Resumen
In this paper, we propose a semi-Markov chain to model the salary levels of participants in a pension scheme. The aim of the models is to understand the evolution in time of the salary of active workers in order to implement it in the construction of the actuarial technical balance sheet. It is worth mentioning that the level of the contributions in a pension scheme is directly proportional to the incomes of the active workers; in almost all cases, it is a percentage of the worker?s incomes. As a consequence, an adequate modeling of the salary evolution is essential for the determination of the contributions paid to the fund and thus for the determination of the fund?s sustainability, especially currently, when all jobs and salaries are subject to changes due to digitalization, ICT, innovation, etc. The model is applied to a large dataset of a real compulsory Italian pension scheme of the first pillar. The semi-Markovian hypothesis is tested, and the advantages with respect to Markov chain models are assessed.