Risk and uncertainty are often used interchangeably for as the cone of uncertainty increases the further out into the future we proceed, so does the probability of a negative expected outcome. However, risk is merely a subset of uncertainty for uncertainty includes both positive and negative outcomes.
That being said, many of the risk models that are utilized employ a range of assumptions to create stationarity whereby statistical and Bayesian inferences are applied. Yet, in many instances, the assumptions are intractable while non-stationarity reigns supreme in our complex adaptive system. The models are typically not recording events that are 20 standard deviations away or greater, rather they are unknown, unknowns or known, unknowns which could not be captured by the underlying specifities of the models themselves.
When dealing with uncertainty, risk management is paramount, specifically the management of downside risks. In managing uncertainty, it is important to build resiliency, robustness, and optionality within the contingency/scenario planning stages. Additionally, it is beneficial to “fit models to problems, not problems into a model.”
As a simple example and thought experiment, when dealing with uncertainty, it is important to understand the present situation as thoroughly as possible. From there, look for similar references to form a base case of the future downside expectation. From there, create a decision tree matrix of additional potential downside outcomes. As new information is provided, you compare this with the fit of the base case model to see where similarities and differences exist. As more information is received, it is compared to the base case to find similarities and differences. Greater divergences will lead to the probability that an alternative downside outcome exists which needs to be checked against the potential negative outcomes and the possibility for unknown, unknowns as new information is received. Conversely, similarities to the base case leads to refinement of the base case expectations. The iterations continue indefinitely and the process is refined. Ultimately, more buffers need to be created with greater uncertainty, greater risk, decreases in transparency, and increases in entropy.
Please refer to the following analyses: Part 1: Algorithms, Machine Learning, and Future of Predictive Analytics https://internationalcapitalmarkets.org/2020/08/19/%f0%9f%94%ba-part-1-algorithms-machine-learning-and-future-of-predictive-analytics/ via @Diamond1_CEO
Why of Course! However, there is a Catch……..22 https://internationalcapitalmarkets.org/2019/09/16/why-of-course-however-there-is-a-catch-22/ via @Diamond1_CEO