Banks and non-depository mortgage companies engage ADI to evaluate Fair Lending risk relative to consumer outcomes from their underwriting activities. We endeavor to align our solutions with the reasons consumers withdraw prior to an approval/denial decision based on the merits.  

In the past several years, the proportion of non-approved HMDA-reported applications that were denied based on review of an applicant’s credit circumstances and the characteristics of a collateralized property has declined. In 2009, denials comprised just over one-half (53 percent) of non-approved first-lien applications. By 2019, denials fell to represent less than one-third (31 percent) of non-approvals. Over the same timeframe, withdrawn applications increased from under one-half to over two-thirds of non-approvals.

A community bank client provided us a loan application-level data set and wanted an assessment of its underwriting Fair Lending risk. Aligning with the national trend, withdrawals represented the majority of non-approved applications in the data set.

Assessing the Challenge

When ADI receives client loan-level data, we can quickly assess whether there are regression-adjusted differences in approval rates for consumers who are members of Prohibited Basis (PB) groups, and Non-PB (NPB) group members. When we conduct underwriting Fair Lending analysis, we pay particular attention to two types of non-approved applications:

  • Denials by credit and collateral underwriters based on the merits discovered through their underwriting processes; and
  • Withdrawals by consumers prior to the underwriter’s approval / denial decision.

Focusing on withdrawals, we pose the question “What factors motivate consumers to withdraw before they learn whether their application is approved?”  The reason(s) for such decisions are often multi-faceted and not easily recorded by choosing items from a dropdown menu. This poses a challenge in designing  Fair Lending solutions for withdrawn PB group member applications.

Designing the Approach

The  client’s data file included information (HMDA field Reason for Denial) as to why consumers were denied loans; however, for withdrawals, no such reasons were documented.  Without direct evidence in the client’s data file for why any given consumer withdrew, our assumption was that the withdrawals were motivated by either pricing or non-pricing factors:

  • Pricing: a pricing-motivated withdrawal is one in which a consumer chooses to originate a given loan with a different lender because it has offered more favorable pricing;
  • Non-Pricing: these withdrawals may reflect a consumer deciding to a) sell her home rather than refinance; b) cancel a purchase due to a disappointing inspection; or c) discontinue interactions with an unhelpful loan officer.

Our approach was to first assess whether or not pricing was a key reason behind applicant withdrawal decisions. We requested that the client include data about pricing offers made to each consumer who withdrew and parallel information for those who originated a loan in the same period. Our goal was to diagnose whether those who withdrew had systematically higher pricing offers than those originating a loan.

Implementing the Solution

Our Bank client originated roughly 1,500 Conventional refinance loans in 2019. Meanwhile, 1,100 applicants withdrew their applications prior to the Bank underwriters reaching an approval or denial decision. Our diagnostic testing revealed that the 1,500 originated loans had APRs that ranged from 2.79 to 7.67 percent with a median of 3.91 percent. Using a range of available data elements such as loan term, market interest conditions, loan-to-value ratios, credit score, property type and others, our regression modeling accurately predicted 70 percent of the variation in consumer APRs. We used a nearly identical approach to predict pricing offers to the 1,100 applicants who withdrew and found similarly predictive relationships.

Next, we combined the two sets of loans and employed suitable regression techniques to assess the diagnostic question outlined above: Did applicants who withdrew have higher pricing than similarly situated applicants who originated a loan in a given period? The regression-based evidence was clear: Consumers who withdrew had pricing offers 15 basis points higher than the 1,500 consumers who ultimately refinanced their existing mortgages.

With this evidence in hand, we asked the next – and most relevant – Fair Lending question: Did PB group members have higher pricing offers than NPB group members seeking similar products?

Our results:

Across eight PB groups – defined by race, ethnicity, age, gender and marital status – there were no differences in regression-adjusted pricing of the combined set of originated loans and withdrawn applications for Conventional refinances, using the industry-standard 95 percent confidence level. We concluded there was no Fair Lending risk in this set of applications.

In the same year, the Bank also originated 1,300 VA-insured refinance loans, while 700 VA refinance applicants withdrew their applications. We conducted similar testing for this set of 2,000 records. Notably, we found that regression-adjusted pricing offers were five basis points higher for those who withdrew, than those who completed the refinance. In brief, we concluded that pricing was a key factor in the decisions to withdraw.

  • We also discovered that consumers 62 or older had significantly higher pricing than those under 62, other factors held constant. While some of the high-risk PB group members applications resulted in an approval, others resulted in a withdrawal.
  • We identified two sets of high risk records: a) PB group member loans; and b) withdrawn applications using steps linked to our regression modeling. In each set, regression-adjusted pricing was significantly higher than that for similarly situated NPB group members.  We recommended the Bank conduct Comparative File Reviews (CFRs) for each set of records. We selected comparator loans in which similarly situated NPB group members had lower pricing. The Bank implemented CFRs in an effort to identify reason(s) for pricing differences in each pair to exonerate underlying risk.

Key Points

  1. There are many pricing and non-pricing reasons why consumers withdraw a mortgage application prior to an approval or denial decision from the lender.  From both a Fair Lending and business development perspective, there may be strong benefits for lenders to further analyze and respond to these outcomes.  
  2. Increasingly at ADI, we find a strong empirical association between loan pricing offers and withdrawal decisions. Consumers who withdraw have relatively less attractive pricing offers, on average. Many non-pricing factors may also prompt a consumer’s decision to withdraw.
  3. When diagnostic testing points to pricing as a likely reason for withdrawal decision, we design and implement a pricing-based testing approach. Where we find that PB group members who withdrew did so for pricing-related reasons, we recommend CFRs as a next step. 

About the Author

Paul Strasberg, Ph.D.

Paul is ADI’s Senior Economist and lead consultant in ADI’s Fair Lending analyses to accurately measure potential Fair Lending risk. You can contact Paul at or 703.740.4907.