The reduction of the number of parameters in high-order Markov chain already inspired several articles. In particular, Raftery proposed an autoregressive modelling which utilizes the same transition matrix, with a coefficient, for every lag. In this paper, we show that a model of the same type,...