In the third Data Analytics Series BIGcast, The CECL Effect, John Best speaks with Joe Breeden of Deep Future Analytics about CECL, data pooling, and predictive analytics.
The Impact of CECL
As Joe Breeden explains in the podcast, CECL stands for Current Expected Credit Loss, and is the new accounting standard for how financial institutions will set loss reserves. Typically, organizations under $10 Billion assets have utilized moving averages to calculate loss reserves, but this model is backward-looking and will not be acceptable for the new regulations. A moving average model will always set your loss reserves too low moving into a recession and too high moving out of the recession.
When discussing how to meet CECL requirements and create a forward-looking model, Breeden states, “There is a lot of flexibility on how you implement it, but there are two things that are pretty much unavoidable”:
- “Lifetime” – Financial institutions need to project what the total loss expectation is for the full life of a loan. If a new mortgage has a life of 30 years, loss needs to be predicted for the full 30 years, rather than just the 12 months or so traditionally done by the moving average method.
- “Current” – Financial institutions now are required to show current economic conditions. Somehow, credit unions now have to show current economic and near future economic conditions for likely about the next two years.
Data Pooling for the Good of your Organization and the Credit Union Movement
Advanced analytics requires massive amounts of data. The more data available, the more certainty in the results. A major advantage of data pooling is that credit unions can greatly increase the amount of data available to perform modeling by participating in a pool. Credit unions do not need to be geographically near each other to benefit from the pools, and as Joe Breeden states in the podcast, “If we get folks spread around the country, in a shared blind repository, then it gives us a better overall view of the scaling of the risk versus economics and other things.” He continues to explain that “We leverage that pool to learn aspects that are in common, like economic sensitivities, but then also to calibrate to the individual… so you get the benefit of the whole, but specific to the individual institution.”
Predictive Analytics: One Step at a Time
Ultimately, credit unions need to perform predictive analytics in order to meet the CECL requirements and ensure a successful future of member service. However, prior to being able to predict future loan losses and other valuable insights, financial institutions must understand the difference and master the first two steps of analytics; descriptive and prescriptive. First, let’s clearly differentiate the different types:
- Descriptive Analytics: As indicated by the name, descriptive analytics is simply used to explain what exactly your data is telling you. This could be how many members the credit union has or how many members walked into a branch on Friday. Not only should these answers be easy and quick to answer, but they should tie out across all departments. Until a financial institution has a single source of truth to access these answers enterprise-wide, prescriptive and predictive analytics will not be feasible.
- Prescriptive Analytics: At this stage, you can begin to look at your data more analytically. Instead of asking how many members visited the branch on Friday, prescriptive analytics can help answer why that number exists. From there, you can dig to the root and assess how to incent or discourage certain behaviors in the future.
- Predictive Analytics: Finally, once the first two steps are mastered, predictive analytics becomes a reality. Now you can create scenarios and use data to make actionable decisions that will affect the future of the credit union and its members. Rather than basing your prices on the market, you can take advantage of forward facing models and set prices properly for your members and set reserves appropriately for your credit union. These models are based on tangible data that takes all kinds of factors into account and prepares you for the economy you will be in, not the economy you were just in.
Unfortunately, the reality is, the majority of credit unions are stuck just trying to get past the descriptive analytics stage. Credit unions suffer from siloed data and a lack of commitment to analytics from top management. As a result, simple questions like the number of credit union members differs between departments. Credit unions need integrated and normalized data from a single source of truth in order to advance capabilities into more specialized and actionable analytics.
Visit http://bigfintechmedia.com/Podcast/the-cecl-effect to listen to the entire podcast.