Forecast accuracy hardly improves with method complexity when completing cohort fertility.

Area of Science:

• Demography
• Social Sciences
• Population Studies

Background:

• Completed fertility forecasts are vital for public policy and social science research.
• The optimal method for forecasting completed fertility remains unknown.
• A wide array of forecasting methods exist, varying in complexity and theoretical grounding.

Purpose of the Study:

• To evaluate and compare the accuracy of various completed fertility forecasting methods.
• To identify methods that provide reliable uncertainty estimates.
• To guide the selection of appropriate forecasting techniques for demographic analysis.

Main Methods:

• Identified and categorized 20 major completed fertility forecasting methods (162 variants).
• Assessed method accuracy and uncertainty using data from 1,096 birth cohorts across 29 countries (Human Fertility Database).
• Validated findings using data from 201 United Nations countries and six world regions.

Main Results:

• Only four methods consistently outperformed the naive 'freeze rates' method in accuracy.
• Just two methods produced reliable, non-downward biased uncertainty estimates.
• Simple extrapolation methods and Bayesian methods were among the top performers, with Bayesian methods being data and computationally intensive.

Conclusions:

• Simple extrapolation methods offer a competitive alternative to more complex Bayesian approaches for fertility forecasting.
• Careful selection of forecasting methods is essential for accurate demographic projections and policy-making.
• Further research may focus on refining methods that balance accuracy with data and computational demands.