Don't Model the Answer You Want to See
The DCG advisory consulting team starts every week with an internal discussion of market trends, regulatory developments, and the real experiences of our bank and credit union clients. Here are the notes from this week’s Monday Morning Meeting.
Check out the meeting notes from previous weeks.
Do you have a question?
This week, the team dug into a topic that has surfaced in several recent conversations with DCG clients: how a risk model’s greatest value can sometimes lie in making teams uncomfortable.
Risk models are designed to identify what could happen, not to validate what you hope will happen.
Financial institutions learned a tremendous amount about customer behavior during the most recent rising rate cycle. Deposits migrated. Betas are non-linear and changed at different points in the cycle. Loan pricing didn't move in lockstep with market rates. And customer behavior proved far more dynamic than many traditional assumptions suggested.
Yet incorporating those lessons into an ALCO model can be met with resistance, since it will make the numbers worse. That is the point! Understand the potential risk before potential becomes reality.
A more dynamic scenario could produce greater earnings exposure. That result could invite different questions from different audiences (ALCO, board, regulators, the street).
But choosing not to examine the risk doesn't eliminate it.
In the last cycle, institutions had stress-test results available that identified vulnerabilities, but in too many cases those scenarios were dismissed as low probability. When actual results later diverged from expectations, the natural question became, "Why were we so wrong?"
Going forward, the better approach may be to start with what customers have actually done.
How have depositors behaved at different points in the rate cycle? How has money migrated between products? How have loan spreads responded to market movements? And how might those behaviors change in the next environment?
Rather than beginning with a hypothetical scenario and debating whether an assumption/scenario is "right," ALCO can begin with institution-specific behavior and ask what that behavior tells us about potential outcomes in different scenarios.
A good risk model shouldn't always make management comfortable. Sometimes its greatest value is showing something management doesn't want to see, early enough to do something about it. The goal isn't to produce the best-looking ALCO results; it's to produce the clearest picture of the risks the institution actually faces.
What are others seeing?
For more insights from Darling Consulting Group, click here.
© 2026 Darling Consulting Group, Inc.
