This paper is devoted to an application of hidden Markov models to historical data. Empirical observations showed that periods of conflict are characterized by an important increase of the norm. A segmentation of the data should bring some elements of answer to the following questions : are there synchronous periods when comparing the normative production with the situations of war? Or are there discordances between the two that can inform us on the State politics? The data we had to deal with was very specific and we propose two new models to analyze it. In the first one, we introduce a zero-inflated Poisson distribution in the context of hidden Markov models. The second model generalizes integer-valued autoregressive processes by adding a hidden Markov chain.