Abstract
The regime switching model developed by Hamilton is applied to winter traffic data and the relationship between the road surface condition and the traffic speed is analyzed. In the regime switching model, the structure of a time series is assumed to be subject to change in accordance with an unobservable Markov chain. The parameters that specify the structure in each state and transition probabilities of the Markov chain can be estimated by the maximum likelihood method. A regime switching model with two states is applied to traffic speed data collected on a national highway in winter and it is shown that each state of the supposed Markov chain coincides quite well with the good and bad road surface conditions.
