Probabilistic County-Level Population Projections.

Area of Science:

• Demography
• Statistical Modeling
• Population Studies

Background:

• Traditional population projections often lack uncertainty assessment.
• National probabilistic methods (e.g., UN) are not directly applicable to subnational levels due to unique data characteristics.
• Subnational projections require accounting for internal migration and specific populations (e.g., college students).

Purpose of the Study:

• To develop a novel Bayesian method for subnational population projections.
• To incorporate migration dynamics and special populations into subnational forecasts.
• To assess the accuracy and calibration of the proposed method against existing deterministic projections.

Main Methods:

• A modified Bayesian approach building upon the United Nations' national probabilistic method.
• Application to Washington State counties, including migration and college populations.
• Comparison with deterministic projections and out-of-sample validation.

Main Results:

• The proposed Bayesian method produces accurate and well-calibrated subnational population forecasts.
• Forecast intervals generated by the new method are generally narrower than traditional growth-based intervals.
• Improved uncertainty quantification for population predictions at the county level.

Conclusions:

• The developed Bayesian method offers a robust framework for subnational probabilistic population projections.
• This approach enhances the reliability of local population forecasts by quantifying uncertainty.
• The method provides valuable insights for regional planning and policy-making.