A binary time series model with a long-memory structure

Markov chains and Mixture Transition Distribution are the most traditional models for binary time series. More recently, Startz (2008) has introduced a novel model for binary data, the so-called Binomial AutoRegressive Moving Average model basically based on the comparison between theoretical and empirical autopersistence functions. However, some economic phenomena show a long memory structure not captured by any of these formulations. For quarterly U.S. binary data on recession, we show that a long-memory model for binary data can substantially improve the fit