Entrada em invalidez sob a perspectiva dos modelos de sobrevivência de longa duração

Abstract

The biometric tables for disability incidence commonly used in the Brazilian pension market often lack adherence to the specific demographic reality of certain pension regimes. This study aimed to model disability incidence estimates from the perspective of long-term survival models, considering the nature of the event as an occurrence experienced by only a portion of the population exposed to risk (cure fraction). The methodology consisted of applying these models to the experience of a specific timeframe from the insured database of the Federal Public Servants’ Social Security Regime (RPPS). The statistical performance obtained was compared with both the classic survival model and the Álvaro Vindas market table, which serves as the minimum criterion parameter for public pension systems. The results indicate that the standard mixture model demonstrated better adherence in the scenario that was methodologically most compatible with the cure fraction hypothesis. This approach outperformed both the Álvaro Vindas table and the limitations of the model without a cure fraction, which artificially forces the survival probability to zero at advanced ages. In conclusion, the adoption of long-term survival models provides a more reliable measurement of risks, mitigating distortions in actuarial valuations and contributing to the solvency of the pension system.