Matrix-Variate Hidden Markov Models: An Application to Employment Data

Hidden Markov models (HMMs) are a powerful tool for modeling heterogeneous time-dependent data structures. Recently, a family of parsimonious matrix-variate HMMs has been introduced in the literature. In this manuscript, we apply this family of models to a dataset measuring employment levels for 187 countries in the world over 26 years. Each country is evaluated using the well-known three-sector model of the economy and by taking into account the gender of the workers. By using the HMMs flexibility, different states are detected and the switches of the countries among the states over time are analyzed.