Stochastic population forecasting based on combinations of expert evaluations within the Bayesian paradigm

Area of Science:

• Demography
• Statistical Modeling
• Population Studies

Background:

• Traditional population forecasts often assume independence between demographic components (fertility, mortality, migration).
• Incorporating expert opinions into forecasting models can enhance accuracy but requires methods to handle inter-expert correlations and component dependencies.

Purpose of the Study:

• To develop a procedure for stochastic population forecasts using an expert-based approach.
• To allow for dependence among demographic components and account for correlations among experts.
• To indirectly derive the dependence structure from expert-provided scenarios.

Main Methods:

• Employs an expert-based approach where expert evaluations serve as data within a Supra-Bayesian framework.
• Utilizes a mixture model to handle dependencies between demographic components and correlations among experts.
• A Markov chain Monte Carlo (MCMC) algorithm is used to approximate the posterior distribution for demographic indicators.

Main Results:

• The proposed method allows for the derivation of dependence structures from expert opinions, rather than imposing them.
• The derived posterior distribution serves as the forecasting distribution.
• An application demonstrates the method's utility in forecasting the Italian population from 2010 to 2065.

Conclusions:

• The Supra-Bayesian approach provides a flexible framework for stochastic population forecasting by integrating expert knowledge.
• The method effectively accounts for complex dependencies in demographic processes and expert judgments.
• This procedure offers a robust tool for generating more realistic and informative population projections.