Comparative Analysis of Classical and Time-Varying Lee–Carter Models for Mortality Forecasting in Indonesia and Japan

Syipa Arista Widya (Department of Mathematics, Universitas Islam Negeri Sumatera Utara, Medan, Indonesia)
Ismail Husein (Department of Mathematics, Universitas Islam Negeri Sumatera Utara, Medan, Indonesia)

Zero : Jurnal Sains, Matematika, dan Terapan
Vol 10, No 2 (2026): Zero: Jurnal Sains Matematika dan Terapan

Reliable mortality forecasting is essential for demographic planning and actuarial decision-making. This study compared the forecasting performance of the Classical and Time-Varying Lee–Carter models using World Population Prospects (WPP) 2024 mortality data for Indonesia and Japan (1970–2023). Both models were estimated using 1970–2015 data and evaluated on 2016–2023 through out-of-sample validation based on MAE, RMSE, and MAPE. For Indonesia, the Time-Varying model yielded a slightly lower MAPE (7.262%) than the Classical model (8.086%), whereas the Classical model produced slightly lower forecasting errors for Japan. Overall, the differences between the models were modest. The findings suggest that, in this study, forecasting performance was associated more closely with population-specific mortality characteristics than with model complexity alone.