Extrapolation beyond observed data

Background: Forecasting highly advanced populations requires extrapolating dynamics into mortality regimes that are sparsely represented or entirely absent from historical training data sets like the HMD.

Question / Future Work: Since the empirical training data does not include the super-low mortality levels that some populations may reach in the future, there is no data-driven trajectory path into or through those very low mortality regimes; the current approach of extrapolating within-cloud behavior to edge points may not adequately reflect true future dynamics, representing an open question for the theory of mortality change.