A multivariate approach to project the long run relationship between mortality indices for Canadian Provinces

The cointegration approach is proposed to model cross-province mortality indices within Canada. We apply and compare the vector autoregressive model (VAR) and the vector of error correction model (VECM) derived from cointegrated models for males and females. Relying on the Johansen cointegration test, the analysis shows clearly that there is a dependence among provincial mortality indices. The two models fit well the females data. However, poor performance has been revealed for men beyond 10 years horizons. We project the mortality indices from both models and compute the annuity from the forecasts. We project the mortality indices from both models and compute the annuity from the forecasts.