Markov switching count data models as an alternative to zero - inflated models of vehicle accident frequencies

In this study, two-state Markov switching count data models are proposed as an alternative to zero-inflated models, in order to account for preponderance of zeros typically observed in accident frequency data. Similar to zero-inflated models, two-state Markov switching models assume an existence of two states of roadway safety. One of the states is a zero-accident state, which is safe. The other state is an unsafe state, in which accident frequencies can be positive and are generated by some given counting process (Poisson or negative binomial). Contrary to zeroinflated models, Markov switching models explicitly consider switching by roadway entities (roadway segments) between the states over time. An important advantage of Markov switching models over zero-inflated models is that the former allow a direct statistical estimation of what states specific roadway segments are in, while the later do not. To demonstrate the applicability of the approach presented herein, a two-state Markov switching negative binomial model and standard zero-inflated negative binomial models are estimated using five-year accident frequencies on Indiana interstate highway segments. The Markov switching model result in a superior statistical fit relative to the zero-inflated models.