Amidst the intricate dance of market fluctuations and industry disruptions, business leaders often find themselves standing at the crossroads of decision-making, surrounded by charts, data, and projections meticulously crafted by expert analysts, armed with risk models designed to mitigate uncertainty and safeguard their organizations against potential dangers. 

The recent turbulence in global markets and the unforeseen emergence of disruptive technologies have shattered long-held assumptions, leaving traditional risk models struggling to keep pace with the complexities of the present-day economic landscape. In this introspection, business leaders recognize the pressing need to advance their risk management strategies to embrace radical uncertainty, a concept in decision theory and economics that points out the inherent limitations of predicting future outcomes or events with certainty, particularly in complex and uncertain environments. It implies that certain scenarios are inherently unpredictable, and complete knowledge or data about the past and present may not be sufficient to accurately forecast the future.

Unlike situations of risk, where probabilities can be assigned to different outcomes based on historical data and statistical analysis, radical uncertainty acknowledges the presence of “unknown unknowns.” These unknowns make it impossible to create reliable probabilistic models for informed decision-making, thereby highlighting the limitations of relying solely on historical data and statistical models to make predictions about the future.

Does this mean accepting radical uncertainty “as is” and letting it be the default excuse for failed risk management? As a model validator, this might lead to abuse of the concept by overly relying on historical data, neglecting tail risks, conducting insufficient analyses, and reviewing and testing risk models, or simply doing nothing.

Proponents of radical uncertainty propose approaches that balance radical uncertainty with model robustness, consistency, and adaptability instead of just aiming for precision in predicting the future.

Scenario analysis

Through the collaboration of different relevant units of the organization, plausible future scenarios that capture various economic conditions and potential disruptions in different portfolios can be created. By assessing the impact of a range of these scenarios, risk management teams can gain insights into the resilience of their risk models and develop strategies to mitigate risks under different conditions.

Sensitivity analysis

Different from scenario analysis, sensitivity analysis facilitates the identification of key inputs and assumptions that have the most significant impact on the risk model’s output or performance and provides a clearer understanding of the sources of uncertainty piece by piece. By understanding which variables are most sensitive, management can focus on gathering more relevant data or improving the modelling approach in those areas. 

Stress testing

Management can subject risk models to stress testing to evaluate their robustness and performance under extreme or adverse yet plausible conditions. This involves simulating scenarios that go beyond historical data, such as severe economic downturns or unexpected events, to assess the model’s sensitivity to different stress factors. Stress testing can help identify vulnerabilities and weaknesses in the model and inform adjustments or improvements to enhance its ability to handle radical uncertainty.

Expert judgment and deliberate ignorance

Seeking insights and opinions from subject matter experts who possess domain knowledge and experience can help identify potential risks and provide qualitative assessments that complement quantitative models. Additionally, accepting “deliberate ignorance” by acknowledging what is not known and explicitly including uncertainties in the model can provide a more realistic representation of radical uncertainty. It is always better that the limitations or “unknown unknowns” are documented in the organization’s risk framework for enhanced risk transparency and awareness and improved model development, aside from just compliance and audit/review purposes.

Bayesian updating

Bayesian updating is an iterative process that involves updating beliefs and probabilities based on new information. Management can apply this approach to risk models by continuously monitoring data and updating model assumptions as new and significant information becomes available. This allows for the incorporation of evolving insights and the adaptation of models to changing conditions, helping to capture the dynamic nature of radical uncertainty.

Regular model validation and review

An institutionalized process for regular model validation and review ensures the reasonable effectiveness of risk models in the face of radical uncertainty. Validation and review include comparing model predictions with actual outcomes and evaluating the model’s performance over time. If significant discrepancies or weaknesses are identified, adjustments can be made to enhance the model’s ability to account for uncertainties.

Embracing radical uncertainty does not mean relinquishing control but rather challenging business leaders to maintain a delicate balance among resilience, adaptability, and innovation as strategic anchors in uncharted territories. In this way, embracing uncertainty becomes a strategic advantage rather than a liability in safeguarding the future endeavors of their organization.


As published in The Manila Times, dated 26 July 2023