Predictive Analytics In Play At BCU

The suburban Chicago shop is building out models to prepare for a surge in delinquencies and charge-offs.

 
 

Top-Level Takeaways

  • Predictive modeling combines internal and external factors to suggest the likelihood of delinquencies and defaults.
  • BCU will use such modeling to prepare for what might be a rough stretch for members, and the credit union’s books, in 2021.

CU QUICK FACTS

BCU
Data as of 03.31.20

HQ: Vernon Hills, IL
ASSETS: $4.0B
MEMBERS: 270,637
BRANCHES: 54
12-MO SHARE GROWTH: 8.8%
12-MO LOAN GROWTH: 7.3%
ROA: 0.56%

BCU ($4.0B, Vernon Hills, IL) is honing its skills in predictive modeling as the SEG-based cooperative tracks and acts on financial issues COVID-19 is posing among its far-flung membership.

John Sahagian, BCU’s chief data officer, says the credit union currently is focused on getting ahead of an anticipated surge of delinquencies and defaults that would follow the expiration of deferrals, pay skips, and mortgage forbearances as well as the end of increased unemployment benefits paid out as part of the CARES Act.

“We’re developing a model to show the likelihood of a loan to enter default,” Sahagian says. “We’re also working to strengthen predictive models we already have for early-stage delinquencies rolling to advanced collections and for recovery likelihood of charged-off loans. These efforts are still a work in progress, but we’re excited about their potential in allowing us to focus our finite resources where they can have the greatest impact.” 

Insights In An Imperfect Setting

John Sahagian, Chief Data Officer, BCU

Reaching members digitally and through its 54-branche network that extends through 18 states and Puerto Rico, BCU has accommodated nearly 90% of member requests for extensions and pay skips. It has fulfilled more than 15,000 such requests, seven times as many as last year at this point.

But working with members in many different geographic areas across the entire country  can be complicated. 

“Unemployment programs are different across the states,” Sahagian says. “Tracking these deposits is difficult, but we’ve still gained some insights. One interesting find is that there is little overlap between members requesting loan extensions and members receiving unemployment benefits.”

That’s the kind of insight that will help BCU predict who’s at risk and proactively respond when possible to every member’s situation.

Regression Analysis In Action

Unemployment figures are some of the external factors the credit union is including in its current projections for loan delinquencies and losses as it builds on its predictive modeling work.

Brett Engel, BCU’s director of finance, credit risk, and profitability, says the credit union built its loss models using regression analysis of the cooperative’s historical performance versus industry and macroeconomic factors.

Join The Credit Union Analytics Conversation

With monthly meetings discussing topics ranging from technology to how to build your team, there’s no limit to what you could learn when you connect with others who want to improve how they use analytics to better serve credit union members.

Join The Analytics Community

But the pandemic has complicated that approach, rendering historical data and models less reliable because of the influence of government aid, the high numbers of loan extensions, the rapid dips and spikes in employment, and the effects on different segments of borrowers.

For example, Engel says, the 2008-2009 financial crisis disproportionately affected prime borrowers whereas today’s crisis is harder on blue collar and service industry employment. 

“We will likely see some payment priority shift,” Engel says. “Borrowers might prioritize mortgages and credit cards over autos rather than the strategic mortgage defaults of 2008-2009.” 

Predictive Analytics And Loan Losses

Engel says the BCU data crunchers are running four Moody’s scenarios — upside, baseline, downside, and severe downside — for their macroeconomic model inputs each month to forecast losses. They also use loan delinquency models that incorporate the impact of fiscal stimulus on disposable personal income.  

Brett Engel, Director of Finance, Credit Risk, and Profitability, BCU

“The purpose of running these scenarios is to help management understand the potential range of outcomes given the vast uncertainty surrounding the crisis,” Engel says.

He says BCU uses Moody’s in the modeling because it’s among the more pessimistic of the 30-plus analysts the credit union follows, especially for unemployment, which Engel says remains a key driver in the credit union’s loss risk calculations.

“We adjust model inputs based on the median unemployment forecast versus the Moody’s forecast,” Engel says. “Over time, we expect analyst forecasts to converge, so our model results will update accordingly.” 

Also figured in is what the credit union calls the “BCU Portfolio Insulation Factor.”  

“We feel BCU is well-positioned in the crisis because we have more exposure to the health care industry, which has historically, and is projected to, experienced lower employment volatility versus the broader market,” Engel says. “So, we make another qualitative adjustment to our loss forecast based on the lower employment volatility based on the weighted average industry exposure of our loan portfolio.” 

Expectations For 2021

Engel says BCU is currently forecasting most of its pandemic-related delinquency and charge-offs to occur in 2021. The credit union is already working to mitigate that. 

“We’ve been proactive in providing hardship accommodations to our members,” Engel says. “Many have already taken advantage of one round of extensions, and we are planning another round for members who still need help.”  

The credit union also is expecting Congress to eventually approve, and the president to sign off on, another $1 trillion aid package. 

“Our baseline loss scenario includes this additional round of stimulus,” Engel says.

Want more credit union strategies? Sign up for the CreditUnions.com free newsletter.