Overview
Bruce D. Spencer is a statistician at Northwestern University who has been a central collaborator with Juha M. Alho in developing the probabilistic framework for assessing official U.S. population forecast accuracy. His core contribution is the approximately linear propagation-of-error model (Alho and Spencer 1985) that formalizes how errors in fertility, mortality, and migration assumptions accumulate through cohort-component population projections, and showed that USBC high-low variant intervals are systematically too narrow for a 67% probabilistic interpretation. He and Alho extended this framework through a series of papers in the late 1980s–1990s covering mortality error models, population databases as stochastic objects, and the magnitude of error from different vital processes.
Key Contributions / Features
- Alho and Spencer (1985) — "Uncertain Population Forecasting," Journal of the American Statistical Association 80(390): 306–314; introduces the approximately linear model V(t+1)=R(t)V(t); three error sources (misspecification, parameter, random); mixed estimation; fertility dominates; USBC intervals too narrow for 67% interpretation. See Alho and Spencer 1985 — Uncertain Population Forecasting.
- Alho and Spencer (1990) — "Error Models for Official Mortality Forecasts," JASA 85(411): 609–616; parallel error-bounding methodology applied to SSA mortality forecasts; establishes that SSA mortality intervals systematically understate uncertainty.
- Alho and Spencer (1991) — "A Population Forecast as a Database: Implementing the Stochastic Propagation of Error," Journal of Official Statistics 7(3): 295–310; extends the approximately linear framework to full stochastic cohort-component implementation.
Related
Sources