  

A quick example of lender creativity in leveraging data is a lender who has interpreted the Reg B disclosure requirement and its operational protocols in a way that makes ordering an AVM as a data point in appraisal QC too much of a burden. This is a lender who understands the value in comparing the appraiser’s estimate against an AVM to get an objective, cost-effective and nearly instantaneous line of comparison, but feels they can no longer do so under current guidance for first mortgages.  

Using an appraisal scoring tool that provides a “value risk score” can give a lender that data point on the anticipated value risk when comparing the appraised value with the AVM, without disclosing the AVM’s expected value. With an understanding of the underlying analytics and through a quick, automated pull of this data point, a lender can generate a quick and informed opinion on the risk related to the appraiser’s value conclusion, and still fall within compliance without the necessity to furnish the additional documentation. From here, a lender can take a variety of next steps to remain in line with guidance, but this is just one small example of how data can be used more efficiently in predictive applications.  

This scenario offers perspective on how lenders can use data more effectively by putting available tools to work early in the origination lifecycle and by developing creative business practices around the application of data. With the investments and expectations of the GSEs to guide them, there is a real opportunity for mortgage lenders to build exciting and profitable data-driven business.  

(Views expressed in this article do not necessarily reflect policy of the Mortgage Bankers Association, nor does it connote an endorsement of a specific company, product or service. MBA NewsLink welcomes your submissions; articles and/or Q/A inquiries should be sent to Mike Sorohan, editor, at msorohan@mortgagebankers.org.)