We investigate the ability of the Lee–Carter model to effectively estimate the gender
gap ratio (GGR), the ratio between themale death rates over the female ones, by using a
Cox–Ingersoll–Ross (CIR) process to provide a stochastic representation of the fitting
errors. The novelty consists in the fact that we use the parameters characterizing the
CIR process itself (long-term mean and volatility), in their intrinsic meanings, as
quantitative measures of the long-term fitting attitude of the Lee–Carter model and
synthetic indicators of the overall risk of this model. The analysis encompasses 25
European countries, to provide evidence-based indications about the goodness of fit
of the Lee–Carter model in describing the GGR evolution.We highlight some stylized
facts, namely systematic evidence about the fitting bias and the risk of the model
across ages and countries. Furthermore, we perform a functional cluster analysis,
allowing to capture similarities in the fitting performance of the Lee–Carter model
among countries.
