A survey of human judgement and quantitative forecasting methods

Area of Science:

• Multidisciplinary research in forecasting, encompassing computer science, operations research, risk analysis, decision science, and psychology.

Background:

• Forecasting research spans human judgment, quantitative methods, and hybrid approaches.
• A significant increase in application-focused forecasting literature has been observed, with a rise in data-driven models.

Purpose of the Study:

• To provide a comprehensive overview of forecasting research across various academic domains.
• To establish common ground for future discussions by connecting disparate research areas.
• To identify key research thrusts and ethical considerations in human/machine forecasting.

Main Methods:

• Systematic literature survey using key search terms across multiple disciplines.
• Analysis of over 280 publications focusing on forecasting methods.
• Comparative review of human judgment, quantitative forecasting, and hybrid models.

Main Results:

• Quantitative, data-driven forecasting models show a substantial increase.
• Neither human judgment nor quantitative methods are universally superior; effectiveness depends on data characteristics.
• Human and algorithmic approaches can be complementary, leading to more accurate and resilient forecasting models.

Conclusions:

• Future research in human/machine forecasting must integrate model selection, human-model interaction, forecaster training, and forecast combination.
• Ethical considerations are crucial when combining quantitative models with human judgment.
• A holistic approach considering all research thrusts is necessary for advancing predictive performance.