On single point forecasts for fat-tailed variables.

Area of Science:

• Epidemiology
• Statistical Modeling
• Risk Management

Background:

• The COVID-19 pandemic exemplifies a multiplicative phenomenon with significant tail risks.
• Current naive empiricism and point forecasts are insufficient for managing such risks.
• Addresses critiques regarding scientific methodology during the pandemic.

Purpose of the Study:

• To highlight common errors in applying naive evidence-based methods to fat-tailed variables.
• To demonstrate the inadequacy of first-order scientific methods for tail risk management.
• To propose policy implications derived from statistical properties of multiplicative phenomena.

Main Methods:

• Analysis of common fallacies in statistical inference for fat-tailed distributions.
• Case study utilizing the COVID-19 pandemic as an example.
• Critique of naive empiricism and point forecasting in risk assessment.

Main Results:

• Naive "evidence-based" empiricism and point forecasts lead to significant errors with fat-tailed variables.
• First-order scientific methods are insufficient for effective tail risk management.
• Mitigating policies for multiplicative phenomena must align with their statistical properties and risks.

Conclusions:

• Advanced statistical approaches are crucial for managing tail risks associated with phenomena like pandemics.
• Policy decisions must be informed by a deeper understanding of statistical properties and associated uncertainties.
• The study underscores the need for more sophisticated methodologies beyond naive empiricism.