Processi Stocastici per modellizzare la mortalità giovanile

In this Thesis I apply double stochastic processes (Cox processes) in order to model the random human mortality at early ages. A Poisson counting process with mean and non-mean reverting stochastic intensity features have been used and calibrated to UK-Wales dataset with cohorts ranging from 1940 to 1976. This approach has been already used in literature but no one has applied to infant, childhood, adolescence and young adult data. I have found evidence that human mortality data from 0 to 35 years are better described and modelled with mean reverting stochastic processes. The second part of the thesis implement the mean-reverting process for forecasting purposes. The reverting process shows increasing precision with recent cohorts but experience an overestimation of the survival probabilities. This research fills the gap on a non-existent literature on early age stochastic mortality, suggesting a model for forecasting and eventually pricing insurance-linked products.