Calibrating “Likely” in Unsafe-and-Unsound Risk Standards

This article was developed using publicly available responses submitted to Requests for Information issued by banking regulators. It summarizes and synthesizes themes, perspectives, and information reflected in those public submissions for informational purposes only. The article does not represent the views of any regulator, respondent, institution, or the Firm, and should not be interpreted as legal, regulatory, or compliance advice.

Executive Summary

Question 5 asks whether “likely” is the right probability standard for deeming a practice unsafe or unsound, whether alternatives (e.g., “reasonably foreseeable”) are preferable, whether a minimum likelihood percentage should be set, and whether a standard not implying forward-looking probability should be considered. The input shows a split: some support retaining “likely” with evidence-based guardrails, while many argue it is ambiguous or too high a bar that could miss low-probability/high-impact risks. Several commenters caution against numeric thresholds and favor severity-aware or foreseeability-based formulations. The core challenge is balancing legal clarity, supervisory flexibility, and tail-risk capture without rigid percentage tests.

Key takeaways:

• One respondent supports “likely” as the appropriate standard for unsafe or unsound practices.
• One respondent links “likely” to binding case law and calls for demonstrable, quantifiable evidence to support determinations.
• Multiple commenters describe “likely” as ambiguous and suggest “reasonably foreseeable” as clearer and more workable.
• Consumer-focused input emphasizes risks that are “reasonably foreseeable,” not only those that are “likely.”
• One respondent warns that setting a minimum probability threshold “invites delay,” while an FDIC filing notes agencies sought comment on whether to set a percentage.
• Respondents highlight tail-risk concerns: a low-probability but very large loss can be as dangerous as a higher-probability smaller loss.
• Several inputs note that expected risk depends on both likelihood and magnitude, and that “likely” sets a higher bar than “possible.”

Bottom line:

Inputs lean against relying solely on “likely,” favoring severity-aware and foreseeability-based standards without fixed percentage thresholds. If “likely” is retained, commenters urge clear, evidence-based guidance aligned with materiality and capable of capturing low-probability/high-impact risks.

The Question (Ref #5)

Is ‘‘likely’’ the appropriate standard to specify the probability of risk required for a practice, act, or failure to act, to be considered an unsafe or unsound practice? Is another term more appropriate, e.g., ‘‘reasonably foreseeable,’’ ‘‘could reasonably,’’ ‘‘imminent,’’ ‘‘abnormal probability’’? Should the agencies specify a minimum percentage of likelihood? If so, what would be an appropriate minimum percentage of likelihood? Should the agencies consider a standard that does not imply an assessment of a forward-
looking probability?

Direct Response to the Catalog Question

Introduction

Question 5 asks whether “likely” is the appropriate standard to specify the probability of risk for deeming a practice unsafe or unsound; whether terms like “reasonably foreseeable,” “could reasonably,” “imminent,” or “abnormal probability” fit better; whether a minimum likelihood percentage should be specified; and whether a standard not implying a forward-looking probability should be considered.

Historic Lessons in the Evidence

Respondents’ reasoning underscores that rigid probability screens can impede timely intervention, especially for tail risks. They argue that ambiguous terminology fosters inconsistency, while numeric thresholds can delay action. Conversely, tying determinations to demonstrable, quantifiable evidence and materiality grounds supervision and mitigates ambiguity, particularly when augmented by foreseeability and severity-focused analysis.

The Challenge

Operationalizing a probability standard in examinations is difficult: “likely” may be read as “very probable,” yet supervisors must capture significant but less probable harms. Fixed percentages risk false precision and delay, while undefined terms risk inconsistency. Agencies must reconcile “if continued” timing language, materiality, and the need to act on foreseeable undue risks without over-reliance on probabilistic estimates.

Evolving Metrics

Inputs emphasize evidence-based judgments and expected loss thinking, “expected risk is a function of the likelihood and magnitude of a risk”, and note that “likely” is a higher bar than “possible.” Some urge demonstrable, quantifiable evidence to support findings, while others call for foreseeability standards that do not hinge on explicit probability estimates.

A Framework Inspired by the Inputs

An implicit pattern emerges: anchor unsafe-or-unsound determinations in materiality; require demonstrable, quantifiable evidence; consider both likelihood and severity; avoid rigid percentage thresholds; and provide clear qualitative guidance on terms like “likely,” “reasonably foreseeable,” and “if continued,” ensuring tail risks are not excluded.

Case Study

Across the documents, one cluster supports “likely” when tied to case law and evidence (e.g., requiring demonstrable, quantifiable analysis), while another cluster warns that “likely” can miss low-probability/high-impact risks and prefers “reasonably foreseeable” or severity-aware approaches. Both clusters discourage fixed probability percentages, favoring qualitative clarity and materiality-based judgments.

Recommendations

1. If retaining “likely,” explicitly align it with materiality and require demonstrable, quantifiable evidence to support determinations.
2. Provide a complementary, foreseeability-based formulation (e.g., “reasonably foreseeable undue risk”) to capture emerging and tail risks not well captured by “likely.”
3. Do not set a fixed minimum probability percentage; inputs warn such thresholds invite delay and false precision.
4. Incorporate expected loss reasoning that weighs both likelihood and magnitude so low-probability/high-impact risks are actionable.
5. Clarify how “likely” relates to “could reasonably be expected” used in supervisory contexts to ensure consistent application.
6. Define the “if continued” clause to reduce timing ambiguity and provide practical examples of qualifying harms.
7. Emphasize qualitative guidance and documentation standards over numeric cutoffs to promote consistency and flexibility.
8. Periodically review application outcomes to ensure tail risks and foreseeable undue risks are captured without overbreadth.

Conclusion

On Question 5, the inputs do not converge on “likely” as the sole probability standard; many urge foreseeability, and severity-aware approaches that avoid rigid numeric thresholds. A pragmatic path is to retain “likely” only with clear, evidence-based guidance tied to materiality, while also using foreseeability language to catch low-probability/high-impact risks. Agencies should avoid minimum percentage requirements and permit standards that do not hinge on explicit probability estimates. This balances clarity, flexibility, and coverage of tail risks.

Follow us, stay informed, stay secure, and let’s navigate the risk landscape together.