Probabilistic mortality forecasting with varying age-specific survival improvements.

Area of Science:

• Demography
• Actuarial Science
• Biostatistics

Background:

• Traditional mortality forecasting often fails to predict turning points or age-shifts in mortality decline.
• Extrapolating past trends provides precise forecasts only under stable mortality conditions.

Purpose of the Study:

• To develop an advanced mortality forecasting model that incorporates mortality dynamics and turning points.
• To provide accurate and reliable life expectancy forecasts with quantified uncertainty.

Main Methods:

• A novel model combining mortality improvement rates and multi-country mortality trend analysis.
• Simulation-based Bayesian inference for model estimation and prediction interval generation.
• Validation against established models like Lee-Carter and UN Bayesian approach using historical data.

Main Results:

• The proposed model accurately forecasts both regular and irregular mortality developments.
• Performance is comparable to or better than existing leading mortality forecasting models.
• Prospective forecasts indicate gradual increases in life expectancy for British and Danish women from 2012 to 2050.

Conclusions:

• The new model offers a robust framework for mortality forecasting, adept at handling complex mortality dynamics.
• It provides valuable insights into future life expectancy trends with reliable uncertainty quantification.
• This approach enhances the precision of long-term demographic and actuarial projections.